Because if you don't the results gather so close together it doesn't make content? I haven't seen a 1080p screen since the Dell U2711 was a thing lmao. Should include some 4K RT with DLSS as all those things really hammer the CPU as well.
Displaying the percentage difference on the screen in the context is very much appreciated. I sometimes struggle processing numbers in speech, so this helps immensely. Thank you very much for incorporating this in the video!
Hey, so while AMD guarantees DDR5200 when you use 2 dimms in a 4 slot board, Intel is saying that 5600 is only officially supported in boards that have 2 memory slots TOTAL. If you have 4 slots ROUTED intel only guarantees the much lower DDR5 figures. This may be important for those seeking RMAs. Has AMD had the better memory speed guarantee this whole time?
@Level1Techs That's from 600 Series 12th gen documentation, not 700 Series 13th/14th gen documentation. Do you have documentation stating the same limitations for 700 Series motherboards paired with 13th/14th gen processors? It does seem Intel officially supports higher capacity (192gb) vs AMD (128gb) as well as higher maximum bandwidth of 89.6gb/s vs AMD 73.4gb/s.
Cuz of your and GNs videos, a while back I decided to sell my older 12600k rig and swap over to a 7800X3D on AM5, rather than upgrading to 13th or "14th" gen. Couldn't be happier with that decision. Cheers! Love your & Tim's content 😀
For me, realistically it's not worth it to swap over because AM5 has no option for DDR4. That was the reason I went with Intel to begin with, so I could reuse my 4X 32GB 3200 RAM I basically picked up for free.
@@GENKI_INU Yeah, I can understand that. But personally, I decided to just sell and regain some of the cost of that mobo + CPU + 32Gb of DDR4 (went for DDR4 as well at the time, cuz they were twice as cheap as DDR5). Got a nice AM5 board and hopefully there will be a nice zen6 or zen7 upgrade for it later down the line
that what im felt right now .. im using 12700k with ddr5 6200mt and i don't see my next path for my next upgrade with intel platform without changing motherboard ..im not going for 13 & 14 gen bcs those gen is really power hungry + stability issue
@@suzu9629 Yup. And with Intel, even ignoring the current problems, a significant upgrade would only be something like the 13 or 14900. But the price of those are pretty much the same as the 7800X3D + a new AM5 board. And at least with that you're getting some new CPU releases. Though at this point, for anyone thinking of switching, I'd recommend waiting a bit. Especially since we might get the new X3D variants later this year
Factorio would be an interesting benchmark. Because according to the devs, they are actually mainly slowed down by ram speed. Especially on high SPM bases the RAM speed gets more important.
@@ABYTE64 Ok so right now it has less players than these stellar games (/s) Brawlhalla, Age of Mythology: Retold, Unturned, Supermarket Together, Tapple - Idle Clicker. You think they should benchmark those games too because they are MORE popular than fucking factorio? Get outta here.
@@beachslap7359 With how many displays (and AI, obviously) we are currently shoving into cases, you might very well have reasons to do that in the future..
Steve not only sat down again, but also had to hide behind the b-roll while explaining the 1080p testing methodology. I totally get it. I'm wondering about the possible inclusion of 5800x3D in the results though.
Well it would be "appropriate gaming power" to the AMD 7600, so it would HAVE to count, and where the game is better under the 5800x3d than it is under the 7600, then THAT SPECIFIC GAME wants a bigger cache in that circumstance. So for two reasons it ought to be included.
this is a gem of a youtube channel and the information you share is extremely well explained in clear language- easy to understand. great channel. always really good uploads.
Only downside is that us sim racers that have or desire VR and triple screen setups are likely more GPU-bound than these tests suggest, but is always handy data regardless!
With DDR5 you can see big improvements especially with latency if you tweak the secondary timings. Buildzoid guides are priceless and worth the time to tweak your DDR5. Tweaking settings like tRC, tRFC, tFAW and others yield great improvements in overall system latency and performance in general.
@@andersjjensen nope, in games they are huge, call of duty has immense gains just look at my content with 12400f running Warzone i get same fps as other ppl i9 also, gains from memory tuning are bigger the lower the cache is, that's so why i can boost fps that much on 12400f that only has 18mb while AMD 7600 or 5600 have 32mb
Well tuned RAM can elevate CPU performance to the next level in gaming scenarios. For example, 7700 with tuned RAM is right on par with 7800X3D + 5200 MHz/CL40 kit. And just about 10% slower than 7800X3D with similar tuned RAM kit. I expect 9700(X) to be faster than 7800X3D once they both get tuned RAM kit.
Why? After that you will just wait for the 10k series and then 10kx3d and so on. Just buy a CPU now and use it for years instead of blindly waiting for the next one.
@@kwedl atm im using a 13700k and with all the talking Intel 13th and 14th gen issues i figure its time to move on... and why would i buy a 7800x3d when its soon to be outdated?
@@fooboomoo I bought DDR-5600, but left them at the default 4800 as I don't know how that overclock affects stability and speed on a 3D rendering workstation.
This would be interesting to compare memory sensitivity for production software, although traditionally, those workstations prefer stability over speed, and are often run at default memory speed/timings. @Hardwareunboxed Going to have any time for this?
Great content, but I wish there was more kits like the 6000 CL36 stuff. I would like to know what are exactly the key factors and were is the point of diminishing return.
The timing of these new videos have been great. I just bought a new CPU. I'm still on the fence on the RAM kit that I want to buy. Hopefully I'll be able to make a decision after watching this video.
Testing at supported JEDEC memory standards really showcases the benefit of 3D v-cache, assetto corsa had a 64% uplift from the 7700X@JEDEC to the 7800X3D@JEDEC. Running at JEDEC is the proper way to test CPUs at their stock configuration.
You should've used lower latency kits for the low speed RAM since there are 5600MHz CL28 kits available. 5200MHz RAM should run at a lower latency than the CL30 the 6000MHz kit was if you want to compare just the effect of extra speed and see if there's a real benefit. The 5200MHz kit was penalized by the high CL40 latency too not just the lower speed.
BIOS updates also can include security fixes, it's amazing how many IT "professionals" don't understand this and go off outdated best practices The thing about best practices, they change over time
Great content, as someone who has a 7800X3D with cheap CL40 memories, it's hard to find good RAM benchmarks. I'm currently running them at 5600, but I still wonder how much performance I could get by fine tuning the settings
Hey Steve, I'm running my RAM at 6400 with excellent stability. The burning question for me is should I dial my memory back and tighten the timings so that it runs at 6000Mhz? Thanks. AMD Ryzen 7800X3D ASUS ROG Strix X670E Gaming Wifi Samsung 990 Pro KLEVV Bolt V DDR5 32GB (2x16GB) 6400MHz CL 32 (KD5AGUA80-64A320H)
Before updating BIOS to the latest version, I like to see if there are reports of problems with the new BIOS. Nothing more frustrating than creating new PC problems for yourself.
