As a reminder: We MANUALLY TUNE the DDR4 kit for all of our reviews because we have controlled every timing for that kit for CPU reviews for years. This means that it outperforms any XMP 3200 kit on the market because we have hand-tuned it (it works on multiple generations of CPU, too). This kit outperforms our 3600 kits and 3800 kits because of the manual timing tuning done. So no, it is not as simple as just saying "you should use a higher frequency kit," because it ignores that just doing XMP doesn't hand-tune timings (like we do on the 3200 kit). We reviewed the i7-12700K (and KF) here: ua-cam.com/video/B14h25fKMpY/v-deo.html Here's our Windows 11 vs. Windows 10 CPU comparison on Alder Lake: ua-cam.com/video/XBFTSej-yIs/v-deo.html We also reviewed the 12600K here: ua-cam.com/video/OkHMh8sUSuM/v-deo.html We kicked off our Alder Lake reviews with the i9-12900K here: ua-cam.com/video/fhI9tLOg-6I/v-deo.html We covered Alder Lake’s architecture here: ua-cam.com/video/htCvo9XJZDc/v-deo.html
Can you please post the full timings you use for reviews? Like an Asrock Timing Configurator screenshot? (Handy program that pulls up all your timings from Windows if you're not familiar with it). I think including these full timings would be an excellent addition to any memory related video. Thanks!
@@GamersNexus one thing I noticed in this video is that you could see your shadow on the right of the screen, a bit more often than I remember seeing in your old studio, although that could be because your new set looks so naked, so it's more noticeable... I'll bet you're going to update your lighting at some point, which should help immensely, if you have the required space and height for it to be workable, not that it really takes away from the content, because it doesn't, but it just feels so unlike you to have a rogue shadow in the frame, as you are so, uh, perfectionist, which is a good thing... But maybe I just never noticed your shadow and the lights in your old studio were also bad for shadows... Either way, that's still a pimpin studio/testing facility you have, and I can't wait to see a video on UA-cam, of you touring the place, once it's completely done being remodeled and whatever else your current plans are...
@@northwiebesick7136 That's just the ghost of videos past. When you turn your head and see it in your room I'd be concerned. It might be looking at you now. 👀
6:42 "In this scenario we observed DDR5 producing the best result with a 7.1% advantage over the DDR5 kit" 🤣 Steve get some rest, moving into this cool new background must be exhausting
With current DDR5 prices, DDR4 may be the clear winner right now if you are going for 32GB. With the money you save from DDR4 you could go to a higher tier CPU or GPU, for better performance overall.
I dont mind paying the extra for DDR5, prob is i cant. Asus the retards prevents use of most air coolers from their mid range boards (the strix models below Hero). Its just stupid cosmetic size of the garbage around the VRMs. Theres only 1 other ITX mobo and its a DDR4 model from Gigabyte. While AIOs coolers are all the rage now, theres still a significant portion of market on air. Asus FFS! Way to lose $
@@user-lh1ef1st9k Its a HTPC that will fit in a floating TV wall unit. Limited space. Im setting up with a 75" tv, hopefully i can take all my pc use to the lounge room. I worry about font size with certain games though and i may need to have 2 good pcs. Ive thought about remote control of the pc from another room, but cant work out the cabling required. (Video to TV, sound to soundbar, wireless Mouse + keyboard control - will wall block? if USB extenders with the receivers on end then ill need some type of powered repeating hubs, the hdmi cable needs power of very high price fibre core
I like how recently you've started adding a quick summary at the beginning of the video. It doesn't stop me from watching the video, on the contrary -- it makes me eager to hear the details. Good move!
I know this wasn't something one should genuinely complain about, but I appreciate it so much that the blue tool panel is no longer catching reflections.
haha, thanks. We've been working on that each shoot. Andrew has been tweaking lighting. Our next one gets rid of the giant hard shadow cast by me when stepping back.
Yeah, the sharp shadow is pretty distracting indeed... Maybe a layer of some matte finish for the panel could help with the gloss? With less stuff catching reflections, there should be more freedom in setting up lights.
Factorio is a game known to be bottlenecked by both memory speed and latency (mainly latency), and as it have ways to reliably benchmark it, testing it in this scenario would have been interesting to see the results.
And hardware unboxed also showed they know nothing about memory when discussing the results. It's shameful to pretend to be an authority rather than admit to yourself you need to learn and grow.
I play modded Factorio with a Ryzen 7 and a very old R9 390 video card. It does cause performance issues I get frame rates dropping to 55 FPS with my 2 year old 5000 science pack/minute megabase.
@@MichaelClark-uw7ex I did play it modded as well, and noticed when my bios disabled my XMP profile as performance tanked. My i7-4770k with 16gb of DDR3-2400 C10 is noticiably faster than when running stock at 1333mhz. As long as you have suffient CPU power (it doesn't need much) the memory is the actual limitation in it.
I would really like to see how alderlake is doing in MMORPGS like, guild wars 2, black desert and so on,... Its a pretty big playerbase that allways get left out and I wasnt able to find benchmarks for any mmo. Keep up the good work! Love your content!
the fact you have to explain the "fairness" of testing a DDR5 equipped/enabled chip with DDR5 is sad. Thanks for the work you guys put in to informing consumers.
Yeah, I find that criticism odd. DDR5 is available, use it. I have always loved AMD, but when I hear "should have used the same ram in both" just seems like wanting to pad numbers in AMD's favor. These are generational changes and advancements, I like seeing them used when they come out, even if it's not the best deal.
It's not really the right type of empirical testing if multiple changes are made. That's not how empirical testing works. You change one thing to comment on that one thing. If you must/have to change multiple things, you thus change the probabilities of the results significance to which part made the difference. Same with JayZ2cents "empirically" testing AMD vs Intel power drawn, then putting a massive watercooler, RGB fans, HDDs etc all on the AMD only system, then running the Intel system barebones... That's not empirical or sensible testing. Here, Intel is moving to DDR5 too. *That's fine, and GamersNexus is absolutely allowed to do the testing, and it's great they are*. It just changes the statement from "this cpu is faster" to "this platform is faster" until we then test DDR4. If Intel released this CPU with only DDR5 options. Then yes, we would say "This CPU and platform is faster" and being human, we would accept the assumed shorthand of "This CPU is faster" because we know we are not talking about it in isolation. But as Intel have allowed DDR4 platforms for the 12gen, it's not possible to talk about it in isolation and only for DDR5 performance, we literally have to include it when making statements, unless saying "only on DDR5". :)
Tradeoffs in memory throughput vs. latency: Compression benefits greatly from throughput over latency because the task batches large segments of memory together. Things like high fps gaming or games with tens of thousands of entities need to shuffle data around more to keep the CPU execution fed, so they miss the cache more and read in smaller chunks from memory when they do. Reducing memory latency reduces the length of the stalls, so in these applications memory latency is more valuable. Additionally, in transitions to DDR3 and DDR4 we have seen higher latency, higher throughput kits than the previous generation for the first year. They eventually settle back around to the same "best latency SDRAM can muster" but with higher throughput.
I would have preferred a big fat disclaimer that the data is not representative of actual performance in most cases at the beginning of the video. Manually tuned B-Die in gear 1 vs not even the best DDR5 kit on the market not manually tuned in gear 2 doesn‘t even come close to being a fair comparison if you really wanna know which platform is capable of performing better overall. This methodology really only makes the testing valuable for the extreme cases where there is a big uplift from the additional bandwidth and maybe clock speed where timings don‘t matter. Still thankful for the testing but I‘m much more looking forward for the future data once you get your hands on one of those 5600C36 D5 kits and got the hang of how to manually tune those. I don‘t know about elsewhere in the world but the pricing of B-Die D4 kits from G-Skill Trident Z vs the best Z5 variants are not that far apart here in Germany so it really is a fair comparison with maxed out 32GB kits.
Thanks for this. The reason I wanted to know about DDR 4v5 was to know how much of the ~25% gains Alder Lake offered were from the RAM vs the CPU itself. Looks like most of it really is the silicon. Good to know!
@Vincent Tuminello can you give me an update on this? I fully upgraded my pc for call of duty warzone, and am completely disheartened today I could notice input latency from the 4800 DDR5 Ram I have, I was getting destroyed compared to sadly my 3200 mhz off brand crap I got with a pre built pc a year ago... I have DDR4 g skill trident 4000 mhz CL14 ram in the box in my closet I was going to return, but i'm strongly considering attempting to install these and seeing if my luck changes, I was getting beat in situations I shouldn't have every single time basically... do you think the latency of DDR5 in its current new state is not good compared to ddr4 for 1st person shooters? I have a z690 motherboard as well
@@brutyfruity unless you are running the game at super high framerates RAM won't have any impact. Try it with overwatch at 400fps cap and you will see massive gains in fps and 0.1 and 1% lows as well as frametime consistency. Overwatch is one of the few games that are massive ram bandwidth hogs if you try playing them at very high framerates. Lower framerates make ram mostly irrelevant. I highly doubt you are bottlenecked by ram at all in warzone.
Could be interesting to see the impact of DDR4/5 when using integrated graphics. Would be more relevant with laptops generally but could therefore make RAM choice more significant since that is a component manufacturers will inevitably use to cut costs or tout the presence of DDR5 for marketing purposes even if it might not make a notable difference for CPU as shown but graphics is not known yet.
