Hope you enjoy the video! I thought some of the data was very interesting at least. If you have any questions, ask! If you want to help support my channel and this type of content, maybe consider sharing it somewhere
@@TechTesters then how come your results different then other j2 cents shows big numbers..i know he is intel shill but still number gap is wide..wonder how many UA-camrs have been playing with data 🙂
What do you mean? They intentionally designed a socket and chip that can run 9000MT and higher RAM and intentionelly included core isolation to futureproof the security of the platform. Of course they know and tested these things. The fact that reviewers cheaped out and ran tests with 6000MT RAM they had lying around from their AMD builds isn't Intel's fault. My conclusion here is that if you want to turn isolation off until that security layer is needed and also buy the optimal RAM for the platform (7200+, low latency, Gear 2) then you will end up with something at parity with the 14th gen at 30-40% less power draw and heat in gaming, but significantly more powerful in multicore stuff like media creation. Seems like a very average launch. The anti-Intel terminally-online commenters are exhausting to read.
@@CyberneticArgumentCreator Intels specs clearly say: Up to 6400MT/s and maximum supported speed 6400MHZ. So no they didn't design it for 9000MT because that would be in the specs. What is possible is not what is supported. What you probably are talking about is CUDIMM that controls its own speed and not controlled directly by the CPU.
@@CyberneticArgumentCreator quoting this video "intel actually suggested that core isolation should be enabled" and here you can see the value of their recommendation. That is why people perform tests, it is science, not some guy "guessing" "oh we dont have hyperthreading so we can utilize cores better when done separately" xDDD. Of course there are many additional variables, but this seems like rushed release, that's it.
nice video as always! Core Isolation requires Virtualization to be ON in BIOS and it's bound to it, so, my guess is that when you don't use VT, Core Isolation stays OFF. I'm a overclocker and also a gamer and never use VT, in fact, you can test this by disabling VT on BIOS and then when you search for Core Isolation in Windows it will say "page not avaiable" or something like that and instead it will show you TPM etc, but not Core Isolation anymore. Sometimes it takes some reboots, but, it will be gone with VT OFF. All the Best, Sergio!
The vast majority of PC users do not want to be forced to constantly fiddle with the BIOS to make their CPU work as it should. They believe it should work properly out of the box -- and they are correct.
I don’t think this video did anything other than XMP. Core isolation is not an Intel thing either, it’s Windows. If you don’t want to bother with XMP in the BIOS, just buy a prebuilt and call it a day and be happy with JEDEC speeds. As an overclocking enthusiast, ARL and its Z890 platform offers a lot more knobs and dials than RPL with Z690/Z790 did so it’s a welcome aspect. It’s just a shame the 285K isn’t as good as expected since the tools are so feature rich. You can individually offset the V/F curves of each individual P-Core this generation and each E-Core cluster, as well as independently alter the Ring Bus. Platform is so nice for enthusiasts, it’s such a shame the CPU is mediocre.
@@lechf I think rangersmith already typed out his reply, but he's still waiting on his comment to post. You'll have to excuse him. The X3D platform is very slow outside of games.
I am a 12th gen owner, to be onest I never heard of core isolation up until this week. I checked, it is fortunately off so I didn't give away performance for no reason over the past 2 years. thanks for informing us
Thank you Nada. The 285K is a paper launch so far, but it's nice to see there's at least some benefit to faster memory despite the high latency design Intel decided to go with. Looking forward to your 9800X3D review.
Hi there, can you share the RAMs model code? I cannot find the 2x16 GB model that you quote on Gskill website or anywhere around! And both the kits i received from MSI/ASUS are the same as yours, but 2x24 GB I am trying to understand if 2x24 can influence performances, because from my testing 8200 MHz CUDIMM have little to none improvement over 6000 MHz UDIMM, while 7200 Mhz UDIMM grant a little +5% as per your video! Thank you for your content, always on point!
So just buy a super expensive motherboard, super expensive RAM, fiddle with the BIOS and maybe if you are lucky your new Intel CPU will be as fast as the last gen AMD in one of your games. Got it.
It is not if you are looking for a new system. X870 and Z890 are close to same price. AMD vs INTEL CPU are close to same price. UDIMM vs CUDIMM the CUDIMM is about double the price. UDIMM 64GB (2x32GB) 6400 MHz US$300 vs CUDIMM 48GB (2 x 24GB) 9600 MHz US$400. (Newegg) Here is the possibility that can change it all. 5080 predicted price is US$1200-1500. Battlemage predicted price is US$500. Point here is that it looks like these two options will be about the same performance. I see INTEL goal being an all INTEL system.
To be fair, it's been like that since 13th gen. If you got good RAM, good motherboard and fiddled with the timing and voltages, you could get 13/14900K(S) to be on average better than 7800X3D. At least now you don't really need the top of the line cooler for both your CPU and your room/house.
Appreciate the continued coverage and look forward to more though I imagine with Zen 5 X3D about a week out the CPU testing feels endless. There's so much for everyone to figure out about these ARL CPUs I feel like it would have been beneficial for Intel to at least have given guidance in review guides as to what matters most. I've seen OC'd E-cores improve some gaming performance, as well as ring and D2D. Memory definitely has a more universal impact, but it'd be interesting to see whether or not latency has similar results to speed.
Great test including the well known core isolation problem for 12/13/14th gens, maybe you can test further with core isolation off and resizable bar on/off, sometimes resizable bar can decrease gaming performance.
Considering that games are particularly random and unpredictable in their memory accesses, cache misses are one of the key factors in performance, especially with complex hierarchical multi core designs. That's why big caches help so much. Since memory latency is the main weakness of Arrow Lake it makes sense to try to improve that first via better RAM, but it will take a lot of effort (and money) to partly compensate for the 30 ns increase in latency (>100 ns vs 70 ns) compared to the previous generation of CPUs. I was hoping for nice well-rounded CPUs, basically energy efficient 14000s, which, despite their problems, were quite good at fetching data from main memory quickly, but I guess we will have to wait another two years for Intel to fix the latency issues in its new chiplet design. At least idle energy consumption is not as horrible as AMD's and generally energy efficiency seems to be particularly great when using lower power limits. Thus I could imagine that non-K Arrow Lake CPUs will show very good results in that regard. So it might still be an interesting platform for some use cases, but apparently not for pure gaming machines.
