*Official Benchmark Thread* - Post it here or it didn't happen :D

[cylix]

I have a Artic 240 III bought and installed on a secondary system i can tell aswell what the users are saying about them, the QC of artic is shit! The pump is doing strange sounds from time to time and you can hear water splashes inside the aio...also the mount system on AMD is  really bad designed, but gamer nexus said about that. But no review said anything about QC ..probably they got special units...

[Papusan]

On 4/9/2024 at 2:56 PM, Reciever said:

 

Well its worth investigating to determine if it was a faulty unit (which is also a concern) or faulty process. Doesnt seem like they are going to invest more time into it though.

 

Hmmm. It seems Tomshardware tested 3 different models. They probably used only one of the contack frames. And we all know that the contact frame doesn't work equal good on every and different motherboards. Some will benefits from it and some offer nothing. Some even offer worse results.

 

 

For the pict jockey @electrosoft😃 End results... The Black crocks will most likely offer me the best value. Remember new/newer doesn't always mean better, LOOL

 

Both Batch# X346P377T - I mean it's same batch as bro @johnksss And this bought here home 10000km from any California store. The world is so damn small, LOOL

 

[electrosoft]

I ordered a few motherboards from Amazon Warehouse: Asus B650 Prime and Asus X670E Strix for $95 and $228 for my daughter's retrofit to a full Team Red build. I figured if I can't get the Asus B650 Prime working properly (They have a high rate of return / problems) I can always go with the Strix.

 

I also picked up a 7600x on ebay for $165.

 

So far after a few headaches (I can see why they have a high return rate), the Asus B650 is holding it down. I can't see how Asus charges $200 for this board though. No debug lights at all and minimal heatsinks on the board. I must admit I do like the overall look though for some reason. The early BIOS revisions were to fix a crapton of memory incompatibilities for EXPO (still an issue), but since I manually tune all my memory it is not an issue.

 

Main issue was reported overheating VRMs. I swapped in my 7950X3D as a test and let it run CB23 for 20 minutes and they do get toasty but the 7600x doesn't even break 50 running the same test (In summary, if you're buying a 7950X/X3D, get a board worthy of it).

 

As for this 7600x, it is the first AM5 chip I've encountered that can do 2200 fclk no problem. With the newest Agesa updates, running 6000 tuned older M-die sticks, fclk 2000 vs fclk 2200 = bandwidth increase of 63k ->69k which makes sense considering the fclk increases by 10%.

 

uclk holds to 6400 1:1 but gives up the ghost at 6600. My pipe dream of perfect ratios for fclk 2200 were quickly dashed. I had to swap in my 2x32GB G.Skill A-die sticks for that test since these first gen M-die's (I picked them up in 2022) give up the ghost at 6600 (uclk = 1/2).

 

Using the already dialed in timings that had been tested on my 13900ks (MSI Z790i Edge), 12900k (Z690 unify), 7800X3D/7950X3D (X670e carbon, Asrock B650 HDV, Asus B650 Prime), I kept getting flickering that intensified during memory tests. Trial and error, I found out Tras was a bit too aggressive for the iGPU so I backed off from 30 to 35 and all was well.

 

I'll run some more tests this week in the test case including swapping in the 7900XTX for final testing before I break down her system for the retrofit.

 

 

 

[cylix]

New rumors that 5090 will launch this year 🫣🫠. How hard do you think it will be to get one? I am contemplating to sell my 4090 now and get some good money on it as to wait until 5090 will reach the market, probably it will lose a chunk  off value then...

I dont play that much anymore  only  some not demanding games like diablo 4 and battlefield 2042, games that i can run on the 3080 Asus Turdbook or get a 3070 or 3080 until 5090. But i dont want to not be able to get one on launch date or short after that..

https://wccftech.com/nvidia-geforce-rtx-5090-rtx-5080-blackwell-gpus-launch-q4-2024-rumor/

[electrosoft]

1 minute ago, cylix said:

New rumors that 5090 will launch this year 🫣🫠. How hard do you think it will be to get one? I am contemplating to sell my 4090 now and get some good money on it as to wait until 5090 will reach the market, probably it will lose a chunk  off value then...

