But what of other applications ? And what about the fact that every benchmark site in the world have only the "old" benchmarks. I doubt everybody will redo their benchmarks.Hardwareluxx did some early tests to assess the impact of the Basline Profile on performance in MT benchmarks and games. Content is in German but like the autor says, it's self-explanatory.
Intel Baseline Profile: Neue Benchmarks nach Intels Vorgaben - Hardwareluxx
Intel Baseline Profile: Neue Benchmarks nach Intels Vorgaben.www.hardwareluxx.de
Expect most games to remain unaffected, while a few titles will feel the hit.
The top entry is the winner.Edit: I just looked at the site. Its not clear to me which is worse, as its not noted "higher is better" or "lower is better" and a limited number of benchmarks, and nothing of the competition either.
No matter what you think, here are the facts:I really don’t get why they have to add “Intel fail safe” for SVID. That’s like half of the performance loss in that 1 setting. That basically adds +50mv or more to vcore across entire curve. These settings appear to be designed for people that already have degraded silicon. Trying to avoid a bunch of RMAs I suppose.
For a fresh brand new CPU, I would keep all the same settings but keep SVID at typical scenario. You won’t experience any degradation / crashes. It’ll still lose some performance but it won’t be as extreme.
Inb4 ARL launch where Intel uses RPL with the Multicore Enhancement setting for ARL's efficiency gain claims while using RPL with the Intel Baseline Profile setting for ARL's performance gain claims.Hope this is the last time Intel depends on maxing out their CPUs to the limit to win benchmarks. Going forward, they will need to improve IPC at the baseline settings otherwise the extra performance will be worthless in any serious application workloads, thanks to them shaking the confidence of their users. No one aware of this fiasco will ever trust a maxed out Intel consumer CPU to finish their work faster lest they want bit errors creeping into it. The days of enjoying fastest workstation performance on Intel consumer CPUs may well be over. Looking forward to Puget System's and TechPowerUp's analysis of the baseline performance downgrade on the 14900K/KS.
This couldn't come at a worse time for Intel, with the impending launch of Zen 5. Now supercharging Arrow Lake with crazy settings is out of the question for them. They have limited time left to accept defeat or do something really insane (like innovate ) to save face by the end of the year. Of course, this all depends on Zen 5 meeting expectations.
I don’t disagree? It’s clear they were negligent.No matter what you think, here are the facts:
1) Users are having crashing problems
2) Intel admits they know the cause
3) Intel has provided a fix for the problem.
I guess Schadenfreude is something best served cold then!3DPM is Ian's code optimised by an ex-Intel guy for AVX-512.
How do you shoot your own foot? Just ask Intel!However in the post Alder-Lake Intel world, Intel no longer support AVX-512 so now AMD wins those benchmarks!
You shoot yourself in the foot by keeping your eyes on the rearview mirror instead of looking ahead through the windshield.How do you shoot your own foot? Just ask Intel!
Return it for running at stock ? It seems like no matter what, this is still an issue, and you should run the Intel "safe" config and put up with the lower performance.Just tried this and my initial impression was wrong. You *have* to run Intel failsafe for SVID otherwise it completely wrecks performance. By increasing the voltage you’re decreasing current and not running into the aggressive current limit.
Without running Intel fail safe I was struggling to even hit 200W of power consumption and performance was down by like 25%. So if you’re enabling IA CEP & SA CEP, you have to run a much more aggressive voltage curve. It’s clear that this configuration is really aggressive and probably intended for a chip that’s already degraded by a noticeable amount and you’re just trying to get something stable.
If you’re installing a new chip or haven’t experienced degradation I would run it with default SVID, IccMax=307A and PL2 <=253W. With this configuration there is little compromise. If your chip can’t run this configuration, I would try to return it and run the settings above and it should be okay.
I’m not sure if that’s stock. Stock is the configuration in the BIOS listed “enforce Intel limits” that enforces current and power limits. This is a step beyond that where you’re running a fail safe configuration where it no longer runs at “stock” because the chip has been fried.Return it for running at stock ? It seems like no matter what, this is still an issue, and you should run the Intel "safe" config and put up with the lower performance.