This is what happens every generation of DDR, the established standard being more mature has already been optimized for latency and had the clock rates pushed to the edge, while the newer generation usually just focuses on higher clock rates for higher raw bandwidth, so for the first period on the market and as the standard matures you'll get better clocked kits with relatively lower latency like when DDR3 first hit the market CAS 5 1066mhz DDR2 was common in the high end, but DDR3 was largely only available in JEDEC standard timings so like CAS 7 for 1066 and CAS 9 for 1333 but as it matured you could get CAS6 1600 kits and cas 7 2000 kits. So some applications highly benefit from high raw bandwidth where some are more negatively impacted by the relatively high latency.
This is what has happened before, however it is not inherent or guaranteed to be this way always. Results in the past cannot predict the future in this case. There is nothing inherent in memory design restrictions, that requires this.
@@zazethe6553 there is some degree of predictability because "DDR5" is merely a JEDEC standard. Just like PCIe 5.0 is merely a standard. Predictably, these DDR standards' primary focus is on doubling bandwidth (this can be achieved in various ways). I don't know what the standard for DDR6 is, or if it has even been decided yet, but I can confidently predict a doubling of bandwidth over the DDR5 standard. The secondary focus is lowering power consumption. That one is less quantifiably predictable.
I'm over here still running ddr3 mhz 1800.. nice to see ddr5 come out and hopefully drive ddr4 prices down. Thats when i swoop in and pick up a new build
I know it doesn't apply here all that much in this round of tests in particular, but what's the difference in igp performance? That relies on the ram heavily. More a curiosity really, but if there is a significant uplift for some reason, it may help people wait when dedicated GPUs are more available.
I installed windows 11 a few days ago and had a big drop in game performance..Checked online and lots of other people seem to be having this problem too..Dont know if you do software stuff but thought it would make a good video..Thanks for all your hard work.
Interesting Benchmark idea: "will the E-cores save your FPS?" the e-cores is great so some random messenger or discord chat won't kill your FPS but how much of a performance hit will it save? i7-12700K, with only 4 P-cores enabled, vs 4 P-cores and 4 E-cores. while we're at it, maybe vs 6 P-cores and 8 P-cores scenario 1: sanitized windows install (no BG tasks) scenario 2: bloated windows install (4GB worth of bg tasks with combination of chrome tabs, RGB software, anti-virus, and an active discord voice call)
The 12900k is like the 5950x. No you don't need 16 cores. No 16 cores may not even help over 12 cores. But what you are getting is better binned CPUs. More total capacity for work. Better silicon. And you have reached the top component for the board you are on. Do you really care that your CPU blender render took 16 minutes instead of 20? No probably not. Do you care that you got 165fps instead of 162? Or that 0.1% lows were 10fps higher? No probably not. A 12 core and a 16 core will have very simmilar frame time graphs aswell. But when you compare the cost of a whole system. $2000 for a 5900x system. Vs $2250 for a 5950x system. Same for intel. Do you save the 12% and lose 25% of your computing power? If you don't do loads of professional work. And you play older games, or you plan on buying newer games a couple of years late when they are on sale. Or you plan to upgrade in 2-3 years. Then probably yes. If you plan on playing new games right when they come out. If you plant on multitasking like a madman (I do this). Or if you just want your CPU to last through multiple GPU upgrades. Go with the 12900k or the 5950x. Both are great. Neither really out performs their lower end counterpart today. But they may in the future.
@TheHammerGuy - Intel has turned to E-cores because they failed to get anymore than 8 power cores on their 10 nanometre node. So they made a Frakenstein's monster of a CPU. For the same price, instead of an 8p/8e or 6p/4e you can get a 16p-core or an 12p-core processor. Waaaay more convenient when you consider the price, and the components you'd have to upgrade and their cost as well as needing a new power supply. As you only need about 140 Watts for a Zen 3 processor.
@@enriquepadilla2542 This just isn't good advice during a performance war. If you have to decide if you really need an i9, you don't, and you would be better off buying a new i5 or i7 again sooner.
@@enriquepadilla2542 I would love to undervolt a 12900k, but not going to pay $1000. 5950 is a better value, if you will keep it busy all day. 12700k looks good to me.
@@Adreno23421 the 4000mhz cl14 is limited to 8gb per channel currently, so 16gb for 180 currently, would run you 360 to get 32gb of it.... pretty pricy , but have not seen where it provides performance boost, have any reference?
Small studio critique: That light, that is to the left of the camera, makes you cast an irritating shadow on the wall when you move back from the table. You should change something to avoid that. 😉
Awesome update. I wasn't concerned with your testing methodology (you explained it all and it made sense for the choices you made), but I was very interested from a consumer point of view if there were downsides to going one direction vs the others (DDR4 vs DDR5, Win 10 vs Win 11) that would make it a bad idea to even consider that. You've covered all those and now its very much down to what you intend to do now/in the future as opposed to it being potentially crippled and losing out. Great stuff.
When you guys do your efficiency testing I'm very curious about where things sit in perf if you restrict the power levels down to some baseline. I've seen a lot of speculation about power efficiency based on the 12900k but the reality is none of these chips are running anywhere near to their most efficient levels up near 5Ghz; they are all spending tons of additional power for the last X% of performance just to be at the top of the benchmark chart. I'm quite curious what that X is, and what that curve looks like between Alder Lake and Zen.
Would like to see something similar to Der8auer's recent video where you assign only E-Cores or only P-Cores to a program and benchmark E-Core VS P-Core: maybe max overclocking (% over listed spec), 4GHz vs 4GHz, 8 e-cores vs 4 p-cores (with and without SMT), etc...
that's an interesting benchmark r7 5800x sanitized test bench vs r7 5800x with 2GB worth of Chrome tabs. and i7 12700K sanitized test bench vs i7 12700K with same parameters as above. "ARE THE E-CORES SAVING YOUR FPS?" I've told some people that the great thing about the hybrid design helps to alleviate BG tasks so some random discord chat wont kill your FPS.
@@TheHammerGuy94 I don't know if that's the best way to do that test, but we'd mostly end up looking at the 1% lows. Methodology aside, I am hyped for cpus in the future which will have like 16+64 cores because background tasks will never bother me ever again... As a massive multitasker that likes to leave things open, this is huge.
Quick tip since I noticed the can of CRC contact cleaner on the shelf- that stuff is magic for removing thermal paste. Dissolved the paste out of an LGA socket on a Craigslist motherboard, and got a thermal paste stain out the carpet!
Have you thought about adding game engine build tests for productivity testing? For example, building the UE5 Valley of the Ancient sample? From the recent SIGGRAPH talk, it's doing a lot of heavy composition when building the nanite meshes (graph partitioning, and variable length compression). Additionally, given the working set size won't fit in the LLC (one of the examples was up to 26GB -> 6GB after compression), it will likely benefit from the extra memory channels (more random addresses per second) more so than bandwidth.
@@Cd5ssmffan UE is used for things other than game development... I know people who use it to do 3D animations for independent movies, and people who use it for concept development in other media companies (including some XF studios). I have many colleagues who use it to generate AI training datasets, one of such projects trains traffic cameras to reduce traffic jams and red lights. I also know people who have used it for IC visualization before sending them to the Fabs and as a demo to raise capital to send a chip for fabrication. And I have used it in the past for simulation visualizations (things like fluid flow for power stations). The latter two cases wouldn't rely on things like Naninte, but do rely on engine building from source. But the media and AI training examples could rely on it.
The new set and lighting look good, Steve! It's easier to see you and the overall picture is clearer. Gamers' Nexus videos always had a slight blurriness to them before.
Probably the biggest application that could bring tangible benefit from switching to DDR5 over DDR4, that I can think of, is ZFS ARC caching to RAM on a NAS, especially if you have compression or dedupe on.
I enjoyed the comparison for the ddr4 to the ddr5 with the Alder Lake chipsets. I hope you continue this in-depth analysis of these new technologies, especially as they begin to mature and manufacturing processes become more streamlined for ddr5.
Yeah, sometimes I'm a bit sad Gamers Nexus doesn't talk in detail about laptops / iGPUs. Intel has come such a long way with its latest Xe graphics that it's totally viable for a low end gaming machine, especially with the current GPU situation.
As much as I enjoyed watching this, I really look forward to seeing the second generation of ddr5 and its performance. Early adoption can be rough. It'll be cool to see what it looks like when they honed the process a little more.
Yep. With WinZip having such a high performance uplift it also shows possibly some optimisations might be possible now that the memory bit width is different. Possibly the likes of GTAV is optimised for DDR4 on PC, thus looses performance in DDR5 setups. Will be interesting to see what happens.
Basically now if you buy DDR4 3600 cl16 which has pretty good price you already beat or are on par with most of currently available ddr5. And all the ddr5 are more expensive. It's just scalped disgustingly.
Yep that's specifically why I skipped DDR5 right now even when buying a 12600K. Ended up going for a cheaper DDR4 board instead and will probably just flip that in a couple years when better DDR5 modules come out.
@@damara2268 close to true. 3600c16 is pretty bad though. 4000c14 is pretty much maxing out gear1 low latency on a ddr4/Alder Lake build. If the ddr5 is the GOOD stuff (6200c34 kit from TGroup or g.skill 6800c38) then it matches/beats it. So, it has to be really fast to beat it, but, it does beat it at the highest end of gen 1 spec of ddr5.