I dont play that much anymore  only  some not demanding games like diablo 4 and battlefield 2042, games that i can run on the 3080 Asus Turdbook or get a 3070 or 3080 until 5090. But i dont want to not be able to get one on launch date or short after that..

https://wccftech.com/nvidia-geforce-rtx-5090-rtx-5080-blackwell-gpus-launch-q4-2024-rumor/

 

What I've garnered so far is 5090 = same node / process as 4090 pumped up more but the majority of gains will be from GDDR7 and it's gonna run hot and heavy. I am curious about the performance gains since the 4090 was such a bump over the 3090. Part of me expects smaller than expected gains as they are cramming in more voltage and while the AI sector is looking at an MCM design, the consumer market looks to get a big ole juicy monolithic die again.

 

I plan on waiting till after it drops (or close to it). I do too much Fallout 76 and WoW to suffer for months without top notch 4k performance. if the 7900xtx wasn't such a crapshow with WoW, I could see using it as a stop gap but for WoW it is garbage with all the driver issues.

 

I've actually had a few people reach out (3) wanting to buy my 4090 because it auto boosts to 2820+ right out of the box and with a basic OC on the stock AIO cooler does 3120+. Mem  >=+1500. I imagine on a chiller and a real block with a real XOC bios it might turn into a monster if it does this on stock gear.

 

The major win for me is basically zero coil whine to the point I not only run my case fully open but I no longer have to use headphones and can play like I used to.

 

With all this said, if a model on launch I want pops up for retail I will most likely pull the trigger but if I miss it or have to wait till Jan/Feb (which is what happened with my 4090), I'm fine with that too. I'll probably lock into MSI, FE or Asus again (in that order).

 

 

 

 

 

 

[Papusan]

Here's more disgusting from intel. SP99 (115+69) MC 81. Just as expected. Batch# X346P377T 🤮🤢🤮 For the time I'm still not in good shape but this doesn't make my mood better. This is just sad.

 

And this defective KS fits correctly where it should. 

 

https://www.overclock.net/threads/overclocking-raptor-lake-refresh-14900k-14700k-14600k-etc-results-bins-and-discussion.1807439/post-29317725

 

This is where Intel should put their new binned KS chips. Should never ever have been shipped out from Intel factories. 

[jaybee83]

On 4/7/2024 at 8:08 PM, electrosoft said:

When some of the mainstream YTers try to say faster memory is really a waste and no real gains, then you get B4B who actually does some tuning comparing 6000 vs 8600 for gaming.

 

This is right along the lines of the gains I saw 6000 vs 8000 tuned on my 13900KS + MSI Z790i Edge. The gains are real and substantial especially in WoW.

 

 

 

On 4/7/2024 at 8:14 PM, Mr. Fox said:

They have to spread misinformation. How else can they legitimize mediocrity? 🤣

 

welp as with all things related to performance on hardware: every program / benchmark / games has different bottlenecks. there are tons of games that do not benefit from any RAM OC while others see a big boost in min fps or even avg 🙂 always depends how and what ure testing i wager.

 

7 hours ago, cylix said:

I have a Artic 240 III bought and installed on a secondary system i can tell aswell what the users are saying about them, the QC of artic is shit! The pump is doing strange sounds from time to time and you can hear water splashes inside the aio...also the mount system on AMD is  really bad designed, but gamer nexus said about that. But no review said anything about QC ..probably they got special units...

 

huh cant say i have any complaints about my AC LFII 420. no pump noise, no water noises, performance as expected. whats the age and mounting orientation? also make sure its not one of those AIO batches affected by the faulty gaskets:   https://www.arctic.de/en/lf-service-kit

 

2 hours ago, cylix said:

New rumors that 5090 will launch this year 🫣🫠. How hard do you think it will be to get one? I am contemplating to sell my 4090 now and get some good money on it as to wait until 5090 will reach the market, probably it will lose a chunk  off value then...