I highly ironic twist, last month MSI released a firmware that allowed disabling Current Excursion Protection for non-K 14th gen CPUs. They even had bar-chart graphs showing the performance uplift, since CEP interfered with undervolting the CPU.Just tried this and my initial impression was wrong. You *have* to run Intel failsafe for SVID otherwise it completely wrecks performance. By increasing the voltage you’re decreasing current and not running into the aggressive current limit.
Without running Intel fail safe I was struggling to even hit 200W of power consumption and performance was down by like 25%. So if you’re enabling IA CEP & SA CEP, you have to run a much more aggressive voltage curve. It’s clear that this configuration is really aggressive and probably intended for a chip that’s already degraded by a noticeable amount and you’re just trying to get something stable.
Why can't there be a baseline profile that's actually just setting the baseline values and nothing else, and another profile for salvaging degraded chips? It feels like everything is second guessing the desired functionality so customers have to second guess the settings as well, which is exactly what led to the current mess to begin with.Just tried this and my initial impression was wrong. You *have* to run Intel failsafe for SVID otherwise it completely wrecks performance. By increasing the voltage you’re decreasing current and not running into the aggressive current limit.
Without running Intel fail safe I was struggling to even hit 200W of power consumption and performance was down by like 25%. So if you’re enabling IA CEP & SA CEP, you have to run a much more aggressive voltage curve. It’s clear that this configuration is really aggressive and probably intended for a chip that’s already degraded by a noticeable amount and you’re just trying to get something stable.
If you’re installing a new chip or haven’t experienced degradation I would run it with default SVID, IccMax=307A and PL2 <=253W. With this configuration there is little compromise. If your chip can’t run this configuration, I would try to return it and run the settings above and it should be okay.
Benchmarks may not get updated until a fresh new batch of reviews comes up with Zen 5 / Arrow Lake (the first one to launch will refresh the benchmark landscape).But what of other applications ? And what about the fact that every benchmark site in the world have only the "old" benchmarks. I doubt everybody will redo their benchmarks.
Edit: I just looked at the site. Its not clear to me which is worse, as its not noted "higher is better" or "lower is better" and a limited number of benchmarks, and nothing of the competition either.
Because that would admit the chips are degraded, opening Asus/Intel to financial or legal trouble.Why can't there be a baseline profile that's actually just setting the baseline values and nothing else, and another profile for salvaging degraded chips?
And just like that we still don't have any way to force baseline settings even with a "baseline" profile being introduced. Completely crazy.Because that would admit the chips are degraded, opening Asus/Intel to financial or legal trouble.
I think a bit more nuance is in order here. My impression is we are technically getting a proper baseline profile, just not the one that enthusiasts such as @H433x0n would be looking for. It's the... uhhm... Intel non-Enthusiast Baseline Profile for Enthusiasts Processors SeriesAnd just like that we still don't have any way to force baseline settings even with a "baseline" profile being introduced. Completely crazy.
The microcode has enabled CEP disablement on a non-K processor and a non-Z chipset motherboard. However, there are still limitations. The microcode only brings the ability to disable CEP on the latest Raptor Lake Refresh non-K, Raptor Lake K-series, and Alder Lake K-series processors. Raptor Lake non-K and older chips are out of luck on this one.
| CPU / Chipset | Z790/Z690 MB | B760/B660 MB |
|---|---|---|
14th Gen K-series | Yes (already supported with previous microcode) | Yes |
14th Gen non-K series (BO stepping) | Yes | Yes |
14th Gen non-K series (other stepping) | No | No |
13th Gen K-series | Yes | No |
13th Gen non-K series | No | No |
12th Gen K-series | Yes | No |
12th Gen non-K series | No | No |
Benchmarks may not get updated until a fresh new batch of reviews comes up with Zen 5 / Arrow Lake (the first one to launch will refresh the benchmark landscape).
That being said, this may not be in the advantage of Intel, because another type of "news" will echo through the media. Here's how the article from the german website is reflected on another US tech website:
Some Intel CPUs lost 9% of their performance overnight | Digital Trends
A new Asus motherboard update aimed at addressing stability issues on Intel CPUs could shave upwards of 9% of your performance.www.digitaltrends.com
You no longer need to understand which is worse or better, the headline says all you need to know.
That's really weird. What kind of power usage causes degradation in Raptor Lake CPUs over the long term? Alder Lake was pushing 200W+ and has been for quite awhile without having these problems . . .Looks like Gigabyte believe wattage should be less than 188w while short run, and within 125 watts while long run.