As far as I know the recommendation is to upgrade the BIOS just when you build the PC, or if the BIOS update fixes a problem that you were experiencing.
For what platform? I assume Intel since the issues, but for AM5 there is no new BIOS as of right now. My Asrock board's only gotten increasingly stable with updates. On 3.01 right now.
As a follow up, are we to be looking forward to a primary and secondary timings comparison? I can understand the basic breakdown though really appreciate the more in-depth analysis.
If I'm not mistaken, Ryzen is more sensitive to latency than frequency. Would it be possible to add a 5200MT/s CL30 to the chart to see how close it gets to 6000MT/s CL30? They idea behind this is that it is possible to get some kits at 6000MT/s but with higher latency (cheaper) or higher clocked kits and tune down to lower latencies (if you want to be fancy and also Zen5 is going to support higher clocks)
It depends on a certain game. Cyberpunk, for instance, loves higher frequency more than tighter timings. While CS2 tend to benefit from tigher timings.
No greeting, no intro, pedal to the metal spewing straight facts! I LOVE IT. ❤❤😍😍 Given, the flow was interrupted by the ad spot, but ya gotta get paid too
I'd be interested to understand how much of the performance comes from the improvement in memory clock vs the reduction in cycles from a lower latency orientated kit
Im using 6200 cl30 2200 fclk myself. 6400 cl30 2167fclk is also stable for my 7600x paired with b650e-f. But I'm waiting for my dedicated ram cooler cause at 6400 it gets up to 55c. Which causes instabilitys with maxed out trefi
@@mruczyslaw50 it's actually running cl28 @ 6200 mybad. But no. 2200 does indeed run faster. Why would running a lower fclk run faster? Unless it's becoming unstable that doesn't make much sense.
Thanks for making this video....I have a 7800x3d while using the xmp profile and was wondering if I really adding extra heat and power for nothing. It help me confirm, because maybe I was doing something wrong. anyway thanks
Interesting stuff. Id love to see testing done with higher speeds though. The 6000 isn't really "high speed" but pretty standard these days for AMDs, but is it worth spending more over the standard 6000 for 7200/7600/8000?
@@Kapono5150 Premium look but Corsair doesn’t want to show the SDRAM manufacturer for their kits for some reason unless they changed that. I remember even Buildzoid ranting about this.
Dominator RAM has always been a very poor $ per performance part. If stuff looking nice is important online product images are free. Waste not want not.
is ddr5 7200 worth it on a 13600k ? basically my motherboard fried my ram so i rmaed my mobo and i will have buy new ram. is 6800 a more reasonable spec ? also any reason to go above 32 gb ?
Thank you Steve for this test, but as a sanity check could you please include a small segment with the same test in the 9gen and 9gen X3D reviews? Thanks Steve!
I currently use 1.15Volt on the soc cpu (7800x3d) with 6000mhz cl30... To tell the truth I don't know if I can go lower because I haven't tried. And with a maximum peak of 1.12 volts on the cpu with 5.05Ghz... I also don't know if it can go lower. I'm happy because the temperatures are fantastic. Let's talk about what the Motherboard wants to use from the factory 1.27V on the Soc and 1.19 on the cpu.... And it is only used to heat up.
I've a question, do diferent dies matter for DDR5 ? Like in DDR4 we had b-die (which i've currently got) do DDR5 has it's equivalent that i should watch out for, i intend to swap platforms for 9800x3d or smg like that.
I got 6% increase moving from DDR5-5200 to DDR5-6400 using my 7800X3D. 6% was enough to matter to me because the price wasn't that different. I use a MSI Tomahawk X670E.
Very interesting results! Could I be so brazen as to ask for further testing separating the effects of memory clocks vs latency? I'm always wondering which to prioritize when choosing between unequal kits.
I got my ram tuned to 6400 cl 30 and noticed a huge bump in 1% lows vs stock. In one benchmark, it went up about 40 FPS. My cpu is a 7800x3d. Was well worth the day it took to optimize.
The importance of updating the bios cannot be overstated. I bought my current PC a year ago (7600x/32gb Corsair V@6400mhz/rtx4070) and I was getting frequent crashes when gaming which I attributed to my GPU being faulty until I found out much later that xmp was the issue and a bios update that came out 2 months after I purchased fixed it for me.
Remember to always go for CL30, or the 6400 CL32 bin (basically the same) if you want to punch in the timings yourself. It's well worth the tiny tiny premium over the worse bin. And setting some super safe subtimings will most likely give you another 1-2% on top of that as well. Something you may as well do since it's utterly free performance.
Before you hit the realistic limits of A or M die, it's but the motherboard that will hit the possible limits Paying upwards of $5 for a whole lotta nothing over an equivalent CL38 kit is pointless and will only cost people extra.. for nothing Getting anything that guarantees Hynix die is already sufficient enough Don't promote for people to buy something they legitimately won't even benefit from
For 6000MHz speeds, CL30-36, 32-36, 30-38, 32-38, and the couple of Kingston models with Hynix inside all behave the same, the only thing that matters for buying DDR5 memory is to have a significant gap between the first and 2nd latency number, so that you're sure you're getting SK Hynix, that's it.
I feel like this needs more clarity on what is making the difference. In the UK the pricing seems to make the most sense at 6000mhz CL36. It's negligibly more (under £4) than 5200mhz CL40, however it's then a 20-25% jump to go from CL36 to CL30. Would that be worth it or do you gain most of the difference with 6000/36? Was the same with DDR4, CL18 was much cheaper than CL16 for the same brand and same speed here..
I’m wondering this too, as I’m getting the feeling this was tested using two different memory kits running at their max rated speeds but with different CAS latencies rather than the same kit with a different XMP profile set. Would check the test setup section that HUB normally includes but it’s absent this time. Still, I don’t think this is a like for like comparison of pure memory clock speeds and more the cumulative effects of higher speed plus better CL you get by splurging on better memory kits. Which is fine if you approach it from a price to performance gain angle but that’s not really what this video was, as it was framed purely in the context of memory speed while ignoring the other variable.