I dont play that much anymore  only  some not demanding games like diablo 4 and battlefield 2042, games that i can run on the 3080 Asus Turdbook or get a 3070 or 3080 until 5090. But i dont want to not be able to get one on launch date or short after that..

https://wccftech.com/nvidia-geforce-rtx-5090-rtx-5080-blackwell-gpus-launch-q4-2024-rumor/

 

hmmm that is always a dangerous game to sell off the gpu before getting the successor / replacement... depends on whats more important to you: to make some profit on selling your 4090 before prices drop or to avoid any system downtime. i for one will go the second route, since i dont want to risk not being able to snatch up a 5090 at MSRP-ish prices or having to go the scalper route (or even worse: not being able to get one at all). the potential profit from the 4090 is not worth to potentially be sitting on a non-working highend system for months.

[ryan]

about 12 sources are saying the 5090 will likely be a 1.7x or 77 percent uplift. if the accuracy turns out, its pretty exciting news, thats a pretty big bump with its supposed 24 000 cuda cores and gddr7. if it does end up being 70 percent faster and I buy one then I will definately be good for 4k gaming for about 5 years and that is worth the 2k price, but price wise no one really knows but I expect the typical you get more but you pay more pricing.

[Mr. Fox]

2 hours ago, jaybee83 said:

welp as with all things related to performance on hardware: every program / benchmark / games has different bottlenecks. there are tons of games that do not benefit from any RAM OC while others see a big boost in min fps or even avg 🙂 always depends how and what ure testing i wager.

Yes sir. And, the talented "influencers" and spreaders of falsehoods are very deliberate in choosing the information that fits their narrative. They are more like salesmen than technicians.

[jaybee83]

19 hours ago, ryan said:

about 12 sources are saying the 5090 will likely be a 1.7x or 77 percent uplift. if the accuracy turns out, its pretty exciting news, thats a pretty big bump with its supposed 24 000 cuda cores and gddr7. if it does end up being 70 percent faster and I buy one then I will definately be good for 4k gaming for about 5 years and that is worth the 2k price, but price wise no one really knows but I expect the typical you get more but you pay more pricing.

depends on your expectations for 4K gaming @5 years. 6 years ago the 2080 Ti was released, compare its performance in today's  AAA titles at 4K with RTX on and ull see what u potentially get with a 5090 after 5 years 🙂 

 

19 hours ago, Mr. Fox said:

Yes sir. And, the talented "influencers" and spreaders of falsehoods are very deliberate in choosing the information that fits their narrative. They are more like salesmen than technicians.

yeah could be deliberate or just plain ignorance 😄 

[electrosoft]

51 minutes ago, ryan said:

about 12 sources are saying the 5090 will likely be a 1.7x or 77 percent uplift. if the accuracy turns out, its pretty exciting news, thats a pretty big bump with its supposed 24 000 cuda cores and gddr7. if it does end up being 70 percent faster and I buy one then I will definately be good for 4k gaming for about 5 years and that is worth the 2k price, but price wise no one really knows but I expect the typical you get more but you pay more pricing.

 

I don't think so for the consumer model but I can certainly hope so. A 77% uplift is massive but an uplift of 50% more cuda cores on the same node is going to be a 600-700w+ monstrosity at stock. I could reasonably see ~25% more cuda cores and GDDR7 carrying a lot of the heavy lifting for a 500-600w card and ~30-50% uplift.

 

$2k has been the rumored price for awhile now and you would be getting some major bang for the buck if that is true but then again I don't want a space heater in my room like the 3090ti was either.

 

If we get that monster uplift (again doubt it), that is an instant buy and sell for myself. We'll revisit this in October/November if it actually launches this year.

 

 

 

 

[Papusan]

Second one. SP105 (118/79) this time and MC58. So looks very low but match my memory sticks. I still prefer higher clock speed over fastest possible ram speed for my benches and hobby. But the scores I have done so far looks good. Close to 45000 in CBR23 with 59/48/50.