This is really usefull to know, because one of my friends is getting a new computer and at least at the moment in Australia DDR5 cost x4 ($800 for 32GB) more then DDR4 (~$200 for 32GB)which is quite rediculous for what might only be 3-5fps more.
I have a feeling DDR5 with lower latency and Gear 1 mode will be very good. Or maybe DDR5 with faster frequency, lower latency, and still in Gear 2 mode.
Der8auer tested several kits of ddr4 and was running anything above 3600 in gear 2 mode, so i assume that's about as fast as you can go in gear 1. Faster ddr5 may also be too much for Alder lake's imc and require gear 4.
Please consider testing the iGPU performance difference between DDR4 and DDR5. I expect to see a decent improvement there but it would be good to know if we should be looking specifically for ddr5 on our next laptop or APU desktop purchase. Also it would be nice to see how different DDR5 clock speeds influence the iGPU performance
I mean.. since you asked.. I’d love to watch videos comparing the 4790k to the 12700k/12900k for gaming and streaming. Also, thank you for helping me learn something new every day!
Good to know that there is not much of a difference between ddr4 and ddr5 for just standard CPU operations. I would be interested seeing how ddr5 would compare to ddr4 for the IGP.
I would appreciate exploring concurrent tasks on the new Alder Lake architecture, even if the testing isn't as precise as the highly standardized+sterilized benchmarks. Many people (perhaps even a large majority) run and use several applications at once, which very well might represent the single largest departure between the environment you collect benchmark data in and the actual use case scenario among users. 'Sterilizing' the testing platform to only run the benchmark and nothing else may have been a natural choice in a simple mindset of controlled collection of raw performance data, but this split up between p-cores and e-cores hints at optimizations to actual real world use cases where a user has a browser open in the background with several (or even a hundred) tabs open on sites with a regular JS workload, they have an IDE running with many files open while compiling, messenger applications are opened and receiving/sending messages, one screen is handling a game while another screen handles video or music playback from a browser, etc etc. Also, another word in for Linux performance as a Linux-only user.
I wish the data would be integrated with the video and not at the front because I love to see the charts and hear what you say about the good and the bad. This has helped me choose cases, CPUs, memory, and everything else I can get.
I wonder if they could re-run this test with the newer production DDR5 and all the software updates that have been happening in windows 11 and the bios settings in the boards
Oops, looks like Alder Lake is actually next gen cpu, not because it uses ddr5, but it actually is a strong performer. Thanks nice look at memory vs cpu
Would love to see testing that isolates the performance of P and E cores separately, IPC comparison gen over gen, and extrapolate theoretical performance if you keep adding E cores (like 16 E cores)
Derbauer just did something like this, sorta. Ran games and benchmarks exclusively on P and E cores and evaluated the performance between. The E cores have quite a bit of performance in them, gonna be interesting seeing how the BIOS updates that disable the E cores so Denuvo games will run on Win 11+12th gen affects performance of those games until more devs either remove Denuvo or there's a fix released by one of the parties involved (info on that issue taken from Jay's video)
I had the initial 'calling foul" reflex when the reviews came out regarding DDR4 v DDR5. Then I thought about it for 40 seconds and of course DDR5 was appropriate the test. Looking forward to a competitive landscape.
Well, even pretty cheap ddr4 (compared to ddr5) is on par or better for gaming than most DDR5 available lol. So probably if they used good ddr4 with 12th gen the gaming benchmarks would show even higher fps for Intel than they showed now
@@damara2268 Yeah thats why this whole video is pointless, to me. I want to see 4000 c14 (high end of Gear 1 on ddr4- 12900k) vs 6800c38 (highest quality ddr5 kit announced). THAT is where it will be interesting and will tell us already whether or not to go ddr5 on this whole platform (including Raptor Lake next year).
@@CrackaSlapYa it's already obvious that ddr5 will have good price/performance in around 2 years. All other DDR iterations also needed 2 years to become actually good. So for alder&raptor ddr5 is not worth it at all. Ddr4 3600mhz cl16 at gear 1 has same or higher performance compared to most DDR5 available and all ddr5 cost and will cost more than this kit during upcoming year. Of course there are some exceptions where ddr5 is a lot better than ddr4 even now but there is low amount of those situations heh.
@@damara2268 it does until 6000+ ddr5 ram. It just depends on if you intend to upgrade every gen or not. If not, then get the ddr5 6000c34 now TGroup ram for $399 32gb kit. Then upgrade the ram in a year (wont take 2 because the demand for it is very low right now). Sure, best value is cheapest ddr4 board + b-die ddr4 + 12700k. Best set-up is g.skill 6800c38 (overclocked to 7000+) + 12900k + Apex/EVGA Dark .
@@CrackaSlapYa but do you understand that with that kind of ddr5 you're looking at about 300$+ extra price(ddr5 Mobo included)? At that point the alder lake platform becomes even worse failure than rocket lake i9 regarding price/performance lol.
I really like the methodology at the end. You should keep this format. I was chomping at the bit to see the test results, but stayed for the methodology because it explained a lot.
About the new technology point - you didn't benchmark RTX cards with DLSS on back then. I know it's not 1 to 1 but it's till new technology you arbitrarily chose to not use as a main way to benchmark those cards. Nothing wrong with that, just a point I think could fit into this discussion.
I believe that is completely different matter. DLSS runs the game at lower resolution than native. When you benchmark video cards all the cards must run the game / program at the same resolution to get comparable results about how powerful the card is. Benchmarking with DLSS enabled against competition would be as useless as running amd card benchmark at 720p and nvidia card at 1440p and comparing the fps.
Do a single Finite Element or Computational Fluid Dynamics Simulation (benchmark) with a few million degrees of freedom and you'll see why you want to go with DDR5
Please also include Microsoft Flight Simulator 2020 in your future benchmarks of these nature. It’s an extremely CPU, GPU, and RAM intensive game compared to other games so would like to see how it performs as well.
The video is ***SO CLEAR***! Thankyou for the banging content and presenting it in great visual and auditory quality. You really are the best all around tech channel.
I liked Hardware Unboxed’s rationalization where he did test DDR5 with the 12900K but for the 12600K he said you’d be spending more money on RAM than the CPU so he just tested it with DDR4 for the sake of a financially realistic build.
Thanks for this review. The CAS latency & memory frequency is useful to note when we compare the memory bandwidth of different platforms like Zen 3, Alder Lake and M1 Pro/Max and how they affect real programs. M1 Pro shows huge advantage in 7zip vs 5900x and 11900.
M1 Pro doesn't use DDR tech for memory, it's basically soldered unified GPU memory, that is an apples and oranges comparison, just unfair comparing it with these x86 chips
Please test how far the 12900k can go in gear 1 with ddr4 and of course how it performs. I hear 4000 works in 1:1 but can it do 4000 15-15-15-35 at safe voltage and so on and so forth? There's no information out on this so far.
It is unfair to use new technology for benchmark IF the new technology rises the price of whole build significantly. From what I could googled at the moment DDR5 is roughly 3x the price of DDR4 and that's assuming you can get it. I'm not even comparing mobo prices. In other example if you put two best F1 drivers, Verstappen and Hamilton in head to head race but gave Toyota Corolla to one and F1 bolid to the other, the results will be obvious. I agree that support for DDR5 is part of specification of this CPU but same can be said about support of DDR4 so testing of both configurations should be made. Thank you Steve for doing those benchmarks unlike other youtubers who decided to don't care about people who cant afford or find DDR5
I feel like we finally reached a point where the regular normal user wouldn't care if is DDR4 or 5.... (in terms of speeds and overall usage of their system) is not a world difference like it was from 2 to 3 or 3 to 4 so...... I personally have mix feelings about DDR5 I feel like I could stick to 4 for probably next 2 years just fine (unless something happens with 5 evolution and software in general that makes 5 more needed but I doubt it) again is not like when i had 2 and 3 came out or same with 3 and 4 where I kinda saw it like "god I need this new tech RIGHT NOW!" a bit sad it's not the case but at the same time thank god is not the case cos everything so expensive when it comes to tech now days.........
All memory standard changes it took a while for the new memory to mature and to really show its potential. Probably in 12 months the differences will be becoming noticeable and the ddr5 pricing will be better. But honestly it does take a good kit of ddr4 to run with the launch ddr5 so the pricing difference is not as extreme as many want you to believe. You are not using bargain basement ddr4 memory kits and timings to match the ddr5 performance even now. A ddr4 3200 cas14 kit 32gb will run you in the $200 or so U.S.D. range now. The only people buying ddr5 right now are building an Alder Lake system. If you want to be able to upgrade to ddr5 in the future and you choose to go with ddr4 now you will have to replace the mb to do so. So again the cost of staying with ddr4 now may actually cost you more down the road and limit your upgrade path for your new platform build. What you want out of your system over the long run needs to factor into the choice you make today.
we start see higher differences when they start use higher speeds for memory. cpu is 4800, like it was in beginning in ddr4 1600, ddr5 can go up to 6400. but when they start make those may take years
This was the video I was hoping for when Intel released and everyone was testing Ryzen with DDR4 and Intel with DDR5, and telling us isn't and apples to apples compassion, but in the same breath saying "it blows it out the water". Even if it doesn't make a difference its still nice to show that. Thank you for the videos tho.