Well I mean, if you buy a 5600 CL 40 ram kit, you can go look up the timings for a 6000 cl30 punch that in for free and save $50+ on ram. Did this exact thing on my 13900k, and went from 6400 cl32 to 7800 cl30 for free and saved over $100. Knowledge is power
People not having 3DVcache should get CS latency of half the number of hundreds in frequency. So 6000 CL30 for Ryzen 7000 series. Here he's speaking about bandwidth, but most of the time improvement is probably more based on latency than bandwidth, which the higher frequency improves at equal timings, and higher frequency also helps latency on the memory controller side if latency in cycles stay the same.
i have question, my gpu does not display on ANY bios version except for the oldest version available to consumers. its very weird as i lose a lot of newer features because of this, none of the hardware is faulty, and all runs perfect on the oldest bios version. MOBO: b650m gaming plus wifi MSI GPU: XFX RX 7900GRE 16GB CPU: Ryzen 5 7600 RAM: DDR5 lexar 2x16gb 6400mhz (i do not have xmp enabled as it is to unstable on my pc so it is set to 5600mhz) PSU: XPG core reactor 2 850W white SSD lexar nm790 2tb nvme
TLDW version: It depends on the game and system as it always has, if your game is using a mass of memory changes then faster ram helps. One thing that most people do not understand these days though is that with especially Nvidia's reluctance to give us actually good sizes on the video cards this becomes more of an impact as VRAM is forced to offload to System RAM and the two have to swap when required far more often. Those shifts in memory are often when you start to see some of the hitches caused by memory itself.
Why are avoiding the GPU benchmark by running 1080p but still testing more GPU demanding games, isn’t it better to test more CPU intense games in this scenario? (Games like Planet Coaster, Cities Skylines, Besiege, Teardown etc.)
Why 5600 c40? The jedec base for 5600 is c46. Same story but not as big of a difference for 5200, it's c42 at jedec. Did you just disable xmp then manually raise the multiplier from 4800 up to 5200 or 5600 and keep the timings from 4800?
CL30 this, CL30 that But does it make a real difference, especially when looking at manual overclocking? Most definitely not. 6000 Mbps CL38 or even CL36 for a handful of kits (e.g. Patriot Viper Venom PVV532G600C36K) !but not all! will suffice The real focus when choosing which memory to pick is it both being an EXPO/XMP capable kit (looking at you CL48 T-Create Classic) AND uses Hynix M or A-die chips. paid 99€ for my Viper Venom kit in Germany, which at the time of purchase was unmatched, and cannot complain one bit with buildzoid's basic Zen 4 DDR5 timings applied. Couldn't do this on basically any of Corsair's offerings despite it being such a popular choice All in all a great video nonetheless, yet these details really do end up annoying me for when it comes to inform the unknowing about what memory to choose for their new build IMPORTANT MENTION: Please do not use more or less than two sticks of memory that were purchased as a kit!
Right now you are in luck that SP032GXLWU60AFDEAE is the current cheapest 6000 Mbps 2x16 GB kit available in the US for $84.97 However this isn't always the case which brings me back to my original point The most that's reasonable to pay extra is a whole $1 over a non-CL30 kit as long as you are able to ensure the use of Hynix A or M die
They know about the Buildzoids easy timings for hynix m- and a-die because they have literally tested them with games. You’re correct about CL but you do actually have to look at the primary timings because they can help you see if the kit is hynix or not. Some RAM manufacturers do show the SDRAM manufacturer for the kit like G.Skill.
Hello Hardware Unboxed! I Purchased Ryzen 7 7800x3d and pre ordered a Corsair Dominator Platinum 6200 Mhz CL36 to upgrade my Am4 system to Am5. but lately I discovered that the Ram I bought was for Intel, SKU: cmt32gx5m2x6200c36w Question is: Will it work? or what will happen if I combine these two? I hope to get answers from you! Big Thanks!
We've tried three different DDR5-6000 CL30 kits on every X670/B650 board and they work on every single one except for the 'Asus Prime B650M-A AX6-CSM WiFi'.
I've got a suggestion for a follow up for this - is more memory or faster memory better? I now have 64GB of DDR5 set up as 4x16 sticks. However, my board doesn't seem to like running all four stick with EXPO on. Would I be better off taking two sticks out and running at full speed?
Depending on the speed difference and the type of application. If we're talking gaming, 32GB should be plenty and it'll be better to run at higher speed. Besides, this is something you can test yourself, right? Just take out two modules and check how well your games run.
One thing these mass media guys dont tell you is dont just throw any ram on your motherboard cpu combo. Always check what the optinsl ram is for your cpu/motherboard. Throwing 6000 on a cpu that doesnt work well with it wont do you any good. Like the 13400f for you budget guys it runs better on ddr4 3200 then it does even on 4800 ddr5. And it wont use more then either. So you can use 3200 or 4800 ddr5 thats your choices. Just using this as an exmaple. Its why ram comparisons mean nothing to me and shouldnt to you either because the cou they sre testing on will always like a certain one more then another. Doesnt mean your cpu will.
@@Hardwareunboxed I'm on about the comparisons yall do with a certain cpu and memory in videos like this and try going oh this is better or that's better when it isn't it's all motherboard/cpu dependant. That's all
In one of your Zen 5 reviews, you compared the jump from Zen 3 to Zen 4 to the jump from Zen 4 to Zen 5. I was expecting you to mention that there was no new DDR technology between the latter. I suspect that much of the performance jump on the former was due to the move from DDR4 to DDR5.
Another thing to keep in mind is that if you find a deal on even higher spec memory, you should have little issue dialing it back to 6000. I did that with my 6400 kit
Would be interested to see some testing of EXPO vs XMP on AM5. I suspect the difference will be negligible but would still be useful to know as XMP rated kits still seem far more common and therefore more affordable than EXPO kits despite the fact that intel's high end is now a pretty hard sell on desktop.
A decent portion of chips can run 6400 and UCLK at 2133 with higher SOC voltage, usually 1.25-1.3v, curious what the benefit is in FPS whether increased voltage is worth it or not.
I fiddled around with Ram the past week and currently I am using a Frankenstein setup of Ram. I use both a original DDR4 16GB 3200 CL16 kit AND a 32GB 4400 Cl26 kit but regulated both to down to 3200 CL15 and this dramatically improved performance and especially drops / lows on my 5800X3D. My new rule of thumb would be go for the max "officially supported" ram speed and shop for lower latency kits, depending on your budget. Latency seems really important for higher framerates.
I went with the black g skill ram that is cl30 6,000mhz 64 gigs of ddr5 but I have 4 sticks with my all white 4090/ 7800x3d build so the ram is slower but I havent updated the bios in a year but I will update it soon.
I watched a test video that focused on testing MiHoYo games. The results showed that X3D, with its large cache, can effectively improve single-core performance and address issues with insufficient memory speed. Therefore, for certain games, Intel cpu should be paired with high-frequency memory, while AMD X3D cpu can perform well with just 6000MHz/CL30 memory.
My Ram stopped working at its rated speed of 3600Mhz. I now use it at 2666Mhz, the PC won't boot at any faster speed. I have noticed no difference in my PC performance or in games (at 4k) at all. Although I do run at 60fps, I don't notice faster fps.