DDR5-6400 doesn't really cost much more nowadays. And the calculation is different on whether you're just upgrading memory, or building a whole new system.
I don't get why anyone would spend so much on a new MB, an indifferent CPU, and top-dollar memory, to end up with this disappointing level of performance. And with no upgrade path promised, I can't see this socket lasting very long. If the 14 series CPUs have genuinely been fixed with the updated micro-code, as Intel say, they look the better options, at least for gamers, if they don't fancy AMD.
@@marwanak10 if you're creating, time is money. so, just buy a Threadripper system with some RTX A6000s that will smoke a 285k. Covers both CPU intensive and GPU intensive workloads. And you get 128 PCIE lanes to connect all the things you'd like. The 285k is consumer class with only 24 lanes just like AMD's consumer class CPUs.
@@marwanak10well there are players who plays competitive titles. Intel is the best for these type of games. Better 1% lows and consistent frame pacing > high avg fps. Also high polling rate mice works the best with an Intel cpu.
I think the fact that Intel abolished its Hyperthreading is hurting it in more games than it benefited from. Hyperthreading was a security nightmare for Intel over the past decade+ and I assume they removed hyperthreading to try and mitigate this issue.
One can tell the productivity and games optimized for Arrow on compiling tools and skills. The laggards which are obvious now holding their hands out for Intel to pay them to recompile. Extorting Intel, AMD has the same issue, at every new architectural generation requiring application hardware compiler optimizations. Whole platform is collaborative. Let the lazy software devs starve until they pick up the tools and optimize their apps. mb
Looks like DDR5 still has a couple more gens left to mature like DDR4 did to 3600, 3200. Great to see the speed changes and compatibility improve on the newer platforms.
Useful video. But really cant help Intel's case. When people put this amount of money into their CPU they dont wanna spend weeks wafting for updates and doing fixes. In my experience when i build pc for people who put this amount of money into their PC, they want it to work day one and be the best and those are just the gaming fans. With professionals, thigs are way more extreme and they have no intention of tweaking things to make them work. People pay A LOT of money, it has to be the best and has to work. Its that simple.
MSI sent you a Unify X but you didnt show 8200-8600 on A die with tight timings. You will see a significant uplift using a tuned A die. Cudimm is kinda like a meme tbh. This platform still has plenty of issues specifically scheduling issues. I try to like my 265K but theres just so much issues with certain games.
I think the metric missing from this presentation is the 14900K. Do these changes mean it’s better than or at least equal to the 14900K? As consumers we should expect an uplift.
Suggesting turning off core isolation seems like bad advice to me. It is a security setting in Windows that is enabled by default. Testing with it enabled seems not only fair, but also more representative for real world performance. From an educational standpoint testing the differences is indeed interesting. It reminds me of the AMD tests of a couple of months back. Using the actual administrator account yielded better results for their products. However, that is not how a normal user runs their system.
I don't think it is enough to make me interested in the 2xx series. I do find the OCing of these chips interesting, it is different enough that it feel new. Reaching 8200 without using CUDIMM seems impressive, did you check if this was actually 100% stable? I guess it was stable enough to run the benchmarks 🤔
It's possible they're using different testing conditions where the difference doesn't show (as much), or maybe turning it off benefits AMD even more (which I haven't tested lately), who knows!
Thank you for the updates. i actually wonder if the timings of the memory matters, because as i built my PC last year i tried to pick the fastest memory with the lowest timings available in the market, so i ended up buying (the sweet spot) a 6400hz with 32CL i guess (can't remember exactly😅). So do timings matter in competitive games like Valorant and csgo ?, because i don't think it matters in games like cyberpunk that benefits more on faster memories in my opinion. Thanks a lot 😊
Maybe 6800 is a sweet spot IMO. I'm so glad I have 64Gb 6400 from G.skill and mostly said it can go up to 6800 even without touching anything in bios. I hope I'm a one of that lucky people lol
I just hope to see faster memory speeds at higher capacities like 64 or even 96GB. I was only able to get 6600MT/s on my 96GB Corsair Dominator Titanium kit with the 14900k and only 6400 after the CPU became unstable due to the faulty microcode that killed it. Anyway now we are only seeing 48GB kits at speeds that are to high actually for the sweetspot in GEAR 2. Intel said the sweetspot was 8000MT/s not 8200 or 8400. I would love to see some Dominator Titanium 96GB CUDIMM kits at or around 8000MT/s 🤩. Idk if that is asking to much and if CUDIMM is even able to do that or if the IMC on the 285K can do it but would be nice!
@@TheTastefulThicknessonly people that has it is UA-cam testers. Damn thing never came in stock for the public to buy. Said the 24th of October then changed to 31st of October and they are still out of stock.
i see just very old games testing on the brand new cpu, the newest is starfield and that game released in september 2023 and performing at the best on arrow lake. maybe because of the hidden quad channel on dual slots. i mean not this video, overall the whole yT landscape testing just and only very very old games.
In my 40 game test with the 265K there are a bunch of newer games. For this one I specifically picked the ones where Intel's results vs AMD were bigger than expected.
Thank you! And I'll reconsider including Battlefield games when EA stops with their useless hardware change limit, right now they make it impossible to test their games extensively with different hardware.