I am so stoked! Like literally cannot wait for the Ultimate DDR5 showdown this Holiday Season, as the fresh sticks of DDR 5 get featured and compared by their onboard power supplies, in the Gamer’s Nexus VRM Christmas Special. It’s goona be so mind blowingly cool! Not just for DDR5, but for everything with voltage regulation. Those little modules are goona be like wax on Santas Sleigh. Gliding into that Yuletide goodness with everyone’s favorite seasonal semiconductor showdown!
Would be interesting to see a DDR3 top spec, vs. DDR4 base spec and top spec, to try to explain what we're seeing with DDR5 now; it's the earliest kits and therefore not the be-all-and-end-all of DDR5 tech.
sadly that will be next year, maybe in q1 maybe in q2, and if you have all of your setup already bought, like me, with ddr5 mobo, can't wait that long. Hopefully 5600 kits come out by the end of this year atleast, which i'm now waiting for.. only missing ddr5 to boot everything up
2:04 - Hey Steve, loving the new set … looking forward to some behind the scenes footage on your relocation. 😄 But just wanted to say I found the "encroaching shadow of doom" distracting. I kept expecting somebody to appear, but it was just you stepping back, as in this example. Hope that helps. 🙏🏻👍🏻
"It's not fair to compare Intel on DDR5 to AMD on DDR4!" Advancements in the platform are also something that has to be considered when review the product. It's fine to want reviews/results from both memory types, but it's not inherently unfair to use the better of the supported memory types with Intel, it's just part of the equation when figuring out the value proposition.
Very interested in seeing some benchmarks where you selectively disable cores of one type or the other. I'm seriously curious how they compare in both per core performance and the actual power usage per performance differences. Also would be nice to see exactly how few high performance cores you need in various games if you have the e-cores running.
Great video again. People are way too quick to rush to the newest technology when not always necessary. I don’t understand the memory timings that much yet, so will have to research that part more. Good to see the new studio, kinks and all. Would be neat to see a follow-up video or the studio progress.
Another reason to pass on 12th gen. Beyond a full platform change, and the software compatibility issues all over the place ... upgrading when DDR5 is not just incredibly expensive, but still in its infancy technically, means early adopting with a comparably crappy kit now, just to change it up again when DDR5 matures, in the future when a whole new CPU generation will be out, probably better than gen12 ... be it gen 13 or 14 ... or the leap past 1700 socket. I'm just not seeing the value if your sitting on a strong 9th gen or the better 10th gen.
It's strongly tempting for anyone with 6th gen or less. If X570 had thunderbolt, I'd upgrade my 5960x to a 5950x. At this point, I've waited so long that I'm just going to continue waiting until Zen 4 and 13th gen Intel are out next year.
I'm on an 8th gen and am feeling a little bit of pain from hardware creeping up in performance, but I feel I am forced to wait all the same by all of the generation's issues.
Don't too fixated on single parameter. Is your good RAM with lower CAS can beat current ram with higher CAS? Obviously they wouldn't because there are improvements that can be done by sacrificing some latency. It's moot point to complain only about single specific parameter on complex system.
thanks for the benchmarks Steve. it is very clear now that DDR5 simply isn't worth the hassle for an alder lake build. however now knowing that DDR4 is far better value and almost identical performance i would love to know how the Alder lake IMC can handle some of the top of the line memory kits. TLDR overclock the shit out of DDR4, drop timings as low as you can and re run these tests to see if the best of DDR4 can beat current DDR5. also i can't wait to see overclocking on the 12700k and 12900k.
"It was a little weird to see that question" Not weird at all. AMD is better than Intel /s Seriously, AMD has been the underdog (and underperforming) competitor for so long, that when they launched a good product line (Ryzen), it birthed a whole new generation of fanatical supporters. Support AMD all you want (hello die hard fanboys), I will just buy the best value for my money. As of right now that seems to be the 5700G ironically, but I'm holding back because the competition is still so hot that any new tech launched this year is gonna be almost obsolete by next year. GN is one of the most, if not the most, impartial reviewer I've seen on UA-cam.
When with performance being the same I think I'd lean slightly towards AMD due to socket lifecycle, fewer board chipsets and not locking overclocking to specific cpus.
Thanks for the benchamark! On some other channel we could see ttha 6000MT ram did not improve performance compared to 5200. It will be interesting to see how DDR5 is going to scale in the future.
Sometimes new technology takes a while to become an obvious speed increase--especially when it needs to work in unison with other things. DDR5 is one of them. Also, the new-ish (not so new anymore) AGP spec. btw, 10:33 has an error. Far Cry 6 does not favor DDR4. :)
My follow up request would be, do games that are known to be RAM dependent, simulation games, like Factorio. Measuring updates per second rather than frames per second on a very large base.
I can't recall the test exactly but someone tested DDR5 4800 vs 6000 and showed a pretty big difference on the 12900K. Since 5800-6000 seems to be the slightly higher end like a 3200mhz kit is above spec for DDR4 I would love to see a comparison of the DDR5 5200 vs 6000 at least and maybe higher. If the results continued to be similar I think waiting till DDR5 drops before purchasing a kit next upgrade would be a better call and just get a D4 board.
The price:performance is exactly why I wanted to see the ddr4 vs ddr5 on this platform. I agree that it’s an advantage of 12th gen having the ability to use ddr5 vs ddr4 on amd, but that was also one of the biggest things with zen3 vs 10/11th gen intel, a price:performance When a cpu (like the new i5) is being touted as “x% ahead of a competitor for the same price” those claims usually aren’t considering the entire platform opportunity cost, but only the cpu itself This was a really good comparison to do It allows somebody to decide on a ddr4+12gen vs ddr4+zen3 vs ddr5+12gen and their pricing for each platform
As a reminder: We MANUALLY TUNE the DDR4 kit for all of our reviews because we have controlled every timing for that kit for CPU reviews for years. This means that it outperforms any XMP 3200 kit on the market because we have hand-tuned it (it works on multiple generations of CPU, too). This kit outperforms our 3600 kits and 3800 kits because of the manual timing tuning done. So no, it is not as simple as just saying "you should use a higher frequency kit," because it ignores that just doing XMP doesn't hand-tune timings (like we do on the 3200 kit).
We reviewed the i7-12700K (and KF) here: ua-cam.com/video/B14h25fKMpY/v-deo.html
Here's our Windows 11 vs. Windows 10 CPU comparison on Alder Lake: ua-cam.com/video/XBFTSej-yIs/v-deo.html
We also reviewed the 12600K here: ua-cam.com/video/OkHMh8sUSuM/v-deo.html
We kicked off our Alder Lake reviews with the i9-12900K here: ua-cam.com/video/fhI9tLOg-6I/v-deo.html
We covered Alder Lake’s architecture here: ua-cam.com/video/htCvo9XJZDc/v-deo.html
Studio is looking great! Looking forward to the moving vlog!
Comeon man i just upgraded this year
@@2511jeremy Upgrade right now, then you can be "obsolete" again in one year.
Can you please post the full timings you use for reviews? Like an Asrock Timing Configurator screenshot? (Handy program that pulls up all your timings from Windows if you're not familiar with it). I think including these full timings would be an excellent addition to any memory related video. Thanks!
Did you swap the bars for the RdR2 results? Because the graphics don''t seem to go along with what you say.
So switching ddr generations is like gamers nexus with new studios. It takes a while before a massive uplift in performance.
"Sigh."
_unzips file_
"The shit I put up with..."
As has been since the dawn of dram; be it DDR, DDR2 ect.
So excited about the potential here! You're right - it'll take a while to fully use it!
@@GamersNexus one thing I noticed in this video is that you could see your shadow on the right of the screen, a bit more often than I remember seeing in your old studio, although that could be because your new set looks so naked, so it's more noticeable... I'll bet you're going to update your lighting at some point, which should help immensely, if you have the required space and height for it to be workable, not that it really takes away from the content, because it doesn't, but it just feels so unlike you to have a rogue shadow in the frame, as you are so, uh, perfectionist, which is a good thing...
But maybe I just never noticed your shadow and the lights in your old studio were also bad for shadows... Either way, that's still a pimpin studio/testing facility you have, and I can't wait to see a video on UA-cam, of you touring the place, once it's completely done being remodeled and whatever else your current plans are...
@@northwiebesick7136 That's just the ghost of videos past. When you turn your head and see it in your room I'd be concerned. It might be looking at you now. 👀
6:42 "In this scenario we observed DDR5 producing the best result with a 7.1% advantage over the DDR5 kit" 🤣 Steve get some rest, moving into this cool new background must be exhausting
hahahaha, oops. Thanks. Didn't even notice that!
@@GamersNexus we did!
Please add an errata to the notes.
With current DDR5 prices, DDR4 may be the clear winner right now if you are going for 32GB. With the money you save from DDR4 you could go to a higher tier CPU or GPU, for better performance overall.
100% valid reason!
I dont mind paying the extra for DDR5, prob is i cant. Asus the retards prevents use of most air coolers from their mid range boards (the strix models below Hero). Its just stupid cosmetic size of the garbage around the VRMs. Theres only 1 other ITX mobo and its a DDR4 model from Gigabyte. While AIOs coolers are all the rage now, theres still a significant portion of market on air. Asus FFS! Way to lose $
Design it by your needS not by your preference 👍
@@user-lh1ef1st9k Its a HTPC that will fit in a floating TV wall unit. Limited space. Im setting up with a 75" tv, hopefully i can take all my pc use to the lounge room. I worry about font size with certain games though and i may need to have 2 good pcs.