"If you still believe cpu limited cpu benchmarking is useless, click away now." ~Spreadsheet Steve I know you really wanted to say something stronger than "click away now" but hopefully the people wanting 4k cpu benchmarks get the message. Hopefully. They are a stubborn bunch who aren't good with instructions and simple concepts but they may figure out how cpu benchmarks work eventually.
I bought a 12600k not long after they came out and when ddr5 when in its infancy. Im running 32gb of 5600mhz cl34. Would it be worth it to upgrade to something like 6400mhz cl32 even though gpu usage is usually 97-100% at 1440p with a 4070ti super? Should i just try to learn to overclock the kit i have?
Yes, you should try to overclock your kit. I use DDR4 3600 CL16 kit with my 12600K and I would think DDR5 would be no different in overclocking it to the speeds of higher clocked RAM. Your current memory of 5600MT at CL34 = a memory latency of 12.14 seconds. I overclocked my DDR4 kit to 3733 MT and tightened the timings down to CL14 which gives me a memory latency of 7.5 seconds. Maybe you should try to OC your DDR5 to 6000 and tighten the timings to CL32 or CL30 if possible, that would bring your memory latency down to 10 seconds.
Interesting results and some were expected. Would it be possible / hard to also put the % for the min fps in the darker part of the bar? in some tests the difference was quite big in others it was not. I can understand it will get more messy with more data in the graphs but i like all that info. Now I'm doing the calculations in my head or when you tell us the difference but when they are also shown it's easier to compare. but i can imagine you also have good reasons not to include them on screen :)
can i simply buy a fast expo ram kit, e.g. 8000, for an amd cpu with a good offer (more future-proof e.g. for ryzen 9000 and 10000) and, if necessary, lower the speed until it is stable? or is this only possible if you manually deal with the dozens of subtimings in the bios and know what you are doing?
For mid or lower end build a 5200mhz ddr5 is better as you can use the price diff to increase your gpu budget. Here a used 6700xt & 6800 has a similar price gap of 32gb ddr5 5200mhz & 6000mhz
Hello first of all thanks for the content your team post frequently ! I'd like to know about your ram test like you've tested DDR5 memory but the faster ram have also better CL so what's the point of the testing they are for sure better if you calculate the absolute latency they are 50% faster like10ns vs 15ns
Steve gets saltier every time he has to explain why he's testing in 1080p 😂
A year from now it's going to be real bad.
@@Hardwareunboxed 😅your Ryzen 9000 Reviews can't come soon enough then.
Lmfao
Because if you don't the results gather so close together it doesn't make content? I haven't seen a 1080p screen since the Dell U2711 was a thing lmao. Should include some 4K RT with DLSS as all those things really hammer the CPU as well.
@@HardwareunboxedYou should do a test at 720P just for sh*ts and giggles😅😂
Displaying the percentage difference on the screen in the context is very much appreciated. I sometimes struggle processing numbers in speech, so this helps immensely. Thank you very much for incorporating this in the video!
Ditto! Getting both speech and visual cues is king.
Couldnt agree more
Gamers Nexus charts gave me PTSD!
Hey, so while AMD guarantees DDR5200 when you use 2 dimms in a 4 slot board, Intel is saying that 5600 is only officially supported in boards that have 2 memory slots TOTAL. If you have 4 slots ROUTED intel only guarantees the much lower DDR5 figures. This may be important for those seeking RMAs. Has AMD had the better memory speed guarantee this whole time?
@Level1Techs That's from 600 Series 12th gen documentation, not 700 Series 13th/14th gen documentation. Do you have documentation stating the same limitations for 700 Series motherboards paired with 13th/14th gen processors? It does seem Intel officially supports higher capacity (192gb) vs AMD (128gb) as well as higher maximum bandwidth of 89.6gb/s vs AMD 73.4gb/s.
That's pretty cheeky. Other than some crazy LN2 boards can you buy anything bigger than ITX with only two dimm slots?
I don't think what they guarantee matters 😉 we tried these CPUs and we can expect what we will get when we either go Intel or AMD
@@duckrutt There are some cheap matx ones.
honestly advertising speeds that you can only get with 2 slots in a 4 slot board is just deceptive
Cuz of your and GNs videos, a while back I decided to sell my older 12600k rig and swap over to a 7800X3D on AM5, rather than upgrading to 13th or "14th" gen. Couldn't be happier with that decision.
Cheers! Love your & Tim's content 😀
For me, realistically it's not worth it to swap over because AM5 has no option for DDR4.
That was the reason I went with Intel to begin with, so I could reuse my 4X 32GB 3200 RAM I basically picked up for free.
@@GENKI_INU Yeah, I can understand that. But personally, I decided to just sell and regain some of the cost of that mobo + CPU + 32Gb of DDR4 (went for DDR4 as well at the time, cuz they were twice as cheap as DDR5). Got a nice AM5 board and hopefully there will be a nice zen6 or zen7 upgrade for it later down the line
that what im felt right now .. im using 12700k with ddr5 6200mt and i don't see my next path for my next upgrade with intel platform without changing motherboard ..im not going for 13 & 14 gen bcs those gen is really power hungry + stability issue
@@suzu9629 Yup. And with Intel, even ignoring the current problems, a significant upgrade would only be something like the 13 or 14900. But the price of those are pretty much the same as the 7800X3D + a new AM5 board. And at least with that you're getting some new CPU releases.
Though at this point, for anyone thinking of switching, I'd recommend waiting a bit. Especially since we might get the new X3D variants later this year
Ditto 👍
Factorio would be an interesting benchmark. Because according to the devs, they are actually mainly slowed down by ram speed. Especially on high SPM bases the RAM speed gets more important.
Yeah but niche games almost never get the benchmarking spotlight put on them. You could always do the test yourself /shrug
While I am in it, the subset of people that want high FPS for shooters and also run 10k SPM Factorio bases is not very big.
@@Azureskies01me wondering what the fuck homeworld 3 is lmao
@@Azureskies01 Factorio is quite popular
@@ABYTE64 Ok so right now it has less players than these stellar games (/s)
Brawlhalla, Age of Mythology: Retold, Unturned, Supermarket Together, Tapple - Idle Clicker.
You think they should benchmark those games too because they are MORE popular than fucking factorio?
Get outta here.
Nothing gets my frequency higher than scissoring RAM DIMMs.
Ayo, that's the dirtiest tech joke I've seen.
Is that an IHS in your pocket or are you trying to see what dem DIMMs do
Had to like your comment to get it to 69 likes. 😜
You are sick!!!
DON'T STOOOOP!
First thing you should do on a PC build is download the latest drivers on your Thermal Paste.
I always do it right after overclocking my case.
That's why we like the Thermal Grizzly Kryosheet, doesn't need drivers.