Speaking of win11 update, did you turn off that "windoze artificial incompetence", that M$$$$ is forcing-feeding everyone, by installing 24H2. Regardless, I always enjoy watching your reviews. You are able to deliver facts, not dreams or broken dreams. Second, does M$$ know, that the E-cores are not at the same place on the die, as Raptor Lake's were? Win11 may still be handling core0-7 as P-core and core 8-15 as E-cores, but they would be very wrong. Another thing to consider is, that Intel removed "hyper threading" from the new architecture, so Meteor Lake does appear as 24 threads and Raptor Lake with hyper threading did show 32 threads. Since I was always bad at math in high-school, where could a tiny drop in performance from a 32 -thread CPU , using 4 nuclear power plants dry per second, vs a 24 -thread CPU, which manages to use more moderate to a lot less power, than before. Let me try math here- 32 - 24 = 0? Naah. Doesn't look right. Are you saying, that all the software currently in use right now, optimized to hell and back to use as many hardware threads as are available, performs better versus a new architecture, that in a worse case may not even recognize the cpu and the current software hasn't have the time to yet to get the optimizations for the new architecture and make a somewhat more equalized comparison possible. Didn't that happen also, when Intel put the memory controller from the chipset of the motherboard to the cpu die and the Core2Duo did "outrun" the new Core i CPUs, changing from the P45 chipset to the Z190 chipset. Do love the gaming performance tests. If current games don't run at 999 FPS, the new CPU won't cut it. It almost makes a difference if a game renders at 700 FPS or 523 FPS. No human will ever be able to notice, sense or comprehend. And my last remark would be, that people running a NVIDIA 4090 for only 6000 $ US, will be sad to hear that their 3000 $ pc did not stop costing $$$$ after you went home from the store. After one year of usage, your pc will have cost twice as much.
yes! It's in the QVL of motherboard manufacturers. Just check it. The best memory with best speed and lowest latency is Corsair Vengeance 64GB (16x4) DDR5 6600Mhz with CAS32
Your FS2020 results for 1080p just don't make sense (I don't play the others). My OC'd 14900KS can do better than that at 4K Full Ultra, and that's before the multithreading improvements in FS2024 (as proven in the Tech Alpha which I played).
Hey. If you check my 265K review, you'll see that you can run into CPU bottlenecks, with a 4090, at 1080p Ultra as much as 4K Ultra. So it actually makes complete sense that you're seeing better results at 4K with an OCd 14900KS than I'm seeing here at 1080p with a 285K.
CUDIMM does a lot but not alone. 5.1 E Cores/4.1 Cache chance for clocking and you see a good jump. Point here is that the CPU is very much underclocked and underpowered. This very small change puts the 285K on top even without BIOS and win updates. So you should see much more. Updates to BIOS and win are almost daily. There is no stock and what was are on hands of testers and reviewers. You are not able to get 285K and if you can you won a lottery. My view is that INTEL wants to test and keep MB manufacturers in check before opening the floodgates. I predict Battlemage will be launched at the same time. Hang tight as I think it will take few weeks or months. Let others do the testing. I predict the performance with Battlemage will match 5080 if not beat it. It will be close. So aim all INTEL system and save. 5080 alone is predicted to sell at US$1200-1500. Battlemage US$500 and bit more for CUDIMM's.
except many reviews had faster ram. Usually you go well in depth in your videos and tests, however with that title, it's only fair you would still put it up against other cpus still.
That would be days more testing. That won't be worth it for a while longer, at least not until AMD's new CPUS are out, Intel sorts some of these inconsistencies out, and probably MS pushes another Windows update or.. 5.
@@TechTestersI'm very curious to see where things end up in a few months when everything gets sorted out. Of course with the 9800X3D around the corner it might not matter much
About core isolation..with my 285K if I turn it OFF then Games with Easy Anti cheat wont start/get a blue screen...would anyone discover the same ? Example: Spacemarine 2
Yep, i mentioned that in my Core Ultra 7 review. We had the same issue in Division 2 and Throne and Liberty. Intel said they are working on a fix that should make the games playable with both settings, but its still early to tell how that will go.
and the i9 14900k can use 6400 memory. so if no big difference using 8400 memory seems like waste of money. if we can get thunderbolt 5 working on 14900k motherboards I prefer to keep my 14900k for at least 5 years. I dont need AI. AI is spying on you and reporting back to big tech.
The only thing that appears to "fix" the i9 Ultra performance is manually tweaking the Cache speed, so if you run for example 8200MHz RAMM, you need to set it at 41x and for 8800MHz must be set at 44x etc... Hope this helps!
@@TechTesters yes try it, but be sure to type all the frequency for both cores and cache manually because this BIOS feels like an early Alpha and far from ready, has quite a lot issues that needs polishing! I highly reccomend using CUDIMM 8800MHz with 44x cache, set P to 55x and E to 46x fixed ratios.
thanks for sharing your data, pretty cool to see ram speed testing. it's a shame the 7800x3d doesn't benefit from higher then 6k ram speeds and certainly doesn't look like the 9800x3d will either. so i suppose Intel has the higher ram speed performance advantage for now but it's still slower in most games. such a shame. decent gains but still doesn't put it in a good spot sadly. even with the improved power utilization of the new chips it's just still not in the same ballpark. i don't want Intel to fade away, we need at least two companies.....it's going suck if there is only one left.
I returned my 265k and z890 strix to microcenter. After trying to install windows 11 5x and then having fortnite either crash or not load at all I decided to switch to Amd.
I find it fascinating that all these Techtubers are bashing Intel for the gaming performance when Intel themselves said 3 weeks ago that there will be little to no performance gains compared to the 13th & 14th gen. They made it publicly known lol.
1 I don't think this video is "bashing". 2 There is a difference between "little to no performance gains" or "performance parity" (direct quote from Intel), and the actual gaming performance regression that a lot of people are seeing.
Are you expecting from reviewers to praise Intel for delivering a performance regression after years of refreshing the same architecture (read 12th Gen Alder Lake)? Yes, they’ve reduced power usage, but the 14th Gen is overclocked heavily from the factory, so it’s not as significant achievement as it seems. Arrow Lake’s power consumption is still much higher than AMD’s, despite Intel using a more advanced node from TSMC.