 

At least this sample have okish E-cores that can easly run 4.8GHz. Thats something isn't it, Huhhh. Not sure I will bother with a 3rd one and return this. Could be I get an equal disgusting chips as the SP99 I got in same shipping. The chance is big getting SP100 and below. It's what it is. Now I need to clean up the SP99 for return to the shop. Or tomorrow.

 

I use kapton tape to reduce the wear from the ILM. Better be safe than sorry. Not fun if they don't want this SP99 trash back due scratches🙂

 

 

 

 

 

 

Maybe we will see Intel offer more tops for the refresh as etc from 15th to 16th gen. Tops over real Cpu performance (more cores). Yup this goes the wrong way. Want tops bro @Mr. Fox? Instead of tick tock we get tick and tops, LOOL

 

Be sure UL/Futuremark will have a bite of this AI craziness. AI benchmarks will be the next big. And you’ll get tops instead of points. 

AI Craze May Have Nerfed AMD’s & Intel’s Upcoming Chips: Strix APUs Originally Had Big Cache Which Boosted CPU & iGPU Performance 🤮🤔🤮

These AI blocks will take up large portions of valuable die space that could've been dedicated elsewhere such as higher core counts, higher iGPU counts, wider caches, and more but it looks like the AI PC craze has made chipmakers take a backseat on standard CPU / iGPU performance and focus more on the NPU side of things.

[ryan]

well its coming soon enough, slightly before gta 6. 4th Q of 2024 so not a long wait, and even if the leaks are wrong im sure its going to be a worthwhile upgrade for you guys and for me, monster level upgrade from my 3060. 5years because I bought mid range 3 years ago and its still going strong at 1080p which was the target, and it will last another 3 years at 1080p. if I get the 5090 and the target is 4k/60 LOW then I really do see it lasting as long as the 1080ti did. alot of people don't care about ultra settings so long as textures are cranked up and with the 32gb of vram and talks of turning ssds into vram I think it will be safe. Im getting a pretty good chunk of money for whiplash and a concussion, just have to do physiotherapy for a year. So im likely going to be spending 5k but I don't want to comprimise on ram/MB/GFX/cpu I will be getting the highest end and it will be interesting benchmarking in the hardware space of desktops. If its anything like laptops it should be fun.

[Mr. Fox]

Not too bad for a CPU that I initially thought wasn't good enough to keep, even if it can't be push to 8400 or higher. I'm probably going to keep it. It's in my work PC, not my benching rig. Passes TM5 and 100% on MemTest Pro Dang Wang and all four passes of MemTest86 Pro with Parallel/HT enabled. SA voltage bug, so this 8200 C36 is totally stable with only 1.120V manually set on VCCSA. Leaving VCCSA set to Auto or manually setting higher than 1.120V causes lock-ups in Windows.

---

Edit: Just tested 8267 and that works fine as well. No changes to voltage or settings, only goosed it a bit from 8200 to 8267 with VCCSA 1.120V.Edit: Just tested 8267 and that works fine as well. No changes to voltage or settings, only goosed it a bit from 8200 to 8267 with VCCSA 1.120V.

Edited April 10 by Mr. Fox
tested 8267 and that works

[electrosoft]

4 hours ago, Mr. Fox said:

Not too bad for a CPU that I initially thought wasn't good enough to keep, even if it can't be push to 8400 or higher. I'm probably going to keep it. It's in my work PC, not my benching rig. Passes TM5 and 100% on MemTest Pro Dang Wang and all four passes of MemTest86 Pro with Parallel/HT enabled. SA voltage bug, so this 8200 C36 is totally stable with only 1.120V manually set on VCCSA. Leaving VCCSA set to Auto or manually setting higher than 1.120V causes lock-ups in Windows.

---

 Edit: Just tested 8267 and that works fine as well. No changes to voltage or settings, only goosed it a bit from 8200 to 8267 with VCCSA 1.120V.Edit: Just tested 8267 and that works fine as well. No changes to voltage or settings, only goosed it a bit from 8200 to 8267 with VCCSA 1.120V.

 

That's not a bad chip at all. Outside of the SA bug (booo!) the actual performance and V/F curve looks rather nice! On your real hardware, this will break 45k easily.

