Asus issues BIOS update to solve Intel CPU game crashing.

[humbug]

As a lot of you may already know there are reports of games crashing with the 13900K and 14900K, it seems to be specifically with Nvidia GPU's, Nvidia have publically thrown Intel under the bus by naming them as the culprit for an "out of video memory" GPU crash error on their GPU's. lol... "Out of video memory" i can see why they were so quick to throw Intel under the bus.
It makes sense as Nvidia drivers rely on the CPU to a much greater extent than for example AMD.

With the background on this given. Yes Asus strikes again, this is not an Intel problem, as such, Intel CPU's come with a TDP "Thermal Design Power" for example the 14900K has a TDP of 125 Watts, lol... yeah, this looks on a box better than a Ryzen 7950X TDP of 170 watts but in reality its more like 250 watts, it looks like Intel are in turn throwing motherboard vendors under the bus by pointing the finger at them for allowing the CPU to run far outside those TDP ratings, i can see Steve Burke already agreeing with Intel, but the truth IMO is Intel knowingly turn a blind eye to all of it, or even encourage it, because in reality at 125 watts the CPU is a lot slower than it is as they run in all reviews at about 250 watts, IE much slower than AMD's CPU's. so yes on the one hand they want to print a lower than AMD TDP on the box while at the same time want to appear like you're also getting the same or better than AMD performance for that.

So, it seems that Asus at least, and possibly others have taken this a little too far and the CPU's are not actually stable, they have issued a BIOS which basically backs the CPU off a little, like you would, well..... an unstable overclock.

Also, Intel, stop pretending your CPU's are something they are not. they either score 39K in Cinebench at 250 watts or they are a 125 watt CPU and score a lot less.

https://twitter.com/i/web/status/1781571284402356719

If you have an Nvidia GPU and get the "out of video memory" crash check your board vendor for a BIOS update.

 

[Tetras]

lol... "Out of video memory" i can see why they were so quick to throw Intel under the bus.

Yes Asus strikes again, this is not an Intel problem, as such, Intel CPU's come with a TDP "Thermal Design Power" for example the 14900K has a TDP of 125 Watts, lol... yeah, this looks on a box better than a Ryzen 7950X TDP of 170 watts but in reality its more like 250 watts, it looks like Intel are in turn throwing motherboard vendors under the bus by pointing the finger at them for allowing the CPU to run far outside those TDP ratings, i can see Steve Burke already agreeing with Intel, but the truth IMO is Intel knowingly turn a blind eye to all of it, or even encourage it, because in reality at 125 watts the CPU is a lot slower than it is as they run in all reviews at about 250 watts, IE much slower than AMD's CPU's. so yes on the one hand they want to print a lower than AMD TDP on the box while at the same time want to appear like you're also getting the same or better than AMD performance for that.

So, it seems that Asus at least, and possibly others have taken this a little too far and the CPU's are not actually stable, they have issued a BIOS which basically backs the CPU off a little, like you would, well..... an unstable overclock.

Also, Intel, stop pretending your CPU's are something they are not. they either score 39K in Cinebench at 250 watts or they are a 125 watt CPU and score a lot less.

Afaik, that's not what is happening here.

Intel has a spec of two things for their power ratings: PL1 and PL2. PL1 is the baseline speed and that aligns with the TDP, then PL2 aligns with the max turbo speed under heavy workloads. Asus (and most other manufacturers on their high-end boards are exceeding even the higher PL2 (e.g. they set it to unlimited). Asus also enable MCE by default, which allows all cores to boost to the highest turbo multiplier.

For the most part, Intel CPUs don't exceed PL2 when playing games and the performance loss of using even PL1 would be minimal, but there were some circumstances (especially when compiling shaders apparently) that they could hammer all of the cores.

I'm pretty sure Gamers Nexus enable these limits and disable any automatic overclocking and HUB said a similar thing in their recent podcast, so their benchmark results shouldn't be much different.

I'm not sure that the Asus profile actually fixes the problem though, because there seem to be two issues: one is the CPU exceeding PL2 and that causing instability due to exceeding what the cooling, CPU or motherboard are capable of and some CPUs seem to be unstable even at their max boost clocks and I don't believe that's related to their power consumption as it can happen even under single core load.

 

[humbug]

PL2 is not what they put on the box, and without MCE it lasts just long enough to complete a Cinebench run. but not long enough to matter a _____ for anything useful.

The reason i go on about all this is because i find it all quite manipulative and cynical, and Steve Burke complaining over and over and every time he finds an opportunity to accuse AMD of "lying" about power consumption for going 5 or 10 watts over while explaining Intel do nothing wrong is systematic of a bias that exists simply because its good click bait and Intel at least understand the value of sucking up to people like him, to stroke their ego.