@TechTesters I wasn't speaking about this specific video. I was referring to the overall comments from the Tech Media regarding the new CPU launch last week.
Unfortunately, Intel's architecture isn't fully leveraging TSMCs 3nm which is why you are seeing consumption is still higher. Your response alone indicates you are clearly out of your depth.
I really hope Intel somehow figures out how to compete with AMD in gaming. Otherwise AMD's gonna keep increasing prices. We already see 7800X3D hitting $500-$600 mark.
CUdimm is comming soon. That will give even a bigger boost in performance. I think all in all the new Intel cpu is great. It just need some good tweaks and adjustments. Turning DLVR of on the core ultra 9 will save a lot of power when using it in multithreaded load like cionebench with a lower voltage input. It can easy save 80 watts as Der8auer showed on his channel. So when Windows is optimizing the new core and games as well the sun will shine again for Intel. To many people are to negative just because it's Intel and are not looking without bias to the product.
The video that was posted 20h ago? Unfortunately, it takes days of testing and retesting with all the new updates to get enough data for this video. Which is also the reason I haven't had time to watch his video just yet. Ill check it out tonight! always interesting to see all the data and different perspectives on the subject.
I sort of feel bad for any gamer that doesn't follow tech news and ends up with one of these CPUs in their system. When AMD is out there. Coming from a recently upset 14900K customer that I got refunded for a 7800X3D. Ps before you reply "bUt NoT aLL bUyERs ArE gAmERzz!" Well I said I feel bad for gamers.
@@Tdisputations 1% lows, a lot are bound to notice. When you're at that level of gear, you're more than likely tracking performance. I'd agree in the low-mid tiers of CPUs but not on the top end.
@@Tdisputations that was at 4k resolution though right? Still a lot of gamers buy on the high end to push more frames on 1440p and 1080p high hz displays. At 4k still better off to sce your pocket a big chunk of change and have better performance.
It is well accepted amongst reviewers that you use a low resolution and a high end GPU to remove any GPU bottleneck from CPU results. For GPU testing Nada uses 1440p and 4K like everyone else. As well as 1080p which is still the most popular resolution on the Steam Hardware survey
Hope you enjoy the video! I thought some of the data was very interesting at least. If you have any questions, ask! If you want to help support my channel and this type of content, maybe consider sharing it somewhere
Can we test with CUDIMM 8400 instead of normal DIMM? Would that make a difference?
Should have tested cudimm too that???
@@krishnav5122
She used CUDIMM 8200MTs
Explained on min 1
Note for future videos: mention U vs CUDIMM and the memory timings. Those can have an impact too, not just the bandwidth.
@@TechTesters then how come your results different then other j2 cents shows big numbers..i know he is intel shill but still number gap is wide..wonder how many UA-camrs have been playing with data 🙂
The most level headed overview of 285k not bashing or singing praise and glad to see that core isolation off is still the way to go.
Interesting results, thanks for the retests.
Did intel even test these things before launch? Doesn't seem like it. Thanks for doing intels job for them. Great video.
What do you mean? They intentionally designed a socket and chip that can run 9000MT and higher RAM and intentionelly included core isolation to futureproof the security of the platform. Of course they know and tested these things.
The fact that reviewers cheaped out and ran tests with 6000MT RAM they had lying around from their AMD builds isn't Intel's fault.
My conclusion here is that if you want to turn isolation off until that security layer is needed and also buy the optimal RAM for the platform (7200+, low latency, Gear 2) then you will end up with something at parity with the 14th gen at 30-40% less power draw and heat in gaming, but significantly more powerful in multicore stuff like media creation. Seems like a very average launch. The anti-Intel terminally-online commenters are exhausting to read.
@@CyberneticArgumentCreator Intels specs clearly say: Up to 6400MT/s and maximum supported speed 6400MHZ. So no they didn't design it for 9000MT because that would be in the specs. What is possible is not what is supported.
What you probably are talking about is CUDIMM that controls its own speed and not controlled directly by the CPU.
@@CyberneticArgumentCreator quoting this video "intel actually suggested that core isolation should be enabled" and here you can see the value of their recommendation. That is why people perform tests, it is science, not some guy "guessing" "oh we dont have hyperthreading so we can utilize cores better when done separately" xDDD. Of course there are many additional variables, but this seems like rushed release, that's it.
nice video as always!
Core Isolation requires Virtualization to be ON in BIOS and it's bound to it, so, my guess is that when you don't use VT, Core Isolation stays OFF. I'm a overclocker and also a gamer and never use VT, in fact, you can test this by disabling VT on BIOS and then when you search for Core Isolation in Windows it will say "page not avaiable" or something like that and instead it will show you TPM etc, but not Core Isolation anymore. Sometimes it takes some reboots, but, it will be gone with VT OFF.
All the Best, Sergio!
The vast majority of PC users do not want to be forced to constantly fiddle with the BIOS to make their CPU work as it should. They believe it should work properly out of the box -- and they are correct.
Agreed. But it's probably worth telling them to enable XMP at least, or disable a simple windows setting like Core Isolation.
@@TechTesters next they will ask the cpu to play games for themselves.
I don’t think this video did anything other than XMP. Core isolation is not an Intel thing either, it’s Windows. If you don’t want to bother with XMP in the BIOS, just buy a prebuilt and call it a day and be happy with JEDEC speeds.
As an overclocking enthusiast, ARL and its Z890 platform offers a lot more knobs and dials than RPL with Z690/Z790 did so it’s a welcome aspect. It’s just a shame the 285K isn’t as good as expected since the tools are so feature rich. You can individually offset the V/F curves of each individual P-Core this generation and each E-Core cluster, as well as independently alter the Ring Bus.
Platform is so nice for enthusiasts, it’s such a shame the CPU is mediocre.