Ive thought about remote control of the pc from another room, but cant work out the cabling required. (Video to TV, sound to soundbar, wireless Mouse + keyboard control - will wall block? if USB extenders with the receivers on end then ill need some type of powered repeating hubs, the hdmi cable needs power of very high price fibre core
With very marginal gains…. Yeah okay
glad to see the background is slowly getting populated by tools
Still finding & organizing everything following the move!
@@GamersNexus nice! studio is looking great
@@BuildingDR this isn't the verge PC build
Ifixit?
@@Dr.RichardBanks Yeah, the Verge has its tools in the foreground.
I like how recently you've started adding a quick summary at the beginning of the video. It doesn't stop me from watching the video, on the contrary -- it makes me eager to hear the details. Good move!
I know this wasn't something one should genuinely complain about, but I appreciate it so much that the blue tool panel is no longer catching reflections.
haha, thanks. We've been working on that each shoot. Andrew has been tweaking lighting. Our next one gets rid of the giant hard shadow cast by me when stepping back.
Yeah, the sharp shadow is pretty distracting indeed... Maybe a layer of some matte finish for the panel could help with the gloss? With less stuff catching reflections, there should be more freedom in setting up lights.
@@GamersNexus I thought that was your new cast member "Shadow Steve"
@@GamersNexus That's what a fill light is for! Easy to setup in theory, hopefully your setup doesn't make it too difficult.
@@Lishtenbird could also try adding a polarizing filter to the camera to potentially reduce glare off the glass
Factorio is a game known to be bottlenecked by both memory speed and latency (mainly latency), and as it have ways to reliably benchmark it, testing it in this scenario would have been interesting to see the results.
Came here to make the same comment. I'm interested in upgrading so I can run pyanadons megabases at yknow...60ups
HardwareUnboxed did that.
And hardware unboxed also showed they know nothing about memory when discussing the results. It's shameful to pretend to be an authority rather than admit to yourself you need to learn and grow.
I play modded Factorio with a Ryzen 7 and a very old R9 390 video card.
It does cause performance issues
I get frame rates dropping to 55 FPS with my 2 year old 5000 science pack/minute megabase.
@@MichaelClark-uw7ex I did play it modded as well, and noticed when my bios disabled my XMP profile as performance tanked.
My i7-4770k with 16gb of DDR3-2400 C10 is noticiably faster than when running stock at 1333mhz.
As long as you have suffient CPU power (it doesn't need much) the memory is the actual limitation in it.
What perfect timing. My GN coasters arrived not 20 minutes ago. Perfect.
Awesome!! Thanks for buying them! Sold out so fast it was crazy.
*takes sip*
*puts drink on coaster*
Those are siiick!
I would really like to see how alderlake is doing in MMORPGS like, guild wars 2, black desert and so on,... Its a pretty big playerbase that allways get left out and I wasnt able to find benchmarks for any mmo. Keep up the good work! Love your content!
I bet very very good as those type of games can get single thread bound.
@@CrackaSlapYa not every mmo... Ashes of Creation (al though in Alpha) is pretty heavy stuff🤪
the fact you have to explain the "fairness" of testing a DDR5 equipped/enabled chip with DDR5 is sad. Thanks for the work you guys put in to informing consumers.
fairness isn't something manufacturers care about these days.
Yeah, I find that criticism odd. DDR5 is available, use it.
I have always loved AMD, but when I hear "should have used the same ram in both" just seems like wanting to pad numbers in AMD's favor.
These are generational changes and advancements, I like seeing them used when they come out, even if it's not the best deal.
Off topic, but will you be having a new studio tour video? Love to see how it all turned out, it's looks amazing from what I see here 😁👍
Yes! Working on editing many of those now!
@@GamersNexus Awesome! Thanks so much Steve, to you and your crew looking forward to it. 😁
@@GamersNexus Fantastic! Definitely keen to see the before and after comparisons. Just the paint job alone gave the space a whole new feel.
Ummm...Sorry Steve that not how the Internet works. If your empirical testing doesnt validate my personal bias, its clearly unfair.
:)
Its clearly fair because my bias is bigger than youra
based
It's not really the right type of empirical testing if multiple changes are made. That's not how empirical testing works. You change one thing to comment on that one thing.
If you must/have to change multiple things, you thus change the probabilities of the results significance to which part made the difference.
Same with JayZ2cents "empirically" testing AMD vs Intel power drawn, then putting a massive watercooler, RGB fans, HDDs etc all on the AMD only system, then running the Intel system barebones... That's not empirical or sensible testing.
Here, Intel is moving to DDR5 too. *That's fine, and GamersNexus is absolutely allowed to do the testing, and it's great they are*. It just changes the statement from "this cpu is faster" to "this platform is faster" until we then test DDR4.
If Intel released this CPU with only DDR5 options. Then yes, we would say "This CPU and platform is faster" and being human, we would accept the assumed shorthand of "This CPU is faster" because we know we are not talking about it in isolation. But as Intel have allowed DDR4 platforms for the 12gen, it's not possible to talk about it in isolation and only for DDR5 performance, we literally have to include it when making statements, unless saying "only on DDR5". :)
Tradeoffs in memory throughput vs. latency:
Compression benefits greatly from throughput over latency because the task batches large segments of memory together. Things like high fps gaming or games with tens of thousands of entities need to shuffle data around more to keep the CPU execution fed, so they miss the cache more and read in smaller chunks from memory when they do. Reducing memory latency reduces the length of the stalls, so in these applications memory latency is more valuable.
Additionally, in transitions to DDR3 and DDR4 we have seen higher latency, higher throughput kits than the previous generation for the first year. They eventually settle back around to the same "best latency SDRAM can muster" but with higher throughput.
I would have preferred a big fat disclaimer that the data is not representative of actual performance in most cases at the beginning of the video. Manually tuned B-Die in gear 1 vs not even the best DDR5 kit on the market not manually tuned in gear 2 doesn‘t even come close to being a fair comparison if you really wanna know which platform is capable of performing better overall. This methodology really only makes the testing valuable for the extreme cases where there is a big uplift from the additional bandwidth and maybe clock speed where timings don‘t matter.
Still thankful for the testing but I‘m much more looking forward for the future data once you get your hands on one of those 5600C36 D5 kits and got the hang of how to manually tune those. I don‘t know about elsewhere in the world but the pricing of B-Die D4 kits from G-Skill Trident Z vs the best Z5 variants are not that far apart here in Germany so it really is a fair comparison with maxed out 32GB kits.
Thanks for this. The reason I wanted to know about DDR 4v5 was to know how much of the ~25% gains Alder Lake offered were from the RAM vs the CPU itself. Looks like most of it really is the silicon. Good to know!
Golden Cove is the new Sandy Bridge.
@@saricubra2867 Yeah man. Even with eCores off, this is an 8 core/16 thread beast with these clocks + ipc.
@Vincent Tuminello can you give me an update on this? I fully upgraded my pc for call of duty warzone, and am completely disheartened today I could notice input latency from the 4800 DDR5 Ram I have, I was getting destroyed compared to sadly my 3200 mhz off brand crap I got with a pre built pc a year ago... I have DDR4 g skill trident 4000 mhz CL14 ram in the box in my closet I was going to return, but i'm strongly considering attempting to install these and seeing if my luck changes, I was getting beat in situations I shouldn't have every single time basically... do you think the latency of DDR5 in its current new state is not good compared to ddr4 for 1st person shooters? I have a z690 motherboard as well
@Vincent Tuminello based. fuck the jab
@@brutyfruity unless you are running the game at super high framerates RAM won't have any impact. Try it with overwatch at 400fps cap and you will see massive gains in fps and 0.1 and 1% lows as well as frametime consistency. Overwatch is one of the few games that are massive ram bandwidth hogs if you try playing them at very high framerates. Lower framerates make ram mostly irrelevant. I highly doubt you are bottlenecked by ram at all in warzone.
Could be interesting to see the impact of DDR4/5 when using integrated graphics. Would be more relevant with laptops generally but could therefore make RAM choice more significant since that is a component manufacturers will inevitably use to cut costs or tout the presence of DDR5 for marketing purposes even if it might not make a notable difference for CPU as shown but graphics is not known yet.
Conclusion; it’s worth it but only purchase it if it’s inline with your plans.
This is what happens every generation of DDR, the established standard being more mature has already been optimized for latency and had the clock rates pushed to the edge, while the newer generation usually just focuses on higher clock rates for higher raw bandwidth, so for the first period on the market and as the standard matures you'll get better clocked kits with relatively lower latency like when DDR3 first hit the market CAS 5 1066mhz DDR2 was common in the high end, but DDR3 was largely only available in JEDEC standard timings so like CAS 7 for 1066 and CAS 9 for 1333 but as it matured you could get CAS6 1600 kits and cas 7 2000 kits. So some applications highly benefit from high raw bandwidth where some are more negatively impacted by the relatively high latency.
This is what has happened before, however it is not inherent or guaranteed to be this way always.
Results in the past cannot predict the future in this case. There is nothing inherent in memory design restrictions, that requires this.
@@zazethe6553 there is some degree of predictability because "DDR5" is merely a JEDEC standard. Just like PCIe 5.0 is merely a standard.
Predictably, these DDR standards' primary focus is on doubling bandwidth (this can be achieved in various ways).