[Mr. Fox]

1 hour ago, electrosoft said:

 

That's not a bad chip at all. Outside of the SA bug (booo!) the actual performance and V/F curve looks rather nice! On your real hardware, this will break 45k easily.

This CPU might be ideal for the Munchkin PC's Z790i Edge with its pedestrian 360mm AIO and limited power phase capacity. I'm not pushing that past 8000 on the memory, so maybe the perfect match of CPU to system. On my benching rig I can just turn on the chiller and force more voltage down its throat to get it where I want it. That said, I am more impressed by the ITX system's abilities than I have been with the more costly builds. It's not equivalent, but better ratio of cost to benefit. Sometimes spending a lot more to get a little more is a tough pill to swallow even though doing more with MOAR is my favorite approach. Nobody likes getting screwed on overpriced products. The ROG Maximus mobos are a classic of example of paying a whole lot more and only getting a little more in return. You do get more, but it is not scaled against the price.

[jaybee83]

here we go boiiiiizzzzz! Granite Ridge Zen 5 CPU support in new Asus Beta Bios versions *w00t w00t*

 

https://videocardz.com/newz/asus-x670-agesa-1-1-7-0-update-enables-initial-support-for-zen5-granite-ridge-cpu-series

 

already got myself the latest 2005 for my X670EEEEEEEEEEEEEEEEEEEEE 😛 😄 

[Raiderman]

3 hours ago, jaybee83 said:

here we go boiiiiizzzzz! Granite Ridge Zen 5 CPU support in new Asus Beta Bios versions *w00t w00t*

 

https://videocardz.com/newz/asus-x670-agesa-1-1-7-0-update-enables-initial-support-for-zen5-granite-ridge-cpu-series

 

already got myself the latest 2005 for my X670EEEEEEEEEEEEEEEEEEEEE 😛 😄 

I read that the other day, about Asus releasing support for zen 5. Should mean a sooner rather than later? I'm ready 🤙

[Papusan]

It seems bios version 9905 looks decent. So no point jumping on newer bios versions. Good CBR23 scores but this Cpu is still not worth delidding. 

 

LOOOL. One wimpy point from 46.000 in CBR-23 with this dude🤗

https://hwbot.org/submission/5532413_papusan_cinebench___r23_multi_core_with_benchmate_core_i9_14900ks_45999_cb/

 

Edit. Yup, here’s +46.000 in CBR-23

https://hwbot.org/submission/5532429_papusan_cinebench___r23_multi_core_with_benchmate_core_i9_14900ks_46036_cb/

 

CBR-23 without baby cores. 8 real cores is better than double up with E-cores. 

https://hwbot.org/submission/5532469_papusan_cinebench___r23_multi_core_with_benchmate_core_i9_14900ks_(8p)_26082_cb

[Papusan]

Shouldn't all modern Intel processors throttle down clockspeed and voltage instead for instant crashing when reaching 100C? Hmmm. Same processors in jokebooks can do it and can run 100C without crashes... Why not with desktops? Hmmm, I expect mostly all the problems people have seen with unstable/crashing 13th and 14th gen processors is because they don't have good enough cooling. This is still a problem Intel and the MB manufactures need to fix. And Intel's trash binning doesn't exactly help either.

 

 

Intel Core i9-14900KS Review pcmag.com

 

Thermally, the Core i9-14900KS was problematic when used with our 240mm water cooler, hitting 100 degrees C quickly and soon crashing, likely due to excessive heat. Used with a 360mm water cooler, the processor still hit 100 degrees C, which is why I didn't add a separate listing for it in the chart. The key difference is that the extra cooling performance helped the processor to overheat less frequently, which would enable it to run at turbo clocks more often and more stably. This is why it managed to run the Cinebench test without crashing and showed a slight performance increase with a 360mm cooler.

 

 

Intel To Discontinue Boxed 13th Gen Core CPUs for Enthusiasts

 

Interestingly, the retirement of the enthusiast-class 13th Generation Core chips is coming before Intel discontinues their even older 12th Generation Core 'Alder Lake' processors. 12th Gen chips are still available to this day in both boxed and tray versions, and the Alder Lake silicon itself is still widely in use in multiple product families. 