We get it bro you have a 5800x3d
@@lechf I think rangersmith already typed out his reply, but he's still waiting on his comment to post. You'll have to excuse him. The X3D platform is very slow outside of games.
I am a 12th gen owner, to be onest I never heard of core isolation up until this week. I checked, it is fortunately off so I didn't give away performance for no reason over the past 2 years. thanks for informing us
Always useful and great vids .. Happy to see u back keep goiing 💜
Thank you Nada. The 285K is a paper launch so far, but it's nice to see there's at least some benefit to faster memory despite the high latency design Intel decided to go with. Looking forward to your 9800X3D review.
Hartelijk dank voor de Intel Core Ultra Gaming Performance-test Tech Mama 🤓💯
Hi there, can you share the RAMs model code?
I cannot find the 2x16 GB model that you quote on Gskill website or anywhere around!
And both the kits i received from MSI/ASUS are the same as yours, but 2x24 GB
I am trying to understand if 2x24 can influence performances, because from my testing 8200 MHz CUDIMM have little to none improvement over 6000 MHz UDIMM, while 7200 Mhz UDIMM grant a little +5% as per your video!
Thank you for your content, always on point!
Thank you so much for this video
Your videos always make me chill and smile :) thank you 🥳
So just buy a super expensive motherboard, super expensive RAM, fiddle with the BIOS and maybe if you are lucky your new Intel CPU will be as fast as the last gen AMD in one of your games.
Got it.
Correct :D
It is not if you are looking for a new system.
X870 and Z890 are close to same price.
AMD vs INTEL CPU are close to same price.
UDIMM vs CUDIMM the CUDIMM is about double the price. UDIMM 64GB (2x32GB) 6400 MHz US$300 vs CUDIMM 48GB (2 x 24GB) 9600 MHz US$400. (Newegg)
Here is the possibility that can change it all. 5080 predicted price is US$1200-1500. Battlemage predicted price is US$500. Point here is that it looks like these two options will be about the same performance. I see INTEL goal being an all INTEL system.
To be fair, it's been like that since 13th gen. If you got good RAM, good motherboard and fiddled with the timing and voltages, you could get 13/14900K(S) to be on average better than 7800X3D. At least now you don't really need the top of the line cooler for both your CPU and your room/house.
Appreciate the continued coverage and look forward to more though I imagine with Zen 5 X3D about a week out the CPU testing feels endless.
There's so much for everyone to figure out about these ARL CPUs I feel like it would have been beneficial for Intel to at least have given guidance in review guides as to what matters most. I've seen OC'd E-cores improve some gaming performance, as well as ring and D2D. Memory definitely has a more universal impact, but it'd be interesting to see whether or not latency has similar results to speed.
Great test including the well known core isolation problem for 12/13/14th gens, maybe you can test further with core isolation off and resizable bar on/off, sometimes resizable bar can decrease gaming performance.
Considering that games are particularly random and unpredictable in their memory accesses, cache misses are one of the key factors in performance, especially with complex hierarchical multi core designs. That's why big caches help so much. Since memory latency is the main weakness of Arrow Lake it makes sense to try to improve that first via better RAM, but it will take a lot of effort (and money) to partly compensate for the 30 ns increase in latency (>100 ns vs 70 ns) compared to the previous generation of CPUs.
I was hoping for nice well-rounded CPUs, basically energy efficient 14000s, which, despite their problems, were quite good at fetching data from main memory quickly, but I guess we will have to wait another two years for Intel to fix the latency issues in its new chiplet design. At least idle energy consumption is not as horrible as AMD's and generally energy efficiency seems to be particularly great when using lower power limits. Thus I could imagine that non-K Arrow Lake CPUs will show very good results in that regard. So it might still be an interesting platform for some use cases, but apparently not for pure gaming machines.
I would love to see some perfomance tests for Raptor Lake with CUDIMM.
Good video!
But going for faster memory means paying more?
So, value /performance is worse or better compared to others?
DDR5-6400 doesn't really cost much more nowadays. And the calculation is different on whether you're just upgrading memory, or building a whole new system.
Do you recommend turning off core isolation for 12th and 13th gen ?
I haven't tested it lately, but from previous content and looking at this data I'd say that turning it off will probably improve FPS there as well.
@TechTesters Ok, thank you.
I don't get why anyone would spend so much on a new MB, an indifferent CPU, and top-dollar memory, to end up with this disappointing level of performance. And with no upgrade path promised, I can't see this socket lasting very long. If the 14 series CPUs have genuinely been fixed with the updated micro-code, as Intel say, they look the better options, at least for gamers, if they don't fancy AMD.
then again why would you go that route over AMD for gaming?
For creators the new Intel makes sense.
@@marwanak10 if you're creating, time is money. so, just buy a Threadripper system with some RTX A6000s that will smoke a 285k. Covers both CPU intensive and GPU intensive workloads. And you get 128 PCIE lanes to connect all the things you'd like. The 285k is consumer class with only 24 lanes just like AMD's consumer class CPUs.
@@Salty_Nutella You're comparing a 2.5k USD system to a 10K+ USD System. That's like saying, why get a BMW just get a Ferrari if you wanna go fast.
@@marwanak10well there are players who plays competitive titles. Intel is the best for these type of games. Better 1% lows and consistent frame pacing > high avg fps. Also high polling rate mice works the best with an Intel cpu.
I think the fact that Intel abolished its Hyperthreading is hurting it in more games than it benefited from. Hyperthreading was a security nightmare for Intel over the past decade+ and I assume they removed hyperthreading to try and mitigate this issue.
One can tell the productivity and games optimized for Arrow on compiling tools and skills. The laggards which are obvious now holding their hands out for Intel to pay them to recompile. Extorting Intel, AMD has the same issue, at every new architectural generation requiring application hardware compiler optimizations. Whole platform is collaborative. Let the lazy software devs starve until they pick up the tools and optimize their apps. mb
i'll wait till Core Ultra Gen3 release. it will be much more refined by then.