@@beachslap7359 With how many displays (and AI, obviously) we are currently shoving into cases, you might very well have reasons to do that in the future..
also make sure your cooler firmware is up to date (especially with air coolers)
Screw long bolts with confidence
Steve not only sat down again, but also had to hide behind the b-roll while explaining the 1080p testing methodology. I totally get it. I'm wondering about the possible inclusion of 5800x3D in the results though.
Well it would be "appropriate gaming power" to the AMD 7600, so it would HAVE to count, and where the game is better under the 5800x3d than it is under the 7600, then THAT SPECIFIC GAME wants a bigger cache in that circumstance. So for two reasons it ought to be included.
Well even the 5800x3D doesn't really benefit all to much from faster memory either, that extra cache really offsets the memory limitations.
Doesn't quite make sense here since this is a DDR5 memory comparison. AM4 doesn't have DDR5 support.
@@dex2531 5800X3D could do it.
/jk
I'm waiting for the 3700x inclusion, screw those plebs who bought a 3600!
updating the BASS
He said BOSS, listen more carefully nest time!
@@paulanderegg5536 got it, *Updates BUSS*
BAAS
Also turn up 4WATT.
update the BUS
this is a gem of a youtube channel and the information you share is extremely well explained in clear language- easy to understand. great channel. always really good uploads.
Another reason why the 7800x3d is the go to CPU for gaming.
I got it for 190$ is it a steal?
@@qdo4more like mug
@@qdo4yeah that’s a screaming good deal! I got mine for 250 and was super happy! Got me a asrock taichi for 200 as well!
@@qdo4 What dark deal did you accept to get such a price?
@@qdo4 no that's a robbery. damn where'd u find that?
I appreciate! that you add the ACC! To your tests. Thank you for that!👍👍👍
Thanks for still testing ACC :) It's hard sometimes to be a Sim-racer and not know what to expect from new hardware. Much appreciated!!!
Only downside is that us sim racers that have or desire VR and triple screen setups are likely more GPU-bound than these tests suggest, but is always handy data regardless!
Been loving the ram videos lately. Now we need a 5000mhz ddr4 vs 5000mhz ddr5 video.
With DDR5 you can see big improvements especially with latency if you tweak the secondary timings. Buildzoid guides are priceless and worth the time to tweak your DDR5. Tweaking settings like tRC, tRFC, tFAW and others yield great improvements in overall system latency and performance in general.
tREFI mostly
Just be aware that the gains from doing that are mostly just visible in synthetic benchmarks and a few scientific calculation tools.
@@andersjjensen nope, in games they are huge, call of duty has immense gains just look at my content with 12400f running Warzone i get same fps as other ppl i9
also, gains from memory tuning are bigger the lower the cache is, that's so why i can boost fps that much on 12400f that only has 18mb while AMD 7600 or 5600 have 32mb
Well tuned RAM can elevate CPU performance to the next level in gaming scenarios. For example, 7700 with tuned RAM is right on par with 7800X3D + 5200 MHz/CL40 kit. And just about 10% slower than 7800X3D with similar tuned RAM kit.
I expect 9700(X) to be faster than 7800X3D once they both get tuned RAM kit.
@@stangamer1151 Spot on!
THANKS FOR INCLUDING COUNTER STRIKE BENCHMARKS
I always make sure to get the sweetspot/highest speed and tightest timing recommended, it's not that expensive.
wake me when the 9000 series x3d chips release
no i won't wake you up. enjoy sleeping.
Why? After that you will just wait for the 10k series and then 10kx3d and so on. Just buy a CPU now and use it for years instead of blindly waiting for the next one.
@@kaimojepaslt wake me up when steve includes the 5800x3d
Sleep through you life, consume products, get excited for the next product.
@@kwedl atm im using a 13700k and with all the talking Intel 13th and 14th gen issues i figure its time to move on... and why would i buy a 7800x3d when its soon to be outdated?
Nice video! I would like to see the same test setups, but with productivity apps (with a focus on 3D rendering).
I second this, could not find any useful benchmarks out there. Went for 5200 over 6000 and now doubt my decision after watching this video 😂
@@fooboomoo I bought DDR-5600, but left them at the default 4800 as I don't know how that overclock affects stability and speed on a 3D rendering workstation.
This would be interesting to compare memory sensitivity for production software, although traditionally, those workstations prefer stability over speed, and are often run at default memory speed/timings. @Hardwareunboxed Going to have any time for this?
Great content, but I wish there was more kits like the 6000 CL36 stuff. I would like to know what are exactly the key factors and were is the point of diminishing return.
The timing of these new videos have been great. I just bought a new CPU. I'm still on the fence on the RAM kit that I want to buy. Hopefully I'll be able to make a decision after watching this video.
Great video. I hope you do an update with the new CPUs and higher memory speed support and whether it's more/less impactful there. Thanks.
Testing at supported JEDEC memory standards really showcases the benefit of 3D v-cache, assetto corsa had a 64% uplift from the 7700X@JEDEC to the 7800X3D@JEDEC. Running at JEDEC is the proper way to test CPUs at their stock configuration.
Should you not be running lower speed modules with lower CL, it is unclear of the test if it is the speed or timings that affect the fps.
You should've used lower latency kits for the low speed RAM since there are 5600MHz CL28 kits available. 5200MHz RAM should run at a lower latency than the CL30 the 6000MHz kit was if you want to compare just the effect of extra speed and see if there's a real benefit. The 5200MHz kit was penalized by the high CL40 latency too not just the lower speed.
BIOS updates also can include security fixes, it's amazing how many IT "professionals" don't understand this and go off outdated best practices
The thing about best practices, they change over time
Great content, as someone who has a 7800X3D with cheap CL40 memories, it's hard to find good RAM benchmarks. I'm currently running them at 5600, but I still wonder how much performance I could get by fine tuning the settings
Probably not that much. My guess its in the middle of 6000 and 5200 results :)
Hey Steve, I'm running my RAM at 6400 with excellent stability. The burning question for me is should I dial my memory back and tighten the timings so that it runs at 6000Mhz? Thanks.
AMD Ryzen 7800X3D
ASUS ROG Strix X670E Gaming Wifi
Samsung 990 Pro
KLEVV Bolt V DDR5 32GB (2x16GB) 6400MHz CL 32 (KD5AGUA80-64A320H)
Before updating BIOS to the latest version, I like to see if there are reports of problems with the new BIOS. Nothing more frustrating than creating new PC problems for yourself.
True because for my Asrock B450m there're 2 new Bios updates that fixes security vulnerabilities but lowers 1% lows and memory stability.
Yup. I didn't update the Bios in mine out of the box.
As far as I know the recommendation is to upgrade the BIOS just when you build the PC, or if the BIOS update fixes a problem that you were experiencing.
@@itisabird I would add vulnerabilites and burning sockets/CPU’s last year.
For what platform? I assume Intel since the issues, but for AM5 there is no new BIOS as of right now. My Asrock board's only gotten increasingly stable with updates. On 3.01 right now.