 

 

Yup, all this just show where we go......... The tech trend continue. Downwards.

Ex-Microsoft veteran calls out Windows 11's "comically bad" performance, 'unfinished' state

Sayan Sen · 19 hours ago

One of the promises Microsoft made about Windows 11 was about the performance benefits it had. A former Microsoft engineer, however, has called Windows 11 out on its "comically bad" performance.

 

Former Microsoft engineer Andy Young is one of those who have noticed the slow and laggy nature of Windows 11 and decided to call out the Redmond giant. Young remarks he has a $1600 Intel Core i9 PC with 128 GB system memory, and yet the Start menu's performance is "comically bad."

[Talon]

3 hours ago, Papusan said:

Shouldn't all modern Intel processors throttle down clockspeed and voltage instead for instant crashing when reaching 100C? Hmmm. Same processors in jokebooks can do it and can run 100C without crashes... Why not with desktops? Hmmm, I expect mostly all the problems people have seen with unstable/crashing 13th and 14th gen processors is because they don't have good enough cooling. This is still a problem Intel and the MB manufactures need to fix. And Intel's trash binning doesn't exactly help either.

 

 

Intel Core i9-14900KS Review pcmag.com

 

Thermally, the Core i9-14900KS was problematic when used with our 240mm water cooler, hitting 100 degrees C quickly and soon crashing, likely due to excessive heat. Used with a 360mm water cooler, the processor still hit 100 degrees C, which is why I didn't add a separate listing for it in the chart. The key difference is that the extra cooling performance helped the processor to overheat less frequently, which would enable it to run at turbo clocks more often and more stably. This is why it managed to run the Cinebench test without crashing and showed a slight performance increase with a 360mm cooler.

 

 

Intel To Discontinue Boxed 13th Gen Core CPUs for Enthusiasts

 

Interestingly, the retirement of the enthusiast-class 13th Generation Core chips is coming before Intel discontinues their even older 12th Generation Core 'Alder Lake' processors. 12th Gen chips are still available to this day in both boxed and tray versions, and the Alder Lake silicon itself is still widely in use in multiple product families. 

 

 

Yup, all this just show where we go......... The tech trend continue. Downwards.

Ex-Microsoft veteran calls out Windows 11's "comically bad" performance, 'unfinished' state

Sayan Sen · 19 hours ago

One of the promises Microsoft made about Windows 11 was about the performance benefits it had. A former Microsoft engineer, however, has called Windows 11 out on its "comically bad" performance.

 

Former Microsoft engineer Andy Young is one of those who have noticed the slow and laggy nature of Windows 11 and decided to call out the Redmond giant. Young remarks he has a $1600 Intel Core i9 PC with 128 GB system memory, and yet the Start menu's performance is "comically bad."

 

I dunno, those guys clearly have no idea what they're doing. 

 

Running Intel Fail Safe makes absolutely no sense. It shouldn't even exist in the menu. As it reads, that setting follows an absolute fail safe voltage guidance that is likely when using an Intel Reference Motherboard for testing/design. The VRM and electrical layout of that board is not going to be anywhere near as robust or capable as the Apex board they were using. They are shoving a ton of voltage and current through the chip, something the reference board is very likely not capable of supplying in the same capacity due to the VRM design. 

 

All they have to do is set Auto, XMP I, set 253w for long/short power, and limit the board to around 400a and the chip would have been running far cooler and stable. If this config is not stable, then it's time for an RMA as the chip is not able to run at SPEC. Simple as that.

 

Unfortunately, they are spreading FUD, they are giving out terrible advice, and perpetuating and furthering ignorance in the community. Simple as that. 

 

I blame the motherboard vendors. They should be setting Intel spec limits by default and forcing you to bypass them on your own afterwards. Asus does in fact as this question upon any CPU swap on FIRST boot. It will ask you if you want Intel limits or Asus optimized. Almost everyone selects Asus limits which remove all limits. Intel should force the vendors to enable stock limits, and then let those that want, bypass them. 