Looks like DDR5 still has a couple more gens left to mature like DDR4 did to 3600, 3200. Great to see the speed changes and compatibility improve on the newer platforms.
Hi Nada, and I hope you are having a nice day. I sure am! I watched not one but two of your videos today.
I loved this thrumbnail xD on your vid
Core isolation pretty much need be off for all CPU helps a lot
Useful video. But really cant help Intel's case. When people put this amount of money into their CPU they dont wanna spend weeks wafting for updates and doing fixes.
In my experience when i build pc for people who put this amount of money into their PC, they want it to work day one and be the best and those are just the gaming fans. With professionals, thigs are way more extreme and they have no intention of tweaking things to make them work.
People pay A LOT of money, it has to be the best and has to work. Its that simple.
Where can i buy intel core ultra 9 285k in europe?Sold out everywhere.
Same in the United States. Seems only UA-cam testers got them
MSI sent you a Unify X but you didnt show 8200-8600 on A die with tight timings. You will see a significant uplift using a tuned A die. Cudimm is kinda like a meme tbh. This platform still has plenty of issues specifically scheduling issues. I try to like my 265K but theres just so much issues with certain games.
I think the metric missing from this presentation is the 14900K. Do these changes mean it’s better than or at least equal to the 14900K? As consumers we should expect an uplift.
Suggesting turning off core isolation seems like bad advice to me. It is a security setting in Windows that is enabled by default. Testing with it enabled seems not only fair, but also more representative for real world performance. From an educational standpoint testing the differences is indeed interesting. It reminds me of the AMD tests of a couple of months back. Using the actual administrator account yielded better results for their products. However, that is not how a normal user runs their system.
I don't think it is enough to make me interested in the 2xx series.
I do find the OCing of these chips interesting, it is different enough that it feel new.
Reaching 8200 without using CUDIMM seems impressive, did you check if this was actually 100% stable?
I guess it was stable enough to run the benchmarks 🤔
Nada, you should work for Intel. How did they miss core isolation? They left performance on the table.
It's possible they're using different testing conditions where the difference doesn't show (as much), or maybe turning it off benefits AMD even more (which I haven't tested lately), who knows!
No comparison to UDIMM?
I have been playing Warhammer Space Marine 2 with core isolation off, because with it on, the stutters are puke inducing.
Teşekkür ederiz. 👍
Thank you for the updates.
i actually wonder if the timings of the memory matters, because as i built my PC last year i tried to pick the fastest memory with the lowest timings available in the market, so i ended up buying (the sweet spot) a 6400hz with 32CL i guess (can't remember exactly😅).
So do timings matter in competitive games like Valorant and csgo ?, because i don't think it matters in games like cyberpunk that benefits more on faster memories in my opinion.
Thanks a lot 😊
is : G.SKILL Trident Z5 CK 48GB (2 x 24GB) 288-Pin CUDIMM DDR5 9600 (PC5 76800) RGB Desktop Memory Model F5-9600C4658H24GX2-TZ5CRK
with cudimm...it jump further with OC ecores do wonders
Please do 4k testing as well?
Maybe 6800 is a sweet spot IMO. I'm so glad I have 64Gb 6400 from G.skill and mostly said it can go up to 6800 even without touching anything in bios. I hope I'm a one of that lucky people lol
The 9950X still seems like a better option. 8200MT memory with no listed price 😮
Love those Mom jeans!
Can we can an update on the (lack of) sleep hours?
thx ;) I am still rocking 13600kf :)
i have 12700K its totaly fine but i hate to use it because Intel is dying its psychological😂
It is a solid CPU, you are good for a few years.
@@techwizard9699 i ordered 7800x3d its for picing up i can cancel the order should i? i only play games and some paperwork like invoices, excel word
Mission impossible to buy 32GB G.Skill Trident Z5 CK DDR5-8200
Shoutout to my DDR4 peeps!
interesting .. U Have 285K Subs .. i've seen this number before 👀👀
Seems it's possible to squeeze some other 5% performance by overclocking+DDR5 8000
If you are spending $2-5k on a system, I am not sure the 285k is the best choice there though
I just hope to see faster memory speeds at higher capacities like 64 or even 96GB. I was only able to get 6600MT/s on my 96GB Corsair Dominator Titanium kit with the 14900k and only 6400 after the CPU became unstable due to the faulty microcode that killed it. Anyway now we are only seeing 48GB kits at speeds that are to high actually for the sweetspot in GEAR 2. Intel said the sweetspot was 8000MT/s not 8200 or 8400. I would love to see some Dominator Titanium 96GB CUDIMM kits at or around 8000MT/s 🤩. Idk if that is asking to much and if CUDIMM is even able to do that or if the IMC on the 285K can do it but would be nice!
Any CUDIMM tests coming up in the future?
This is with CUDIMM. Do you mean CUDIMM vs regular memory at similar speeds? :)
@@TechTestersActually that would be quite an interesting comparison
I'm patiently waiting 🙏
@@markozindovic9227 I don't actually have a DDR5-8200 kit that isn't CUDIMM atm. I might look into it later, first I have some other content planned!
@@TechTesters Gotcha
I guess UD 7600c36 vs 8200c40 CuD would do it. (Both 2x24)
8000+ on normal UD may not even work @@TechTesters
It would be awesome if u could overclocking the ngu, d2d and cores all through bypass mode! 🙌🏼
I think I want a CPU test break first :D
@ haha, yeah, the BIOS screen can dauntingly boring, 🤭
I can't wait to not buy this cpu
Its selling just fine without you
@@TheTastefulThickness you're the only one and User Benchmark that bought this CPU.
@@pkpnyt4711 its selling out.
@@TheTastefulThicknessonly people that has it is UA-cam testers. Damn thing never came in stock for the public to buy. Said the 24th of October then changed to 31st of October and they are still out of stock.