I don't know what the standard for DDR6 is, or if it has even been decided yet, but I can confidently predict a doubling of bandwidth over the DDR5 standard.
The secondary focus is lowering power consumption. That one is less quantifiably predictable.
I'm over here still running ddr3 mhz 1800.. nice to see ddr5 come out and hopefully drive ddr4 prices down. Thats when i swoop in and pick up a new build
I know it doesn't apply here all that much in this round of tests in particular, but what's the difference in igp performance? That relies on the ram heavily. More a curiosity really, but if there is a significant uplift for some reason, it may help people wait when dedicated GPUs are more available.
Yep, really sad that we didn't see integrated graphics comparison. If it's even possible to test properly.
Glad to see this testing! Roman's testing of DDR4 vs DDR5 on Alder Lake was fascinating; I'm looking forward to seeing this coverage
For the graphs:
could you please always keep the order the same; meaning DDR5 always on top and not the best result on top?
I installed windows 11 a few days ago and had a big drop in game performance..Checked online and lots of other people seem to be having this problem too..Dont know if you do software stuff but thought it would make a good video..Thanks for all your hard work.
Interesting Benchmark idea:
"will the E-cores save your FPS?"
the e-cores is great so some random messenger or discord chat won't kill your FPS
but how much of a performance hit will it save?
i7-12700K, with only 4 P-cores enabled, vs 4 P-cores and 4 E-cores. while we're at it, maybe vs 6 P-cores and 8 P-cores
scenario 1: sanitized windows install (no BG tasks)
scenario 2: bloated windows install (4GB worth of bg tasks with combination of chrome tabs, RGB software, anti-virus, and an active discord voice call)
The 12900k is like the 5950x. No you don't need 16 cores. No 16 cores may not even help over 12 cores. But what you are getting is better binned CPUs. More total capacity for work. Better silicon. And you have reached the top component for the board you are on.
Do you really care that your CPU blender render took 16 minutes instead of 20? No probably not. Do you care that you got 165fps instead of 162? Or that 0.1% lows were 10fps higher? No probably not. A 12 core and a 16 core will have very simmilar frame time graphs aswell. But when you compare the cost of a whole system.
$2000 for a 5900x system. Vs $2250 for a 5950x system.
Same for intel.
Do you save the 12% and lose 25% of your computing power? If you don't do loads of professional work. And you play older games, or you plan on buying newer games a couple of years late when they are on sale. Or you plan to upgrade in 2-3 years.
Then probably yes.
If you plan on playing new games right when they come out. If you plant on multitasking like a madman (I do this). Or if you just want your CPU to last through multiple GPU upgrades. Go with the 12900k or the 5950x. Both are great. Neither really out performs their lower end counterpart today. But they may in the future.
@TheHammerGuy - Intel has turned to E-cores because they failed to get anymore than 8 power cores on their 10 nanometre node.
So they made a Frakenstein's monster of a CPU.
For the same price, instead of an 8p/8e or 6p/4e you can get a 16p-core or an 12p-core processor. Waaaay more convenient when you consider the price, and the components you'd have to upgrade and their cost as well as needing a new power supply. As you only need about 140 Watts for a Zen 3 processor.
@@enriquepadilla2542 thanks for this comment I went with a i9
@@enriquepadilla2542 This just isn't good advice during a performance war. If you have to decide if you really need an i9, you don't, and you would be better off buying a new i5 or i7 again sooner.
@@enriquepadilla2542 I would love to undervolt a 12900k, but not going to pay $1000. 5950 is a better value, if you will keep it busy all day. 12700k looks good to me.
I'd love to see this test revisited in a few years once DDR5 has matured a bit
And also using some nice DDR4 4000mhz cl14 or better, because DDR4 is also still improving.
@@Adreno23421 the 4000mhz cl14 is limited to 8gb per channel currently, so 16gb for 180 currently, would run you 360 to get 32gb of it.... pretty pricy , but have not seen where it provides performance boost, have any reference?
Small studio critique:
That light, that is to the left of the camera, makes you cast an irritating shadow on the wall when you move back from the table.
You should change something to avoid that. 😉
They have, he said so in a reply.
We already said in the comments that we changed it.
@@GamersNexus good on you!
Awesome update. I wasn't concerned with your testing methodology (you explained it all and it made sense for the choices you made), but I was very interested from a consumer point of view if there were downsides to going one direction vs the others (DDR4 vs DDR5, Win 10 vs Win 11) that would make it a bad idea to even consider that. You've covered all those and now its very much down to what you intend to do now/in the future as opposed to it being potentially crippled and losing out. Great stuff.
When you guys do your efficiency testing I'm very curious about where things sit in perf if you restrict the power levels down to some baseline. I've seen a lot of speculation about power efficiency based on the 12900k but the reality is none of these chips are running anywhere near to their most efficient levels up near 5Ghz; they are all spending tons of additional power for the last X% of performance just to be at the top of the benchmark chart. I'm quite curious what that X is, and what that curve looks like between Alder Lake and Zen.
Good remark.
These chips can be undervolted by some margin. Stock 12900k is way off of its sweet spot and 12700k is efficiency monster with slight tuning.
Would like to see something similar to Der8auer's recent video where you assign only E-Cores or only P-Cores to a program and benchmark E-Core VS P-Core: maybe max overclocking (% over listed spec), 4GHz vs 4GHz, 8 e-cores vs 4 p-cores (with and without SMT), etc...
that's an interesting benchmark
r7 5800x sanitized test bench vs r7 5800x with 2GB worth of Chrome tabs.
and
i7 12700K sanitized test bench vs i7 12700K with same parameters as above.
"ARE THE E-CORES SAVING YOUR FPS?"
I've told some people that the great thing about the hybrid design helps to alleviate BG tasks so some random discord chat wont kill your FPS.
@A L E X A N D R A、🌹 Death to ASMR and for those who like it!
@@TheHammerGuy94 I don't know if that's the best way to do that test, but we'd mostly end up looking at the 1% lows. Methodology aside, I am hyped for cpus in the future which will have like 16+64 cores because background tasks will never bother me ever again... As a massive multitasker that likes to leave things open, this is huge.
Quick tip since I noticed the can of CRC contact cleaner on the shelf- that stuff is magic for removing thermal paste. Dissolved the paste out of an LGA socket on a Craigslist motherboard, and got a thermal paste stain out the carpet!
"thermal paste stain in carpet", aight sure bud 😏
yea its honestly works well for cleaning dang near anything but work great of oily and greasy things
Amazing content and breakdown as usual Steve. Thanks for all the hard work.
Have you thought about adding game engine build tests for productivity testing?
For example, building the UE5 Valley of the Ancient sample? From the recent SIGGRAPH talk, it's doing a lot of heavy composition when building the nanite meshes (graph partitioning, and variable length compression). Additionally, given the working set size won't fit in the LLC (one of the examples was up to 26GB -> 6GB after compression), it will likely benefit from the extra memory channels (more random addresses per second) more so than bandwidth.
game devs are losers go away
@@Cd5ssmffan UE is used for things other than game development... I know people who use it to do 3D animations for independent movies, and people who use it for concept development in other media companies (including some XF studios). I have many colleagues who use it to generate AI training datasets, one of such projects trains traffic cameras to reduce traffic jams and red lights. I also know people who have used it for IC visualization before sending them to the Fabs and as a demo to raise capital to send a chip for fabrication. And I have used it in the past for simulation visualizations (things like fluid flow for power stations).
The latter two cases wouldn't rely on things like Naninte, but do rely on engine building from source. But the media and AI training examples could rely on it.
@@Cd5ssmffan This comment on a channel called "Gamers Nexus" is, uh, unusual
The new set and lighting look good, Steve! It's easier to see you and the overall picture is clearer. Gamers' Nexus videos always had a slight blurriness to them before.
Probably the biggest application that could bring tangible benefit from switching to DDR5 over DDR4, that I can think of, is ZFS ARC caching to RAM on a NAS, especially if you have compression or dedupe on.
I enjoyed the comparison for the ddr4 to the ddr5 with the Alder Lake chipsets. I hope you continue this in-depth analysis of these new technologies, especially as they begin to mature and manufacturing processes become more streamlined for ddr5.
I know this is probably the last thing in the priority list but I would love to see the igpu performance with ddr4 vs ddr5.
Yeah, sometimes I'm a bit sad Gamers Nexus doesn't talk in detail about laptops / iGPUs. Intel has come such a long way with its latest Xe graphics that it's totally viable for a low end gaming machine, especially with the current GPU situation.
I second this
I endorse this as well.
I'm curious as well.
I'm waiting for DDR16
Please adjust your lighting and or diffuser. The shadows, especially yours, is very distracting..... Great content as usual..
As much as I enjoyed watching this, I really look forward to seeing the second generation of ddr5 and its performance. Early adoption can be rough. It'll be cool to see what it looks like when they honed the process a little more.
Yep. With WinZip having such a high performance uplift it also shows possibly some optimisations might be possible now that the memory bit width is different. Possibly the likes of GTAV is optimised for DDR4 on PC, thus looses performance in DDR5 setups. Will be interesting to see what happens.
@@TechyBen The problem is that current latency for DDR5 is bad, just like DDR4 when it was launched or any other DDR generation.
Basically now if you buy DDR4 3600 cl16 which has pretty good price you already beat or are on par with most of currently available ddr5.
And all the ddr5 are more expensive.
It's just scalped disgustingly.
