• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Instabilty with Intel 13th/14th flagship cpus in some games (unreal engine+)

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
As the titles states, I was having a weird crashing issue on my pc (13900k/4090) in fortnite. I could play a round, maybe two and hard CTD. It happens often enough I won't really play it (past month or so). Apparently this effects other titles, and right this second, I domt recall if it was a specific engine or what.

To that end, Intel says they are looking into it... but in the meantime mobo vendors including asus and msi with Gigabyte have added a feature to use the Intel stock specs (power limits) which seems to motigate the issue. Here's one article...



Edit: what Intel said...
 
Last edited:
Looks like Intel's "freedom" is working against them. I kinda disliked AMD's enforced stock power limits but now it looks like mobo manufacturers can't be relied upon to use them sensibly. I don't know if Intel's policy changed in recent gens but previously adjusting power limit by itself was not considered overclocking. Directly adjusting voltage or clock ratios was, so I disliked MCE and similar.
 
Is it intel or the liberties motherboard partners are taking with their BIOS'? I guess intel blesses it so... intel.

But yeah, a lot of unreal engine games, and others.

I'm going to try the workaround tonight as it's an option on my asus board with the latest bios.
 
We're missing one piece of info for me to point the blame. Based on my past understanding Intel allowed unlimited power settings as not overclocking. Thus the CPU should be stable at any power setting if you don't do anything else considered overclocking by Intel. If the mobo manufacturer does anything with voltage or clock ratios, that's overclocking and 100% the mobo manufacturer's fault. The grey zone here is I don't know if Intel changed their policies with recent gens, as they now express power differently than in the past.

The question now is, does Intel still guarantee stability if you only adjust power limit, including unlimited? Or do they only guarantee up to "max turbo power" I think it is called now.

Edit: forgot to explicitly say it, if the CPU is unstable while keeping in power limit without overclocking, that's clearly Intel's problem.
 
Funny thing is, it's DEFAULT behavior (increased PLs) on most (overwhelming majority I'd say) boards. There's no adjusting to it...it just is.

Edit: added intels response to first post...
 
Last edited:
If I remember correctly, it's about power spikes in the "performance" mode, which is generally a higher TDP mode (253W+). The problem was mainly related (and started from it) to unlocked power limits (mainly max current), causing stability issues. It was described more broadly on some other websites in the last few months.

Intel guarantees the same max frequency, but power states have more restrictions, and the time during which max frequencies are available is shorter. So it meets the max frequency specs, but the average performance during longer, higher loads is worse.

I only add that I have no problems at all with the i9 13900HX, but it's a mobile CPU. My gaming PCs or those that run 24/7 at home, use only Ryzen CPUs or mobile Intels. I'm using 14700K/14900K only for benchmarks. Even though I haven't seen stability issues on my test rigs, it's hard to say if they're really 100% stable in everything. Besides popular benchmarks, they collect dust.
 
Funny thing is, it's DEFAULT behavior (increased PLs) on most (overwhelming majority I'd say) boards. There's no adjusting to it...it just is.

Edit: added intels response to first post...
You probably see more/different boards than I do. In my experience with older boards, as a generalisation I find non-OC boards (e.g. B chipset) tend to set PL1 and PL2 to Intel defaults, although the mobo manufacturer can offer a higher performance mode with those limits raised e.g. Asus APE. This is not overclocking and I think this is ok.

With overclocking (Z chipset) boards it is less clear. Do they default to Intel limits or go straight to unlimited? This is before looking at things like MCE. On my X299 board if XMP is off it sets Intel values, but XMP on, MCE off, it sets unlimited power limit. The user can set power limits as desired. The only question is what are defaults.

That Intel statement, feels more or less in line with what I expect. Basically Intel get mobo makers to behave more like they do on AMD boards. Out of box power limits, overclocking warnings. I guess they were happy turning a blind eye to it for the higher reviewed performance, but they'll have to move on from there.

Also I guess I was looking at it over simply. It's not just a singular power limit, but various other limiters and settings which also factor in.
 
Sorry, I should have been more clear...overwhelming majority of z790 boards. I dont think about the "B" chipsets when we're talking flagship-class unlocked processors.

As far as who sets what, it varies by vendor and sku. Asrock loves 4096/4096 (unlimted) more for than others, though it's not up and down the stack iirc. It just varies wildly. And like woomack said/you acknowledged, theres a bit more to it.

But yeah these are newer issues and not talking any mobo overclocking like the old mce.

Edit: to be clear, I'm not saying unlimited is the overwhelming majority, but raised limits/durations/amps certainly are.
 