As a follow up, are we to be looking forward to a primary and secondary timings comparison?
I can understand the basic breakdown though really appreciate the more in-depth analysis.
If I'm not mistaken, Ryzen is more sensitive to latency than frequency.
Would it be possible to add a 5200MT/s CL30 to the chart to see how close it gets to 6000MT/s CL30?
They idea behind this is that it is possible to get some kits at 6000MT/s but with higher latency (cheaper) or higher clocked kits and tune down to lower latencies (if you want to be fancy and also Zen5 is going to support higher clocks)
Yes CL is more important, at least in AMD plataform.
It depends on a certain game. Cyberpunk, for instance, loves higher frequency more than tighter timings. While CS2 tend to benefit from tigher timings.
Great review! Did you play against SnT variant with Sneak Attack and the new Ulamog?
No greeting, no intro, pedal to the metal spewing straight facts! I LOVE IT. ❤❤😍😍
Given, the flow was interrupted by the ad spot, but ya gotta get paid too
And paid, hopefully!
@@krjal3038 Forgot english doesnt like to spell things like they're said.
I'd be interested to understand how much of the performance comes from the improvement in memory clock vs the reduction in cycles from a lower latency orientated kit
I used 6200MHz 1:1 by overclocking FCLK by 66MHz, on Ryzen 5 7500F, some people successfully used 6400MHz at 2133MHz FCLK but mine wasn't stable.
Latest BIOS? That can improve stability.
i have 7950x cpu , 2133 fabric . 6400 cl32 mem . wasn't stable but updates made it 10000% stable. when it was unstable i just ran 6000 cl28
Im using 6200 cl30 2200 fclk myself. 6400 cl30 2167fclk is also stable for my 7600x paired with b650e-f. But I'm waiting for my dedicated ram cooler cause at 6400 it gets up to 55c. Which causes instabilitys with maxed out trefi
@@НААТ Then 6200 at 2066MHz FCLK should be faster because of 1:1 ratio
@@mruczyslaw50 it's actually running cl28 @ 6200 mybad. But no. 2200 does indeed run faster. Why would running a lower fclk run faster? Unless it's becoming unstable that doesn't make much sense.
Can you do the same test but with 7200Mhz RAM on the 7800X3D? I want to see if it's worth buying
Thanks for making this video....I have a 7800x3d while using the xmp profile and was wondering if I really adding extra heat and power for nothing. It help me confirm, because maybe I was doing something wrong. anyway thanks
Interesting stuff. Id love to see testing done with higher speeds though. The 6000 isn't really "high speed" but pretty standard these days for AMDs, but is it worth spending more over the standard 6000 for 7200/7600/8000?
Corsair Dominator Titanium is very premium kit. Love them
@@Kapono5150 Premium look but Corsair doesn’t want to show the SDRAM manufacturer for their kits for some reason unless they changed that. I remember even Buildzoid ranting about this.
Dominator RAM has always been a very poor $ per performance part. If stuff looking nice is important online product images are free. Waste not want not.
is ddr5 7200 worth it on a 13600k ? basically my motherboard fried my ram so i rmaed my mobo and i will have buy new ram. is 6800 a more reasonable spec ? also any reason to go above 32 gb ?
Thank you Steve for this test, but as a sanity check could you please include a small segment with the same test in the 9gen and 9gen X3D reviews? Thanks Steve!
Thanks Steve.
@@grievesy83 Back to you, Steve.
Being 6400 mts the new sweet spot it could have been interesting to compare it to 6000 cl30 vs cl32 of the 6400 mts
May be 5% to 7% percent improvement
I currently use 1.15Volt on the soc cpu (7800x3d) with 6000mhz cl30... To tell the truth I don't know if I can go lower because I haven't tried. And with a maximum peak of 1.12 volts on the cpu with 5.05Ghz... I also don't know if it can go lower. I'm happy because the temperatures are fantastic. Let's talk about what the Motherboard wants to use from the factory 1.27V on the Soc and 1.19 on the cpu.... And it is only used to heat up.
That antec case is sexy
I've a question, do diferent dies matter for DDR5 ? Like in DDR4 we had b-die (which i've currently got) do DDR5 has it's equivalent that i should watch out for, i intend to swap platforms for 9800x3d or smg like that.
Thanks Steve 👍
I got 6% increase moving from DDR5-5200 to DDR5-6400 using my 7800X3D. 6% was enough to matter to me because the price wasn't that different. I use a MSI Tomahawk X670E.
So, how can I download more ram?
Very interesting results! Could I be so brazen as to ask for further testing separating the effects of memory clocks vs latency? I'm always wondering which to prioritize when choosing between unequal kits.
People who say X3D CPUs don't care about good memory can shut up now. Sure it's not a huge deal in all games but it's still an advantage.
"Good" memory?
Or fast memory?
I got my ram tuned to 6400 cl 30 and noticed a huge bump in 1% lows vs stock. In one benchmark, it went up about 40 FPS. My cpu is a 7800x3d. Was well worth the day it took to optimize.
But what about same speed memory with different latency?
The importance of updating the bios cannot be overstated. I bought my current PC a year ago (7600x/32gb Corsair V@6400mhz/rtx4070) and I was getting frequent crashes when gaming which I attributed to my GPU being faulty until I found out much later that xmp was the issue and a bios update that came out 2 months after I purchased fixed it for me.
Remember to always go for CL30, or the 6400 CL32 bin (basically the same) if you want to punch in the timings yourself. It's well worth the tiny tiny premium over the worse bin. And setting some super safe subtimings will most likely give you another 1-2% on top of that as well. Something you may as well do since it's utterly free performance.
Before you hit the realistic limits of A or M die, it's but the motherboard that will hit the possible limits
Paying upwards of $5 for a whole lotta nothing over an equivalent CL38 kit is pointless and will only cost people extra.. for nothing
Getting anything that guarantees Hynix die is already sufficient enough
Don't promote for people to buy something they legitimately won't even benefit from
For 6000MHz speeds, CL30-36, 32-36, 30-38, 32-38, and the couple of Kingston models with Hynix inside all behave the same, the only thing that matters for buying DDR5 memory is to have a significant gap between the first and 2nd latency number, so that you're sure you're getting SK Hynix, that's it.
I feel like this needs more clarity on what is making the difference.
In the UK the pricing seems to make the most sense at 6000mhz CL36. It's negligibly more (under £4) than 5200mhz CL40, however it's then a 20-25% jump to go from CL36 to CL30.
Would that be worth it or do you gain most of the difference with 6000/36?
Was the same with DDR4, CL18 was much cheaper than CL16 for the same brand and same speed here..
How much of this is speed vs CL? cl 40 vs cl 32 or 30 when comparing speed?