Yep that's specifically why I skipped DDR5 right now even when buying a 12600K. Ended up going for a cheaper DDR4 board instead and will probably just flip that in a couple years when better DDR5 modules come out.
@@damara2268 close to true. 3600c16 is pretty bad though. 4000c14 is pretty much maxing out gear1 low latency on a ddr4/Alder Lake build. If the ddr5 is the GOOD stuff (6200c34 kit from TGroup or g.skill 6800c38) then it matches/beats it. So, it has to be really fast to beat it, but, it does beat it at the highest end of gen 1 spec of ddr5.
This is really usefull to know, because one of my friends is getting a new computer and at least at the moment in Australia DDR5 cost x4 ($800 for 32GB) more then DDR4 (~$200 for 32GB)which is quite rediculous for what might only be 3-5fps more.
I have a feeling DDR5 with lower latency and Gear 1 mode will be very good. Or maybe DDR5 with faster frequency, lower latency, and still in Gear 2 mode.
Der8auer tested several kits of ddr4 and was running anything above 3600 in gear 2 mode, so i assume that's about as fast as you can go in gear 1. Faster ddr5 may also be too much for Alder lake's imc and require gear 4.
That background looks way cleaner! I love it.
Who else is still on DDR3 like me ?
Are you still on ddr3? 😂😂
@@KurayamiDawn yes till this day
@@GJCanave the dedication 😲🫡 probably gonna ask the same after 2 years 🤣🤣
You always make the videos comparing technologies I want to see. Thank you 🙏
Please consider testing the iGPU performance difference between DDR4 and DDR5. I expect to see a decent improvement there but it would be good to know if we should be looking specifically for ddr5 on our next laptop or APU desktop purchase. Also it would be nice to see how different DDR5 clock speeds influence the iGPU performance
I mean.. since you asked.. I’d love to watch videos comparing the 4790k to the 12700k/12900k for gaming and streaming.
Also, thank you for helping me learn something new every day!
Good to know that there is not much of a difference between ddr4 and ddr5 for just standard CPU operations. I would be interested seeing how ddr5 would compare to ddr4 for the IGP.
It definitely had a massive effect on digital audio benchmarks, you saw a 20-30% increase in task operation speed between DDR4 and DDR5
Great videos, no punches pulled and truthful reviews!!
I would appreciate exploring concurrent tasks on the new Alder Lake architecture, even if the testing isn't as precise as the highly standardized+sterilized benchmarks. Many people (perhaps even a large majority) run and use several applications at once, which very well might represent the single largest departure between the environment you collect benchmark data in and the actual use case scenario among users. 'Sterilizing' the testing platform to only run the benchmark and nothing else may have been a natural choice in a simple mindset of controlled collection of raw performance data, but this split up between p-cores and e-cores hints at optimizations to actual real world use cases where a user has a browser open in the background with several (or even a hundred) tabs open on sites with a regular JS workload, they have an IDE running with many files open while compiling, messenger applications are opened and receiving/sending messages, one screen is handling a game while another screen handles video or music playback from a browser, etc etc.
Also, another word in for Linux performance as a Linux-only user.
I wish the data would be integrated with the video and not at the front because I love to see the charts and hear what you say about the good and the bad. This has helped me choose cases, CPUs, memory, and everything else I can get.
Suggestion: Do something about the shadow on the right side of the frame in the wide shots
Thanks GN. The question was "How much of the performance is DDR5?". Here we see in data that the answer is "Not Much". Asked and answered. Well done.
I wonder if they could re-run this test with the newer production DDR5 and all the software updates that have been happening in windows 11 and the bios settings in the boards
Oops, looks like Alder Lake is actually next gen cpu, not because it uses ddr5, but it actually is a strong performer. Thanks nice look at memory vs cpu
Yep. People were not just complaining the tests were unfair, but that the statements needed the results from DDR4 to actually make full sense of them.
Would love to see testing that isolates the performance of P and E cores separately, IPC comparison gen over gen, and extrapolate theoretical performance if you keep adding E cores (like 16 E cores)
Derbauer just did something like this, sorta. Ran games and benchmarks exclusively on P and E cores and evaluated the performance between. The E cores have quite a bit of performance in them, gonna be interesting seeing how the BIOS updates that disable the E cores so Denuvo games will run on Win 11+12th gen affects performance of those games until more devs either remove Denuvo or there's a fix released by one of the parties involved (info on that issue taken from Jay's video)
@@nicwilson89 I think Derbauer also tested AVX-512 and that can make the P-cores even faster.
Thanks for making this video!
I had the initial 'calling foul" reflex when the reviews came out regarding DDR4 v DDR5. Then I thought about it for 40 seconds and of course DDR5 was appropriate the test. Looking forward to a competitive landscape.
Well, even pretty cheap ddr4 (compared to ddr5) is on par or better for gaming than most DDR5 available lol.
So probably if they used good ddr4 with 12th gen the gaming benchmarks would show even higher fps for Intel than they showed now
@@damara2268 Yeah thats why this whole video is pointless, to me. I want to see 4000 c14 (high end of Gear 1 on ddr4- 12900k) vs 6800c38 (highest quality ddr5 kit announced). THAT is where it will be interesting and will tell us already whether or not to go ddr5 on this whole platform (including Raptor Lake next year).
@@CrackaSlapYa it's already obvious that ddr5 will have good price/performance in around 2 years.
All other DDR iterations also needed 2 years to become actually good.
So for alder&raptor ddr5 is not worth it at all.
Ddr4 3600mhz cl16 at gear 1 has same or higher performance compared to most DDR5 available and all ddr5 cost and will cost more than this kit during upcoming year.
Of course there are some exceptions where ddr5 is a lot better than ddr4 even now but there is low amount of those situations heh.
@@damara2268 it does until 6000+ ddr5 ram. It just depends on if you intend to upgrade every gen or not. If not, then get the ddr5 6000c34 now TGroup ram for $399 32gb kit. Then upgrade the ram in a year (wont take 2 because the demand for it is very low right now). Sure, best value is cheapest ddr4 board + b-die ddr4 + 12700k. Best set-up is g.skill 6800c38 (overclocked to 7000+) + 12900k + Apex/EVGA Dark .
@@CrackaSlapYa but do you understand that with that kind of ddr5 you're looking at about 300$+ extra price(ddr5 Mobo included)?
At that point the alder lake platform becomes even worse failure than rocket lake i9 regarding price/performance lol.
I really like the methodology at the end. You should keep this format. I was chomping at the bit to see the test results, but stayed for the methodology because it explained a lot.
About the new technology point - you didn't benchmark RTX cards with DLSS on back then. I know it's not 1 to 1 but it's till new technology you arbitrarily chose to not use as a main way to benchmark those cards. Nothing wrong with that, just a point I think could fit into this discussion.
I believe that is completely different matter. DLSS runs the game at lower resolution than native. When you benchmark video cards all the cards must run the game / program at the same resolution to get comparable results about how powerful the card is.
Benchmarking with DLSS enabled against competition would be as useless as running amd card benchmark at 720p and nvidia card at 1440p and comparing the fps.
Let's gooo! Been waiting for this one.
Do a single Finite Element or Computational Fluid Dynamics Simulation (benchmark) with a few million degrees of freedom and you'll see why you want to go with DDR5
I haven't seen this bench, but I can imagine lol
Please also include Microsoft Flight Simulator 2020 in your future benchmarks of these nature. It’s an extremely CPU, GPU, and RAM intensive game compared to other games so would like to see how it performs as well.
Prediction: Once Zen 4 arrives, DDR5 will have matured enough so that they don't have to offer motherboards with either DDR4 or 5.
The video is ***SO CLEAR***!
Thankyou for the banging content and presenting it in great visual and auditory quality.
You really are the best all around tech channel.
I liked Hardware Unboxed’s rationalization where he did test DDR5 with the 12900K but for the 12600K he said you’d be spending more money on RAM than the CPU so he just tested it with DDR4 for the sake of a financially realistic build.
Thanks for this review. The CAS latency & memory frequency is useful to note when we compare the memory bandwidth of different platforms like Zen 3, Alder Lake and M1 Pro/Max and how they affect real programs.
M1 Pro shows huge advantage in 7zip vs 5900x and 11900.
M1 Pro doesn't use DDR tech for memory, it's basically soldered unified GPU memory, that is an apples and oranges comparison, just unfair comparing it with these x86 chips
Please test how far the 12900k can go in gear 1 with ddr4 and of course how it performs. I hear 4000 works in 1:1 but can it do 4000 15-15-15-35 at safe voltage and so on and so forth? There's no information out on this so far.
AMD 5800x can do it why not 12900k
So ..what exactly are those 3200 DDR4 timings so that your fans can use em too with the same mobo.
Hey Steve, when will the 4 coaster be available again on your store? I want to buy a pack.
It is unfair to use new technology for benchmark IF the new technology rises the price of whole build significantly. From what I could googled at the moment DDR5 is roughly 3x the price of DDR4 and that's assuming you can get it. I'm not even comparing mobo prices. In other example if you put two best F1 drivers, Verstappen and Hamilton in head to head race but gave Toyota Corolla to one and F1 bolid to the other, the results will be obvious.