 

[Tetras]

and without MCE it lasts just long enough to complete a Cinebench run. but not long enough to matter a _____ for anything useful.

I believe MCE only allows the cores to boost to the max single core turbo multiplier, I don't think it has any impact on the length of the boost itself.

I'm pretty sure TAU has been infinite for K CPUs since 12th gen *link, but even if it isn't, I'd expect someone like Asus to be setting it as infinite anyway.

For users who understand the former PL1/PL2 methodology, it still technically exists under the hood here, where Base is PL1 and Turbo is PL2, but Tau is effectively infinite for K processors.

Unlimited PL2 and infinite TAU = good way to cook your VRM

Since a lot of the buyers of these CPUs will be getting them for workstation apps, you could argue it is irresponsible.

I don't know what Steve has been saying, but reviewers are definitely right to complain about the motherboard manufacturers, though as you say, Intel should be insisting that they set these boards to 100% stock settings by default, so they're ultimately the party responsible.

The reason i go on about all this is because i find it all quite manipulative and cynical, and Steve Burke complaining over and over and every time he finds an opportunity to accuse AMD of "lying" about power consumption for going 5 or 10 watts over while explaining Intel do nothing wrong is systematic of a bias that exists simply because its good click bait and Intel at least understand the value of sucking up to people like him, to stroke their ego.

For me, Intel's TDP means one thing: how crap a cooler can they fit to an OEM PC and get acceptable performance for pootling around in Word or Chrome. Hence, a 13900 has a laughable 65 watt TDP. I'd remove TDP entirely and put base power (PL1) and turbo power (PL2) on the box.

 

[humbug]

I think given that they are advertised as 125 watt CPU's they should be reviewed under 125 watt coolers.

The hilarious thing is Steve Walton (Hardware Unboxed) actually did something along those lines for AMD, back when AMD shipped all of their CPU's with Box Coolers Steve decided they should be reviewed under those coolers, it had no real performance penalty for them but for whatever reason AMD stopped that.... its quite sad as those coolers were nice little coolers, people actually look for them on the flea market to put in small builds.

Honestly IMO if you're going to advertise something as such then it should be tested as such, they soon stop that nonsense.

 

[Tetras]

Honestly IMO if you're going to advertise something as such then it should be tested as such, they soon stop that nonsense.

It is hard to know exactly who to blame for this (like where in the pipeline the problem is), like if you look @ Intel's slides when they released new CPUs they do have a table for base and turbo power and that is how they're listed on Intel Ark, but yet you go to a retailer and see TDP: 65 (or whatever) next to the name.

I think given that they are advertised as 125 watt CPU's they should be reviewed under 125 watt coolers.

It made more sense when TAU was not unlimited, since the idea was that the limited turbo was within the cooler capacity, but since TAU is basically unlimited now, even on this forum we see people buy a 14900K and assume the kind of coolers they used on their 6700K is going to work

 

[GoogalyMoogaly]

It is tricky to know what review sites should do as running at 250W or whatever is probably default behaviour for a lot of motherboards, so should you be tweaking that for a review or using it as-is to give a better representation of what a user would see by default?
You wouldn't expect an AMD CPU to be reviewed with PBO turned on as if it were the default case and you wouldn't expect a 4090 to be overclocked and tweaked to within an inch of its life for a review. At least not the standard review, maybe as an extra result you could do all those things.

You'd expect them to be reviewed as they'd run out of the box wouldn't you? Or would you? If RAM is rated at 6000MHz but you need to enable XMP or EXPO in order for it to run at that otherwise it defaults to 4800MHz or whatever you'd expect the RAM to be reviewed at 6000MHz wouldn't you, even if it means changed motherboard settings?

So I don't know whose fault I'd say this is, maybe a little of column A and a little of column B (i.e. a bit Intel and a bit the motherboard manufacturers). Putting in another profile to cover yourself seems sensible, but would you make it the default? That wouldn't make Intel or the motherboard manufacturer look good would it?

 

[humbug]

You make a good point on the memory but XMP / EXPO is such a common thing now IMO running it is not a problem, so long as its reasonable, for example with early AM5 while they could do 6000 it was not reasonable to expect that, they wouldn't always do it and even if it did it needed more work that just setting that in EXPO, 5600 was reasonable and that is what people like Hardware Unboxed used, some accused them of being bias because they would run Intel at 6400, but with Intel 6400 was a reasonable expectation, so its reasonable to do that.