I’m wondering this too, as I’m getting the feeling this was tested using two different memory kits running at their max rated speeds but with different CAS latencies rather than the same kit with a different XMP profile set. Would check the test setup section that HUB normally includes but it’s absent this time. Still, I don’t think this is a like for like comparison of pure memory clock speeds and more the cumulative effects of higher speed plus better CL you get by splurging on better memory kits. Which is fine if you approach it from a price to performance gain angle but that’s not really what this video was, as it was framed purely in the context of memory speed while ignoring the other variable.
Well I mean, if you buy a 5600 CL 40 ram kit, you can go look up the timings for a 6000 cl30 punch that in for free and save $50+ on ram. Did this exact thing on my 13900k, and went from 6400 cl32 to 7800 cl30 for free and saved over $100. Knowledge is power
do you think that 6000C30 kit will be "OLD" for the new AMD 9000 series chip? should we upgrade 6000 C30 kit for the new processors?
People not having 3DVcache should get CS latency of half the number of hundreds in frequency. So 6000 CL30 for Ryzen 7000 series. Here he's speaking about bandwidth, but most of the time improvement is probably more based on latency than bandwidth, which the higher frequency improves at equal timings, and higher frequency also helps latency on the memory controller side if latency in cycles stay the same.
So is the majority of this gains due to the mhz increase or the reduced timings. What has bigger impact?
Superb, informative content as always
i have question, my gpu does not display on ANY bios version except for the oldest version available to consumers. its very weird as i lose a lot of newer features because of this, none of the hardware is faulty, and all runs perfect on the oldest bios version. MOBO: b650m gaming plus wifi MSI GPU: XFX RX 7900GRE 16GB CPU: Ryzen 5 7600 RAM: DDR5 lexar 2x16gb 6400mhz (i do not have xmp enabled as it is to unstable on my pc so it is set to 5600mhz) PSU: XPG core reactor 2 850W white SSD lexar nm790 2tb nvme
Thanks for video.
Have a nice day!
Was the cyberpunk test done with high crowd density? My 12700 drop to sub 60 during gameplay sometimes.
I wonder if Steves children understand the reasons of testing @1080p. Wrath of Steve. 🤔😂
TLDW version: It depends on the game and system as it always has, if your game is using a mass of memory changes then faster ram helps. One thing that most people do not understand these days though is that with especially Nvidia's reluctance to give us actually good sizes on the video cards this becomes more of an impact as VRAM is forced to offload to System RAM and the two have to swap when required far more often.
Those shifts in memory are often when you start to see some of the hitches caused by memory itself.
I think it's time to consider a dark mode intro because I just got flashbanged by my 65" OLED in the middle of the night. 😵💫
@@InvntdXNEWROMAN in windows adjust the sdr/HDR slider and you can have sdr much dimmer while still keeping the max available brightness for HDR.
@@ronjatternot a problem, LG TVs has different settings between SDR and HDR
@@Frigobar_Ranamelonico not if you leave windows in hdr (which is the correct way to do things)
@@ronjatter i onestly Don't like how auto HDR works on windows.
I prefer to use it only when I'm outputting HDR content
what should I upgrade my tforce 36cl 5200 mhz and 7700x to ? should I get 9700x or 7800x3d and what memory with them/
Why are avoiding the GPU benchmark by running 1080p but still testing more GPU demanding games, isn’t it better to test more CPU intense games in this scenario? (Games like Planet Coaster, Cities Skylines, Besiege, Teardown etc.)
Why 5600 c40? The jedec base for 5600 is c46. Same story but not as big of a difference for 5200, it's c42 at jedec. Did you just disable xmp then manually raise the multiplier from 4800 up to 5200 or 5600 and keep the timings from 4800?
CL30 this, CL30 that
But does it make a real difference, especially when looking at manual overclocking? Most definitely not.
6000 Mbps CL38 or even CL36 for a handful of kits (e.g. Patriot Viper Venom PVV532G600C36K) !but not all! will suffice
The real focus when choosing which memory to pick is it both being an EXPO/XMP capable kit (looking at you CL48 T-Create Classic) AND uses Hynix M or A-die chips.
paid 99€ for my Viper Venom kit in Germany, which at the time of purchase was unmatched, and cannot complain one bit with buildzoid's basic Zen 4 DDR5 timings applied. Couldn't do this on basically any of Corsair's offerings despite it being such a popular choice
All in all a great video nonetheless, yet these details really do end up annoying me for when it comes to inform the unknowing about what memory to choose for their new build
IMPORTANT MENTION:
Please do not use more or less than two sticks of memory that were purchased as a kit!
Well most user would not even know about memory timing other than just turn on expo or xmp anyway. So recommending CL30 does make sense.
@@Sintrania CL30 kits typically got far too loose subtimings which matter a whole lot more than the first timing of many in sequence
Right now you are in luck that SP032GXLWU60AFDEAE is the current cheapest 6000 Mbps 2x16 GB kit available in the US for $84.97
However this isn't always the case which brings me back to my original point
The most that's reasonable to pay extra is a whole $1 over a non-CL30 kit as long as you are able to ensure the use of Hynix A or M die
They know about the Buildzoids easy timings for hynix m- and a-die because they have literally tested them with games. You’re correct about CL but you do actually have to look at the primary timings because they can help you see if the kit is hynix or not. Some RAM manufacturers do show the SDRAM manufacturer for the kit like G.Skill.
Yesn't. Voltage can also be a good indicator as 1.25V should always end up being the one you want if it ain't CL30 or 32 for 6400 Mbps
Hello Hardware Unboxed! I Purchased Ryzen 7 7800x3d and pre ordered a Corsair Dominator Platinum 6200 Mhz CL36 to upgrade my Am4 system to Am5. but lately I discovered that the Ram I bought was for Intel, SKU: cmt32gx5m2x6200c36w
Question is: Will it work?
or what will happen if I combine these two?
I hope to get answers from you! Big Thanks!
Insightful. Good range of games reviewed. Is it possible to include the 'First descendant' to the games being tested as well?
EXPO might be worth it when boards are actually able to boot with it.
We've tried three different DDR5-6000 CL30 kits on every X670/B650 board and they work on every single one except for the 'Asus Prime B650M-A AX6-CSM WiFi'.
@@Hardwareunboxed Unlucky! I got the Prime B650M-A AX. Won't boot with anything faster than 4800 MHz.
Have you updated the BIOS to the latest version? The Prime boards are trash sadly.
@Hardwareunboxed Yup, latest BIOS. Ah well, live and learn I guess. Thanks anyway!
I've got a suggestion for a follow up for this - is more memory or faster memory better?
I now have 64GB of DDR5 set up as 4x16 sticks. However, my board doesn't seem to like running all four stick with EXPO on. Would I be better off taking two sticks out and running at full speed?