Last edited:
"Intel CPUs Are Crashing & It's Intel's Fault: Intel Baseline Profile Benchmark"

00:00 - Welcome to Hardware Unboxed
01:26 - Ad-Spot
02:25 - The problem
06:32 - Enabling Intel Baseline Profile
08:14 - Cinebench 2024
09:46 - Baldur's Gate 3
10:26 - Cyberpunk 2077: Phantom Liberty
11:04 - Hogwarts Legacy
11:32 - The Last of Us Part I
11:49 - Star Wars Jedi: Survivor
12:15 - Assetto Corsa Competizione
12:29 - Spider-Man Remastered
12:39 - Watch Dogs: Legion
12:48 - Starfield
13:05 - Performance Discussion
14:48 - Gigabyte Press Release
18:43 - Intel Admit Unlimited Power is “In Spec”
22:05 - Final Thoughts


01.jpg

02.jpg
 
"We are going to be very crisp in our definition"

I like that.

AMD does something similar with their unlocked chips in the way you can configure boosting, and power limits. That is why I used to say a stock 5900X is weak, you have to wake it up with more power, more boost, all of which AMD says is ok to use. And I have tested it myself at its absolute thermal and power limits.

On Intel they do all the work for you lol :D
 
That Anandtech article is the basis of my earlier question: have Intel more recently changed their policy that only adjusting power limit is NOT overclocking? Be careful that article is from 2019. Coffee Lake was the current gen desktop at the time and Comet Lake was released later that year. https://www.anandtech.com/show/1458...ng-an-interview-with-intel-fellow-guy-therien

However that might not be the complete picture. As Woomack mentioned in an earlier post, there are other limiters and settings that mobo manufactures can also tinker with. Is it the power limit, or it is other things that are being done at the same time? Again we don't have the info. Maybe the power limit was never the problem but all the other stuff around it. It needs testing and I feel calling out Intel in that way is premature at best, if not irresponsible.
 
It needs testing and I feel calling out Intel in that way is premature at best, if not irresponsible.
Funny, I thought the same but reversed when everyone immediately started blaming motherboard manufacturers instead of Intel? I agree with he's logic, just because a motherboard can make a CPU run full tilt for longer doesn't mean it's out of spec, if the speed it reaches is what it's supposed to reach. If it's pulling more power than it should and becoming unstable even at "stock" values simply because the motherboard is allowing it to pull more power, then it's clearly Intel's fault no?
 
The mobo isn't only adjusting the power limit, but some other things as well. Adjusting power limit by itself might be allowed by old Intel rules. I don't know if that applies to the other things. Look at MCE, which has been an overclock on Asus mobos going back possibly a decade if not more. even on my new AM5 build the Asus bios makes it confusing what is or isn't an overclock.
 
So....yeaaaaaaaaaah...

I updated my BIOS to support the new 'functionality' that came from this. For some reason I was thinking, because you supposedly lose performance, that temps would be lower, power, etc. That didn't seem to happen. Voltage, according Coretemp and CPUz, was way higher... it locked the P cores to 5.5 GHz, E to 4.3. Then, when gaming (Fornite) it just thermally throttled banging off 100C easily.

The good news is that Fornite stopped crashing. But there's no way I'm running my PC like this 24/7 (I may play Fortnite once a week if PUBG is down for maintenance).

Did anyone else with the affected chips/paltfoems update their BIOS and try this?
 
It could be the change to SVID setting. I'm not familiar with it myself but I have seen reports that the Intel Fail Safe setting can increase voltage compared to auto, with the obvious consequences of doing that.
 
It could be the change to SVID setting. I'm not familiar with it myself but I have seen reports that the Intel Fail Safe setting can increase voltage compared to auto, with the obvious consequences of doing that.
It's quite obvious the SVID change did that. I did NOT expect it to behave this way (increased temps, voltage, etc.) for w/e reason. It doesn't throttle down (THAT is Intel spec?), even.

You'd think the baseline intel specs would use less power than what the motherboard liberally sets by default, considering the description of what's happening.
 
I hesitate to write much since I'm only going off 2nd hand information and not personal experience. Intel Fail Safe is just that, meant to get your system running stable. To my understanding that is the Intel recommended voltage setting. Whatever the mobo does on Auto could be likened to undervolting to try and get more performance out. I think it was stated for AMD, but I'm sure the same applies for Intel, that stock voltages are set so that worst case production silicon will work. Better ones could work with less voltage, but they're not going to bin every individual device to that level.

I guess it is an evolution of sorts like we had to go through with overclocking. Go back say 10 years, what did we do? Increase the ratios and voltage. In the modern era it seems that increasing various limits while reducing voltage is the new meta to squeezing more perf out.
 
Right, I get all of that too...I just didn't think through what, exactly, the motherboard 'profile' (liberties) handles.

Logic told me that if it's, in part, because of 'unlimited' power limits causing the instability in the engine, they will limit power somehow. So, as mentioned earlier, there's a lot more going on than just PL adjustments as we're using more power. I saw voltage and clock speed behavior change; an increase in the former is what surprised me - regardless of how AMD overclocks and how you can get more out of Intel...
 
So if this is an Intel 13th and 14th gen problem... then how could they have gone an entire generation and a half without anyone noticing until now?

I mean there are a lotttttttt of Fortnite players out there...

Surely they would've raised enough hell to make Intel cry BEFORE the 14th gen, no?
 
Back