[Papusan]

4 hours ago, Talon said:

I blame the motherboard vendors. They should be setting Intel spec limits by default and forcing you to bypass them on your own afterwards. Asus does in fact as this question upon any CPU swap on FIRST boot. It will ask you if you want Intel limits or Asus optimized. Almost everyone selects Asus limits which remove all limits. Intel should force the vendors to enable stock limits, and then let those that want, bypass them. 

 

Hmm. Intel gained from it. Higher scores for the reviews (who know what Intel's engineers have said to the MB OEMs). 

 

https://wccftech.com/intels-lga-1851-socket-pictured-debut-meteor-lake-ps-arrow-lake-core-ultra-200-ready/

 

Not much have changed with LGA 1851 vs 1700. 

 

 

 

The Cablemod saga continue. Loads of burned angled adapters/4090's. 

 

[Papusan]

17 hours ago, Talon said:

I blame the motherboard vendors. They should be setting Intel spec limits by default and forcing you to bypass them on your own afterwards. Asus does in fact as this question upon any CPU swap on FIRST boot. It will ask you if you want Intel limits or Asus optimized. Almost everyone selects Asus limits which remove all limits. Intel should force the vendors to enable stock limits, and then let those that want, bypass them. 

nvidia point to Intel😄 Hmmm. Why haven't wee see much of it the first 12-15 months after launch of 13900K in the fall 2022? Only changes is new bioses, drivers and Windows updates. 

NVIDIA Points Intel Raptor Lake CPU Users to Get Help from Intel Amid System Instability Issues

Today, 12:07

According to recently published help guide, spotted by the X/Twitter user @harukaze5719, NVIDIA has addressed reported stability problems users are experiencing with Intel's latest 13th and 14th generation Raptor Lake Core processors, especially the high-performance overclockable K-series models. In a recent statement, NVIDIA recommended that owners of the affected Intel CPUs consult directly with Intel if they encounter issues such as system instability, video memory errors, game crashes, or failures to launch certain applications. The problems seem particularly prevalent when running demanding workloads like gaming on Unreal Engine 5 titles or during shader compilation tasks that heavily utilize the processor and graphics capabilities. Intel has established a dedicated website to provide support for these CPU instability cases. However, the chipmaker still needs to issue a broad public statement and provide a definitive resolution.

The instability is often attributed to the very high frequencies and performance the K-series Raptor Lake chips are designed to achieve, which are among the fastest processors in Intel's lineup. While some community suggestions like undervolting or downclocking the CPUs may help mitigate issues in the short term, it remains unclear if permanent fixes will require BIOS updates from motherboard manufacturers or game patches. As the two tech giants work to pinpoint the root causes, owners of the affected Intel Raptor Lake processors are advised to reach out to Intel's support channels for troubleshooting and potential workarounds for now.

[Mr. Fox]

1 hour ago, Papusan said:

nvidia point to Intel😄 Hmmm. Why haven't wee see much of it the first 12-15 months after launch of 13900K in the fall 2022? Only changes is new bioses, drivers and Windows updates. 

NVIDIA Points Intel Raptor Lake CPU Users to Get Help from Intel Amid System Instability Issues

The thread was created in February and there has been only one response, on the date of the article by TechPowerUp. Other than 14900KS poor quality silicon (and for a period leading up to KS bin scavenging causing shopper to not be able to purchase a decent 14900K bin) I am not aware of the issues mentioned. I find myself wondering to what degree this is TechPowerUp clickbait.

 

That said, if the SA bug (for lack of a better term) is common, it could be that BIOS default value for VCCSA is too high for the processors that use freakishly low VCCSA. I haven't turned the small errors away yet, but I was able to  move from 8266 to 8400 memory overclock last night with only 1.130V on VCCSA with no lockups. This could be what many are experiencing, not knowing their BIOS default is setting for VCCSA too high. I also can't help but wonder if it is why some believe they have a weak IMC that keeps them from running 8000+ and they are simply mistaken, meaning the only reason they can't is because their VCCSA is set too high by default and they don't know that is what the problem is.

