Thursday, February 22nd 2024
Intel Core i9-13900K and i7-13700K Gaming Stability Issues Linked to Power Limit Unlocks
Users of Intel's 13th Gen unlocked K-series processors such as Core i9-13900K, i7-13700K, are reporting stability issues when gaming even at stock clock-speeds. Hassan Mujtaba of Wccftech and Tom's Hardware have isolated the issues to power limit unlocks. Most Z690 and Z790 chipset motherboards include BIOS-level unlocks for the power limits, particularly the Maximum Turbo Power (interchangeable with PL2). By default, the i9-13900K and i7-13700K come with a PL2 value of 253 W, but you can get the motherboard to unlock this to unlimited, which basically tells the processor that it has 4096 W of power on tap, so not technically a "stock" configuration anymore.
Of course, neither your PSU nor your CPU VRM are capable of delivering 4096 W, and so the processor tends to draw as much power as it needs, to maintain the best possible P-core boost frequencies, before running into thermal limits. At stock frequencies with stock boost bins, unlocked power limits can drive the power draw of i9-13900K as far high as 373 W under a multithreaded load, in our testing, when compared to 283 W with the power limits in place. It turns out, that unlocking the power limits can come with long-term costs, besides the literal cost of electricity—the processor's stability with gaming workloads can degrade with certain hardware combos and settings.
Wccftech's Hassan Mujtaba has a novel fix for this—to undervolt the processor. In his experience, undervolting the processor and restoring the stock power limits in the motherboard BIOS settings restored gaming stability. Meanwhile, Tom's Hardware suggests a slightly different approach besides restoring power limits—to reduce the P-core boost multiplier by 2.0x (i.e. reducing the maximum boost frequency by 200 MHz). Both these approaches are claimed to restore gaming performance stability for the i9-13900K and i7-13700K. Although there are no confirmed sightings of the issue in the wild for the newer 14th Gen chips, it stands to reason that even the 14th Gen i9-14900K and i7-14700K could be affected by this issue. We still don't know why it doesn't affect the 12th Gen chips, since they feature fundamentally the same power design as the 13th- and 14th Gen chips.
Sources:
Hassan Mujtaba (Twitter), Tom's Hardware
Of course, neither your PSU nor your CPU VRM are capable of delivering 4096 W, and so the processor tends to draw as much power as it needs, to maintain the best possible P-core boost frequencies, before running into thermal limits. At stock frequencies with stock boost bins, unlocked power limits can drive the power draw of i9-13900K as far high as 373 W under a multithreaded load, in our testing, when compared to 283 W with the power limits in place. It turns out, that unlocking the power limits can come with long-term costs, besides the literal cost of electricity—the processor's stability with gaming workloads can degrade with certain hardware combos and settings.
73 Comments on Intel Core i9-13900K and i7-13700K Gaming Stability Issues Linked to Power Limit Unlocks
It just clocks all my 8 P-Cores to 5Ghz
I'm not imaging Intel are saints here, but most likely this is more of a cat-and-mouse game, where motherboard manufacturers search or demand features they know they can (ab)use to gain an edge over the competition. You don't like that? No worries, you can always jump ship to AMD.
I think the real story here is that tech writers recycle old stories by using sensation headlines that sound current.
They present everything in the most sensational style possible, often making intentional misrepresentations to drive engagement.
The story at Tom's is based on a tweet from a guy that(despite his position) doesn't know how to get his system stable, and a thread in the Steam forums.
Now it's a fake story here. I've seen too many 13700K doing fine to buy into this.
I think the problem is too many people watch a video and decide to build a PC, but they don't really know what they are doing.
Gamers Nexus explained "Accidental Pre-Overclocking vs. Spec"at the time of 10th gen. That guy laughed painfully as he noted that he has to explain this "every single CPU launch from Intel". Asus boards have been heading this direction for like 20 years(going above specs for better performance), and IMO you're in over your head if your don't already know this.
From what I've seen around, Intel is more fun for tinkerers, the whole manually adjusting voltage, power limits seems to scratch the hitch of some people. I know a mad man who put a heavily tweaked 13900k into a 8l box, cooled with a is60evo with only a bottom fan. It's still faster than a stock 13700k. Another went in the opposite direction and push his 13900k to the limits for daily driving and use a Mo-RA 420. I wouldn't do that, but there's a community of people who love that shit and are bored by "plug and play" CPUs.
I'm not saying that reviewers are wrong, just that there's a gaping contrast between two realities here.
This seems like such jet lag on the problem and concern being raised on this issue for these past generation Intel chip parts though. It's unsurprising that issues like this exist though with the way MB makers aren't setting default bios setting to adhere to recommended safety guidance's of AMD/Intel in the first place.
As far as microcode update pushed through Windows by Intel or AMD likewise that's really a separate topic of discussion. I don't see how the two are directly relatable to discussion.
You said it yourself, H chipset getting locked was predictable. And yhea...that was directly going against Intel product segmentation. Yes, that was 100% motivated by money, but AiB might have gotten in trouble with Intel if they didn't comply. You bought a product with a feature that wasn't supposed to be allowed. The AiB were in the wrong from the moment that they tried to be "good guys". In the same spirit, Intel also blocked the workaround (BCLK OC) that MSI found for 13th gen non K/B chipset.
The issue is a bit different here, since it doesn't seem that people were aware that this could be a problem until recently, and that's an issue that also concerns the Z chipsets who are supposed to let you do what you want. Gimping the Z chipset without a valid reason would have been an order of magnitude worse in public reception than not allowing the H and B chipset to do things that it wasn't supposed to do at the time.
As long as you are not trying to enable OC on a chipset/CPU that's not supposed to do it, Intel give AiB a fair amount of wiggle room on the unlocked chipset, like how they let them decide if they want to block or allow undervolting protection (which basically makes undervolting useless if enabled).I'm not so black and white about the explicit consent stuff, though. It depends on why said functionality is being removed/modified. Since not every user is capable of making a sound decision about it, especially if it's about system stability. The chipset stuff seems to be justified on a legal pov. Intel sued nvidia for the chipset used in the 2008 unibody macbook, since they were not in fact allowed to make chipsets that could work with their CPU. So there might be something legit about Intel blocking OC on chipsets that were not in fact supposed to allow OC. Does it suck for the customer ? Yes, but that's not their problem.
Look, Asus took their precaution from the start, and said that this feature is not guaranteed to work in the future
People think CPUs have temperature/clock/power limits for safety reasons but that's not the only thing, they have to regulate the temperature because the properties of the silicon change and these things are designed to operate correctly only within certain parameters. Processors start to spit out incorrect results long before they're actually "damaged" by the heat.
Temperature is also a proxy for power draw, if two pieces of silicon reach the same temperature limit in the same time but one consumes twice as much power that means the control system for the one that consumes more power has to respond faster and is probably more prone to failure.
So in short its only really news if the CPUs are crashing when you at spec'd power limits 'and' spec'd clocks 'and' spec'd voltages.
To the wccftech dude many of us already undervolt, you late to the game dude.Not sure about unlocked to 4096w, but I am pretty sure my ASRock does a pl1=pl2 by default, is that spec on these new chips?
This page doesnt say e.g.
www.intel.com/content/www/us/en/products/sku/230500/intel-core-i713700k-processor-30m-cache-up-to-5-40-ghz/specifications.html