@@MRwilliejoyner This is a lie. The chips launched and were on sale.
i see just very old games testing on the brand new cpu, the newest is starfield and that game released in september 2023 and performing at the best on arrow lake. maybe because of the hidden quad channel on dual slots. i mean not this video, overall the whole yT landscape testing just and only very very old games.
In my 40 game test with the 265K there are a bunch of newer games. For this one I specifically picked the ones where Intel's results vs AMD were bigger than expected.
Great video but you need to test Battlefield, BF2042 loves high speed ddr
Thank you! And I'll reconsider including Battlefield games when EA stops with their useless hardware change limit, right now they make it impossible to test their games extensively with different hardware.
OMG it's steve jobs!
Speaking of win11 update, did you turn off that "windoze artificial incompetence", that M$$$$ is forcing-feeding everyone, by installing 24H2.
Regardless, I always enjoy watching your reviews. You are able to deliver facts, not dreams or broken dreams.
Second, does M$$ know, that the E-cores are not at the same place on the die, as Raptor Lake's were? Win11 may still be handling core0-7 as P-core and core 8-15 as E-cores, but they would be very wrong.
Another thing to consider is, that Intel removed "hyper threading" from the new architecture, so Meteor Lake does appear as 24 threads and Raptor Lake with hyper threading did show 32 threads.
Since I was always bad at math in high-school, where could a tiny drop in performance from a 32 -thread CPU , using 4 nuclear power plants dry per second, vs a 24 -thread CPU, which manages to use more moderate to a lot less power, than before.
Let me try math here- 32 - 24 = 0? Naah. Doesn't look right.
Are you saying, that all the software currently in use right now, optimized to hell and back to use as many hardware threads as are available, performs better versus a new architecture, that in a worse case may not even recognize the cpu and the current software hasn't have the time to yet to get the optimizations for the new architecture and make a somewhat more equalized comparison possible.
Didn't that happen also, when Intel put the memory controller from the chipset of the motherboard to the cpu die and the Core2Duo did "outrun" the new Core i CPUs, changing from the P45 chipset to the Z190 chipset.
Do love the gaming performance tests. If current games don't run at 999 FPS, the new CPU won't cut it.
It almost makes a difference if a game renders at 700 FPS or 523 FPS. No human will ever be able to notice, sense or comprehend.
And my last remark would be, that people running a NVIDIA 4090 for only 6000 $ US, will be sad to hear that their 3000 $ pc did not stop costing $$$$ after you went home from the store. After one year of usage, your pc will have cost twice as much.
Can Arrow Lake handle 4 dimms at 6400?
yes! It's in the QVL of motherboard manufacturers. Just check it. The best memory with best speed and lowest latency is Corsair Vengeance 64GB (16x4) DDR5 6600Mhz with CAS32
Your FS2020 results for 1080p just don't make sense (I don't play the others). My OC'd 14900KS can do better than that at 4K Full Ultra, and that's before the multithreading improvements in FS2024 (as proven in the Tech Alpha which I played).
Hey. If you check my 265K review, you'll see that you can run into CPU bottlenecks, with a 4090, at 1080p Ultra as much as 4K Ultra. So it actually makes complete sense that you're seeing better results at 4K with an OCd 14900KS than I'm seeing here at 1080p with a 285K.
CUDIMM does a lot but not alone. 5.1 E Cores/4.1 Cache chance for clocking and you see a good jump. Point here is that the CPU is very much underclocked and underpowered. This very small change puts the 285K on top even without BIOS and win updates. So you should see much more. Updates to BIOS and win are almost daily.
There is no stock and what was are on hands of testers and reviewers. You are not able to get 285K and if you can you won a lottery. My view is that INTEL wants to test and keep MB manufacturers in check before opening the floodgates. I predict Battlemage will be launched at the same time. Hang tight as I think it will take few weeks or months. Let others do the testing. I predict the performance with Battlemage will match 5080 if not beat it. It will be close. So aim all INTEL system and save. 5080 alone is predicted to sell at US$1200-1500. Battlemage US$500 and bit more for CUDIMM's.
Tech mommy is back!
You need to use CUDIMMs to get a scaling boost for some reason the 200 series loves the clock driver ram
These are CUDIMMs :)
Faster RAM costs a lot of money which is not worth it based on your video!
except many reviews had faster ram. Usually you go well in depth in your videos and tests, however with that title, it's only fair you would still put it up against other cpus still.
That would be days more testing. That won't be worth it for a while longer, at least not until AMD's new CPUS are out, Intel sorts some of these inconsistencies out, and probably MS pushes another Windows update or.. 5.
@@TechTestersI'm very curious to see where things end up in a few months when everything gets sorted out. Of course with the 9800X3D around the corner it might not matter much
Will be interested to see how much difference CUDIMMs make…
Just to clarify, these are CUDIMM modules :)
@ I don’t think you mentioned that at all. Important detail.
Arrow lake get good in games only when oc p cores to 5.7
About core isolation..with my 285K if I turn it OFF then Games with Easy Anti cheat wont start/get a blue screen...would anyone discover the same ?
Example: Spacemarine 2
Yep, i mentioned that in my Core Ultra 7 review. We had the same issue in Division 2 and Throne and Liberty. Intel said they are working on a fix that should make the games playable with both settings, but its still early to tell how that will go.
@TechTesters thanks a lot..except These my 285k runs rallye perfekt.. no Bugs.. No issues..super stable
Buy new CPU that performs poorly, Needs new expensive motherboard and high speed ram. Intel core Ultra or AMD Ryzen 7000 series??
I think we can all agree that you are an amazing person
And cute
In summary - go for AMD.
and the i9 14900k can use 6400 memory. so if no big difference using 8400 memory seems like waste of money. if we can get thunderbolt 5 working on 14900k motherboards I prefer to keep my 14900k for at least 5 years. I dont need AI. AI is spying on you and reporting back to big tech.