Depending on the speed difference and the type of application. If we're talking gaming, 32GB should be plenty and it'll be better to run at higher speed. Besides, this is something you can test yourself, right? Just take out two modules and check how well your games run.
One thing these mass media guys dont tell you is dont just throw any ram on your motherboard cpu combo. Always check what the optinsl ram is for your cpu/motherboard. Throwing 6000 on a cpu that doesnt work well with it wont do you any good. Like the 13400f for you budget guys it runs better on ddr4 3200 then it does even on 4800 ddr5. And it wont use more then either. So you can use 3200 or 4800 ddr5 thats your choices.
Just using this as an exmaple. Its why ram comparisons mean nothing to me and shouldnt to you either because the cou they sre testing on will always like a certain one more then another. Doesnt mean your cpu will.
I think you'll find basically all tech media made specific recommendations for specific CPUs, so really not sure what you're on about.
@@Hardwareunboxed I'm on about the comparisons yall do with a certain cpu and memory in videos like this and try going oh this is better or that's better when it isn't it's all motherboard/cpu dependant. That's all
From day one we made it clear that the sweet-spot for Zen 4 was DDR5-6000 and that remains true today.
"mass media guys" you mean the guys you learned everything about PC's from?
Oh boy another cIueIess c0mment 😅
In one of your Zen 5 reviews, you compared the jump from Zen 3 to Zen 4 to the jump from Zen 4 to Zen 5. I was expecting you to mention that there was no new DDR technology between the latter. I suspect that much of the performance jump on the former was due to the move from DDR4 to DDR5.
If I use Ryzen 7600X, is it worth to upgrade to DDR5 6000 or it's not worth it unless you have the higher end CPUs?? I'm currently using DDR5 4800.
Another thing to keep in mind is that if you find a deal on even higher spec memory, you should have little issue dialing it back to 6000. I did that with my 6400 kit
Excellent, one of the exact questions I had in my head, now I am informed in my buying decision.
man how do you manage to get such low fps on watchdog legion with 7800x3d and 4090? I get around 267 average
Would be interested to see some testing of EXPO vs XMP on AM5. I suspect the difference will be negligible but would still be useful to know as XMP rated kits still seem far more common and therefore more affordable than EXPO kits despite the fact that intel's high end is now a pretty hard sell on desktop.
Tech chanel of the year , seriously you guys have amswered every hot topic that needed to be answered
A decent portion of chips can run 6400 and UCLK at 2133 with higher SOC voltage, usually 1.25-1.3v, curious what the benefit is in FPS whether increased voltage is worth it or not.
I fiddled around with Ram the past week and currently I am using a Frankenstein setup of Ram. I use both a original DDR4 16GB 3200 CL16 kit AND a 32GB 4400 Cl26 kit but regulated both to down to 3200 CL15 and this dramatically improved performance and especially drops / lows on my 5800X3D. My new rule of thumb would be go for the max "officially supported" ram speed and shop for lower latency kits, depending on your budget. Latency seems really important for higher framerates.
I went with the black g skill ram that is cl30 6,000mhz 64 gigs of ddr5 but I have 4 sticks with my all white 4090/ 7800x3d build so the ram is slower but I havent updated the bios in a year but I will update it soon.
My 7950X has 64GB DDR5-5600...I'm happy with it, but maybe should look to check what spec the kit is rated for
Coincidentally I've been tinkering with RAM speeds only today! 😄 Thank you, great video! 🙂
I watched a test video that focused on testing MiHoYo games. The results showed that X3D, with its large cache, can effectively improve single-core performance and address issues with insufficient memory speed. Therefore, for certain games, Intel cpu should be paired with high-frequency memory, while AMD X3D cpu can perform well with just 6000MHz/CL30 memory.
My Ram stopped working at its rated speed of 3600Mhz. I now use it at 2666Mhz, the PC won't boot at any faster speed. I have noticed no difference in my PC performance or in games (at 4k) at all. Although I do run at 60fps, I don't notice faster fps.
One of the few review streams I like before watching. Memory speed impact impact was exactly what I was wondering about.
"If you still believe cpu limited cpu benchmarking is useless, click away now."
~Spreadsheet Steve
I know you really wanted to say something stronger than "click away now" but hopefully the people wanting 4k cpu benchmarks get the message. Hopefully. They are a stubborn bunch who aren't good with instructions and simple concepts but they may figure out how cpu benchmarks work eventually.
Great content as always, thank you Steve!
I bought a 12600k not long after they came out and when ddr5 when in its infancy. Im running 32gb of 5600mhz cl34. Would it be worth it to upgrade to something like 6400mhz cl32 even though gpu usage is usually 97-100% at 1440p with a 4070ti super? Should i just try to learn to overclock the kit i have?
Yes, you should try to overclock your kit. I use DDR4 3600 CL16 kit with my 12600K and I would think DDR5 would be no different in overclocking it to the speeds of higher clocked RAM.
Your current memory of 5600MT at CL34 = a memory latency of 12.14 seconds.
I overclocked my DDR4 kit to 3733 MT and tightened the timings down to CL14 which gives me a memory latency of 7.5 seconds.
Maybe you should try to OC your DDR5 to 6000 and tighten the timings to CL32 or CL30 if possible, that would bring your memory latency down to 10 seconds.
I guess with am5 the higher speed ddr5 memory is really good for better 1% lows for a smoother gaming experience. Great work as always Steve!
Interesting results and some were expected.
Would it be possible / hard to also put the % for the min fps in the darker part of the bar? in some tests the difference was quite big in others it was not. I can understand it will get more messy with more data in the graphs but i like all that info. Now I'm doing the calculations in my head or when you tell us the difference but when they are also shown it's easier to compare.
but i can imagine you also have good reasons not to include them on screen :)
So I just picked up a 7800 X 3D with neo cl34 6400 speed G skill Neo Ram should I go back to the store and get CL 30 6000 megh
no just set speed to 6000 cl30 in bios.
can i simply buy a fast expo ram kit, e.g. 8000, for an amd cpu with a good offer (more future-proof e.g. for ryzen 9000 and 10000) and, if necessary, lower the speed until it is stable? or is this only possible if you manually deal with the dozens of subtimings in the bios and know what you are doing?
For mid or lower end build a 5200mhz ddr5 is better as you can use the price diff to increase your gpu budget.
Here a used 6700xt & 6800 has a similar price gap of 32gb ddr5 5200mhz & 6000mhz
Hello first of all thanks for the content your team post frequently !
I'd like to know about your ram test like you've tested DDR5 memory but the faster ram have also better CL so what's the point of the testing they are for sure better if you calculate the absolute latency they are 50% faster like10ns vs 15ns
"It's silly to test at 1080p with a 4090!"
*Gets roasted by the fire in Steve's eyes*