I agree that support for DDR5 is part of specification of this CPU but same can be said about support of DDR4 so testing of both configurations should be made. Thank you Steve for doing those benchmarks unlike other youtubers who decided to don't care about people who cant afford or find DDR5
I feel like we finally reached a point where the regular normal user wouldn't care if is DDR4 or 5.... (in terms of speeds and overall usage of their system) is not a world difference like it was from 2 to 3 or 3 to 4 so...... I personally have mix feelings about DDR5 I feel like I could stick to 4 for probably next 2 years just fine (unless something happens with 5 evolution and software in general that makes 5 more needed but I doubt it) again is not like when i had 2 and 3 came out or same with 3 and 4 where I kinda saw it like "god I need this new tech RIGHT NOW!" a bit sad it's not the case but at the same time thank god is not the case cos everything so expensive when it comes to tech now days.........
DDR3 to DDR4? Not when my CPU w/motherboard runs DDR3 in 4 channel!!!
All memory standard changes it took a while for the new memory to mature and to really show its potential.
Probably in 12 months the differences will be becoming noticeable and the ddr5 pricing will be better.
But honestly it does take a good kit of ddr4 to run with the launch ddr5 so the pricing difference is not as extreme as many want you to believe.
You are not using bargain basement ddr4 memory kits and timings to match the ddr5 performance even now.
A ddr4 3200 cas14 kit 32gb will run you in the $200 or so U.S.D. range now.
The only people buying ddr5 right now are building an Alder Lake system. If you want to be able to upgrade to ddr5 in the future and you choose to go with ddr4 now you will have to replace the mb to do so. So again the cost of staying with ddr4 now may actually cost you more down the road and limit your upgrade path for your new platform build.
What you want out of your system over the long run needs to factor into the choice you make today.
we start see higher differences when they start use higher speeds for memory. cpu is 4800, like it was in beginning in ddr4 1600, ddr5 can go up to 6400. but when they start make those may take years
I'm more curious about dd5 on an igpu as that might actually have an effect
This was the video I was hoping for when Intel released and everyone was testing Ryzen with DDR4 and Intel with DDR5, and telling us isn't and apples to apples compassion, but in the same breath saying "it blows it out the water". Even if it doesn't make a difference its still nice to show that. Thank you for the videos tho.
"Isn't it unfair to run AMD with all of its cores enabled, because Intel doesn't have that many cores?" This one made me laugh! 🤣🤣🤣
I am so stoked! Like literally cannot wait for the Ultimate DDR5 showdown this Holiday Season, as the fresh sticks of DDR 5 get featured and compared by their onboard power supplies, in the Gamer’s Nexus VRM Christmas Special. It’s goona be so mind blowingly cool! Not just for DDR5, but for everything with voltage regulation. Those little modules are goona be like wax on Santas Sleigh. Gliding into that Yuletide goodness with everyone’s favorite seasonal semiconductor showdown!
A true exhibition of voltage transition!
Great video! Seeing your shadow on the right of the frame is distracting - it almost looks like someone keeps walking in and out of view.
Would be interesting to see a DDR3 top spec, vs. DDR4 base spec and top spec, to try to explain what we're seeing with DDR5 now; it's the earliest kits and therefore not the be-all-and-end-all of DDR5 tech.
I think the DDR5 kits with 7000MHZ and good timings are worth waiting for.
sadly that will be next year, maybe in q1 maybe in q2, and if you have all of your setup already bought, like me, with ddr5 mobo, can't wait that long. Hopefully 5600 kits come out by the end of this year atleast, which i'm now waiting for.. only missing ddr5 to boot everything up
I think it will be a couple of years before the prices level off on those kits, early adopters will have to pay a premium.
I was mostly just curious about cost per performance between DDR4 and 5. Great testing as always!
My cats would like to know how the box performs.
Depends how big the box is! If cat fits, cat sits.
@@GamersNexus That goes inline with Fergus and Milo's testing regime. I will let them know that their testing is on par.
2:04 - Hey Steve, loving the new set … looking forward to some behind the scenes footage on your relocation. 😄 But just wanted to say I found the "encroaching shadow of doom" distracting. I kept expecting somebody to appear, but it was just you stepping back, as in this example. Hope that helps. 🙏🏻👍🏻
"It's not fair to compare Intel on DDR5 to AMD on DDR4!" Advancements in the platform are also something that has to be considered when review the product. It's fine to want reviews/results from both memory types, but it's not inherently unfair to use the better of the supported memory types with Intel, it's just part of the equation when figuring out the value proposition.
2*16GByte Corsair Vengeance DDR5-5200 CL38 is the DDR5-RAM that was used here, in case you where wondering.
Very interested in seeing some benchmarks where you selectively disable cores of one type or the other. I'm seriously curious how they compare in both per core performance and the actual power usage per performance differences. Also would be nice to see exactly how few high performance cores you need in various games if you have the e-cores running.
Great video again. People are way too quick to rush to the newest technology when not always necessary. I don’t understand the memory timings that much yet, so will have to research that part more. Good to see the new studio, kinks and all. Would be neat to see a follow-up video or the studio progress.
Another reason to pass on 12th gen.
Beyond a full platform change, and the software compatibility issues all over the place ... upgrading when DDR5 is not just incredibly expensive, but still in its infancy technically, means early adopting with a comparably crappy kit now, just to change it up again when DDR5 matures, in the future when a whole new CPU generation will be out, probably better than gen12 ... be it gen 13 or 14 ... or the leap past 1700 socket.
I'm just not seeing the value if your sitting on a strong 9th gen or the better 10th gen.
It's strongly tempting for anyone with 6th gen or less. If X570 had thunderbolt, I'd upgrade my 5960x to a 5950x. At this point, I've waited so long that I'm just going to continue waiting until Zen 4 and 13th gen Intel are out next year.
I'm on an 8th gen and am feeling a little bit of pain from hardware creeping up in performance, but I feel I am forced to wait all the same by all of the generation's issues.
9th gen and 10th gen have crappy IPC compared to the Golden Cove in Alder Lake, i would upgrade by that alone.
You don’t need to use DDR5 though. It’s up to the motherboard you select.
The light that creates the body shadow on the right may need adjusting. Little distracting when you lean back.
I remember when good ram had lower CAS Latency
Don't too fixated on single parameter. Is your good RAM with lower CAS can beat current ram with higher CAS? Obviously they wouldn't because there are improvements that can be done by sacrificing some latency. It's moot point to complain only about single specific parameter on complex system.
thanks for the benchmarks Steve. it is very clear now that DDR5 simply isn't worth the hassle for an alder lake build. however now knowing that DDR4 is far better value and almost identical performance i would love to know how the Alder lake IMC can handle some of the top of the line memory kits. TLDR overclock the shit out of DDR4, drop timings as low as you can and re run these tests to see if the best of DDR4 can beat current DDR5. also i can't wait to see overclocking on the 12700k and 12900k.
"It was a little weird to see that question"
Not weird at all. AMD is better than Intel /s
Seriously, AMD has been the underdog (and underperforming) competitor for so long, that when they launched a good product line (Ryzen), it birthed a whole new generation of fanatical supporters.
Support AMD all you want (hello die hard fanboys), I will just buy the best value for my money. As of right now that seems to be the 5700G ironically, but I'm holding back because the competition is still so hot that any new tech launched this year is gonna be almost obsolete by next year.
GN is one of the most, if not the most, impartial reviewer I've seen on UA-cam.
When with performance being the same I think I'd lean slightly towards AMD due to socket lifecycle, fewer board chipsets and not locking overclocking to specific cpus.
Excellent video, as always!
Thank you.
Okay now do the fastest DDR4 kit vs the fastest DDR5 kit. DDR4 4400CL16-19-19-19 vs DDR5 6500 or higher for example
Thanks for the benchamark! On some other channel we could see ttha 6000MT ram did not improve performance compared to 5200. It will be interesting to see how DDR5 is going to scale in the future.
Sometimes new technology takes a while to become an obvious speed increase--especially when it needs to work in unison with other things. DDR5 is one of them. Also, the new-ish (not so new anymore) AGP spec. btw, 10:33 has an error. Far Cry 6 does not favor DDR4. :)
My follow up request would be, do games that are known to be RAM dependent, simulation games, like Factorio. Measuring updates per second rather than frames per second on a very large base.
Interesting! How about downclocking the "E" cores and overclocking the "P" cores.
Yeah I really want to see how this would work out with some well done overclocking on the p cores
I can't recall the test exactly but someone tested DDR5 4800 vs 6000 and showed a pretty big difference on the 12900K. Since 5800-6000 seems to be the slightly higher end like a 3200mhz kit is above spec for DDR4 I would love to see a comparison of the DDR5 5200 vs 6000 at least and maybe higher. If the results continued to be similar I think waiting till DDR5 drops before purchasing a kit next upgrade would be a better call and just get a D4 board.
kind of pointless the human eye can't even see more than 24fps anyway. Not even if you're a riced out benchmark queen.
Not true lol
The price:performance is exactly why I wanted to see the ddr4 vs ddr5 on this platform.
I agree that it’s an advantage of 12th gen having the ability to use ddr5 vs ddr4 on amd, but that was also one of the biggest things with zen3 vs 10/11th gen intel, a price:performance
When a cpu (like the new i5) is being touted as “x% ahead of a competitor for the same price” those claims usually aren’t considering the entire platform opportunity cost, but only the cpu itself
This was a really good comparison to do
It allows somebody to decide on a ddr4+12gen vs ddr4+zen3 vs ddr5+12gen and their pricing for each platform