 

I am not sure it is a bug in the truest sense. It may be the low VCCSA is what was intended and "normal" is higher VCCSA due to poor silicon quality being common and those that use very low VCCSA are the ones that are the best samples. These are usually CPUs with very good SP ratings on P/E cores and IMC SP.

 

Perhaps the problem is more accurately described as incompetent firmware engineering by motherboard OEMs and lack of R&D/QC on their part. Assuming status quo is going to work today exactly as it did yesterday, instead of actually verifying it does is what I would categorize as incompetence and an absence of R&D/QC on their part.

[Talon]

3 hours ago, Mr. Fox said:

The thread was created in February and there has been only one response, on the date of the article by TechPowerUp. Other than 14900KS poor quality silicon (and for a period leading up to KS bin scavenging causing shopper to not be able to purchase a decent 14900K bin) I am not aware of the issues mentioned. I find myself wondering to what degree this is TechPowerUp clickbait.

 

That said, if the SA bug (for lack of a better term) is common, it could be that BIOS default value for VCCSA is too high for the processors that use freakishly low VCCSA. I haven't turned the small errors away yet, but I was able to  move from 8266 to 8400 memory overclock last night with only 1.130V on VCCSA with no lockups. This could be what many are experiencing, not knowing their BIOS default is setting for VCCSA too high. I also can't help but wonder if it is why some believe they have a weak IMC that keeps them from running 8000+ and they are simply mistaken, meaning the only reason they can't is because their VCCSA is set too high by default and they don't know that is what the problem is.

 

I am not sure it is a bug in the truest sense. It may be the low VCCSA is what was intended and "normal" is higher VCCSA due to poor silicon quality being common and those that use very low VCCSA are the ones that are the best samples. These are usually CPUs with very good SP ratings on P/E cores and IMC SP.

 

Perhaps the problem is more accurately described as incompetent firmware engineering by motherboard OEMs and lack of R&D/QC on their part. Assuming status quo is going to work today exactly as it did yesterday, instead of actually verifying it does is what I would categorize as incompetence and an absence of R&D/QC on their part.

 

It's actually a very real issue, I've personally seen it myself. Unfortunately this issue happens when running completely unrestricted on a system with less than adequate cooling. The problem lies with the motherboard vendors allowing 4095w on PL1/PL2 and over 500A at "default". 

 

After a fresh driver install, many games have to recompile shaders and this hits the CPU hard, like unrestricted hard. This is why some are attributing games crashing to new drivers. It has nothing to do with Nvidia. It's the 100c temps some are seeing at unlimited power and their games are crashing, or even BSOD. 

 

The simple and easy fix is to simply run the CPU at Intel's recommended spec of 253w or 320w extreme modes on KS. This should be on by default. We have too many Xbox kiddos coming over to PC, they have no idea what they're doing, and then crying when their 240mm AIO crashes when loading up Fortnite after they spent the evening hammering the CPU with 400w loads in Cinebench. Force motherboard makers to run defaults and let those that want to, go in there and tweak. Seeing "default" Intel CPUs drawing 380w in headlines is just stupid and is not helping Intel's PR. The truth is, at 253w, the 14900K is very close to AMD's 7950x in multicore and power draw. It's the lower end where AMD tends to shine. Yet the unlimited power draw paints a completely different story. Intel needs better people in charge over there. They need to get their heads out of their asses. 

 

I would say it is completely possible Intel has poorly binned and the higher temps/current/power are causing edge case CPUs to crash, but we will have to wait and see. For now, running them at SPEC and as designed should fix all issues. 

 

For those that may be wondering, here is Intel's difficult to locate spec sheet. As you can see, there is a 253w PL1/PL2 "Extreme Config" for K/KF chips. and a 320w PL1/PL2 config for KS chips. 

 

https://edc.intel.com/content/www/us/en/design/products/platforms/details/raptor-lake-s/13th-generation-core-processors-datasheet-volume-1-of-2/processor-line-thermal-and-power-specifications/