These days AM5 can do 7000 even 8000 with some getting "its over 9000", AMD have put a lot of work in to that and its paid off, so i think its reasonable to review at 6000, 6400, maybe even 7000, i think HUB now do it at 6400.

 

[Proto]

The ability to run constant PL2 is down to the motherboard quality and cooling. The consumer should obviously be using their own intellect along with the relatively large amounts of their money.

Although Intel should provide general guidance, they haven’t necessarily done anything wrong, as that guidance would have been to the motherboard manufacturers.

It’s going beyond PL2 where the problem exists, and perhaps the way reviewers continue to manipulate consumers.

 

[VinylRichTea]

I know where people are coming from, as I used to have a problem with this; however, it does help to learn the proper way to overclock and undervolt plus I have KS level 13900K (SP104 - P-Cores 113, E-Cores 88, on the VF Point Offset it has 5.8Ghz at 1.388v) as I got lucky with the bin, and it was the only one I bought. However, a contact frame, a Corsair link 420mm AIO, a delid and a copper IHS really made it shine. Cinebench R23 runs go from 39400-39700 and top out at 80c and that's on a power limit of 260w (upped it from 253w) on LLC4 with a bit of fine tuning on IA AC LL and IA DC LL. Here's my Geekbench 6 run, which smashed a lot of 14900ks multicore runs: GB6 Run

 

[Grim5]

If my understanding from AHOC's video is correct, the motherboard vendors are undervolting the CPUs too far, because Intel lets them screw with the voltages and doesn't validate them

Intel gives them a software tool which they are supposed to use for benchmarking and stress testing to validate their voltage curve profiles, but AHOC thinks either they are not using the tool at all, ignoring the tool's results or the tool simply sucks and is not a good stress test for the cpu

 

[peahead]

Gigabyte have also just released a bios update today, i guess all of them will at some point.

 

[Tee Hee Johnson]

Could be waiting a while for an ASRock bios!

 

[ATIorNvidia]

Wonder if we'll get re-reviews of 14th gen locked to 253w

 

[Grim5]

Intel has released a statement and it suggests that out of the box pre overclocking for Intel motherboards is over

"Intel has observed that this issue appears to be related to out-of-specification operating conditions.
Analysis of affected processors shows some parts experience shifts in operating voltages outside of Intel specified operating conditions.

• We are still investigating, but Intel has observed the majority of reports of this issue are from users with unlocked/overclock capable motherboards.
• Intel has observed 600/700 Series chipset boards often have BIOS defaults set by the motherboard manufacturer to disable thermal and power delivery safeguards designed to limit processor exposure to sustained periods of high voltage and frequency, for example:
  – Disabling Current Excursion Protection (CEP)
  – Enabling the IccMax Unlimited bit
  – Disabling Thermal Velocity Boost (TVB) operating conditions.
  – Additional settings observed which may increase the risk of system instability:
  – Disabling C-states
  – Increasing PL1 and PL2 beyond Intel recommended limits

Intel requests that system and motherboard manufacturers provide end users with a default BIOS profile that matches Intel's recommended settings.

• Intel strongly recommends customer’s default BIOS settings from the manufacturer should ensure operation within Intel’s specified recommended settings.
• In addition, Intel strongly recommends motherboard manufacturers to implement warnings for end users alerting them when there is any unlocked or overclocking feature usage.

 

[ltron]

If my understanding from AHOC's video is correct, the motherboard vendors are undervolting the CPUs too far, because Intel lets them screw with the voltages and doesn't validate them

Intel gives them a software tool which they are supposed to use for benchmarking and stress testing to validate their voltage curve profiles, but AHOC thinks either they are not using the tool at all, ignoring the tool's results or the tool simply sucks and is not a good stress test for the cpu

How does that work when our CPUs will be of different silicon quality? At that point they are just guessing.

In any case I strongly believe stock should be fully stock. If the motherboard manufacturers want to put overclocking and/or undervolting profiles in the BIOS then they should be clearly labelled as such.

 

[Emlyn_Dewar]

Asus. Not even once.

 

[CaNsA]

Asus. Not even once.

Interesting take.
My last few boards have been Asus and I've not had any trouble.

 

[ltron]

Interesting take.
My last few boards have been Asus and I've not had any trouble.

I've had loads of trouble with them to the point where I try to avoid them; Gigabyte by contrast has been the most reliable.

However, I accept this is anecdotal and is not necessarily indicative of the wider trend.

 

[CaNsA]

I've had loads of trouble with them to the point where I try to avoid them; Gigabyte by contrast has been the most reliable.

I had nothing but trouble with my Gigabyte kit.

However, I accept this is anecdotal and is not necessarily indicative of the wider trend.

Agreed