Still sucks.
Intel needs a miracle.
The only thing that appears to "fix" the i9 Ultra performance is manually tweaking the Cache speed, so if you run for example 8200MHz RAMM, you need to set it at 41x and for 8800MHz must be set at 44x etc... Hope this helps!
I'll definitely give that a go, thanks! But first, a little break from CPU content!
@@TechTesters yes try it, but be sure to type all the frequency for both cores and cache manually because this BIOS feels like an early Alpha and far from ready, has quite a lot issues that needs polishing! I highly reccomend using CUDIMM 8800MHz with 44x cache, set P to 55x and E to 46x fixed ratios.
very 80s look
thanks for sharing your data, pretty cool to see ram speed testing. it's a shame the 7800x3d doesn't benefit from higher then 6k ram speeds and certainly doesn't look like the 9800x3d will either. so i suppose Intel has the higher ram speed performance advantage for now but it's still slower in most games. such a shame.
decent gains but still doesn't put it in a good spot sadly. even with the improved power utilization of the new chips it's just still not in the same ballpark.
i don't want Intel to fade away, we need at least two companies.....it's going suck if there is only one left.
👍👍❤❤❤
I returned my 265k and z890 strix to microcenter. After trying to install windows 11 5x and then having fortnite either crash or not load at all I decided to switch to Amd.
😊
I find it fascinating that all these Techtubers are bashing Intel for the gaming performance when Intel themselves said 3 weeks ago that there will be little to no performance gains compared to the 13th & 14th gen. They made it publicly known lol.
1 I don't think this video is "bashing".
2 There is a difference between "little to no performance gains" or "performance parity" (direct quote from Intel), and the actual gaming performance regression that a lot of people are seeing.
Are you expecting from reviewers to praise Intel for delivering a performance regression after years of refreshing the same architecture (read 12th Gen Alder Lake)? Yes, they’ve reduced power usage, but the 14th Gen is overclocked heavily from the factory, so it’s not as significant achievement as it seems. Arrow Lake’s power consumption is still much higher than AMD’s, despite Intel using a more advanced node from TSMC.
@TechTesters I wasn't speaking about this specific video. I was referring to the overall comments from the Tech Media regarding the new CPU launch last week.
Unfortunately, Intel's architecture isn't fully leveraging TSMCs 3nm which is why you are seeing consumption is still higher. Your response alone indicates you are clearly out of your depth.
Long time no see
nop .. 8200 are expenseve now and is crap for future ... CuDIMM coming later in stables bios an show power
These are CUDIMMs though?
Your English is awesome. But i prefer Super Flower )
Intel is dying slowly but surely, sad
I really hope Intel somehow figures out how to compete with AMD in gaming. Otherwise AMD's gonna keep increasing prices. We already see 7800X3D hitting $500-$600 mark.
CUdimm is comming soon. That will give even a bigger boost in performance. I think all in all the new Intel cpu is great. It just need some good tweaks and adjustments. Turning DLVR of on the core ultra 9 will save a lot of power when using it in multithreaded load like cionebench with a lower voltage input. It can easy save 80 watts as Der8auer showed on his channel.
So when Windows is optimizing the new core and games as well the sun will shine again for Intel. To many people are to negative just because it's Intel and are not looking without bias to the product.
This is a CUDIMM kit. But definitely still waiting for some Windows updates for 24H2.
Intel’s new cpus are great. The internet is just trying to pump out AMD’s supplies before their new release
No its shit.
@ hmm I haven’t come to the same conclusion. Maybe mine is faulty and is getting much better results than these UA-camrs
Must have watched the JayzTwoCents video.Jay also used 8400 MT
The video that was posted 20h ago? Unfortunately, it takes days of testing and retesting with all the new updates to get enough data for this video. Which is also the reason I haven't had time to watch his video just yet. Ill check it out tonight! always interesting to see all the data and different perspectives on the subject.
So 6400 and 7200 is the sweet spot. But it the faster ram is more then 5% higher in price then it makes no sense
NO! its 8800MHz CUDIMM with cache manual set to 44x for 1:1 with memory controller.
Depends on whether you calculate the cost of just the memory, or the cost of your total system :P
I sort of feel bad for any gamer that doesn't follow tech news and ends up with one of these CPUs in their system. When AMD is out there. Coming from a recently upset 14900K customer that I got refunded for a 7800X3D.
Ps before you reply "bUt NoT aLL bUyERs ArE gAmERzz!" Well I said I feel bad for gamers.
Don't worry, you get used to whataboutisms :P
A lot of gamers will not notice a difference at the settings they actually play on.
@@Tdisputations 1% lows, a lot are bound to notice. When you're at that level of gear, you're more than likely tracking performance. I'd agree in the low-mid tiers of CPUs but not on the top end.
@ Depends on the game, I guess. I saw a 40 game average 1% lows of about 117fps on hardware unboxed for the 265k. That wouldn’t be noticeable to me.
@@Tdisputations that was at 4k resolution though right? Still a lot of gamers buy on the high end to push more frames on 1440p and 1080p high hz displays. At 4k still better off to sce your pocket a big chunk of change and have better performance.
hello 🌷
Yammy yammy
Lol………..Intel’s dead architecture.
mb
If you turn core isolation off on a PC with sensitive data, you're mad. 100% mad
who ever plays at 1080p is stuck in the dark ages,,,get a grip
It is well accepted amongst reviewers that you use a low resolution and a high end GPU to remove any GPU bottleneck from CPU results.
For GPU testing Nada uses 1440p and 4K like everyone else. As well as 1080p which is still the most popular resolution on the Steam Hardware survey
@@clifflenoir4323 hugs
Now retest with CUDIMM.
Did you watch the whole video, read the description and/or read the pinned comment? :)
So, it's still a fail then.
I wouldn't call it a win anyway :P
Your charts are not clear actually.