Power limits, Intel Default Settings and their role in the problem of degrading Intel processors
Last week, Intel announced discovery of another source exposing desktop Raptor Lake processors (Core 13th and 14th generation) to high voltage that leads to irreversible damage and instability. But there’s one more thing: The company says that to protect your CPU from damage, you should use the so-called Intel Default Settings that restrict power limits. So is the CPU in danger without them, and does disabling them risk degrading?
For context, a little recap: while AMD CPUs work by having a defined maximum power consumption that the CPU will not exceed, Intel’s specs are very loose. CPUs have power limits, the most important of which is the PL2 limit, which tells what the real maximum power consumption during boost is or should be (while Intel TDP values are a completely irrelevant today). This limit for desktop processors like the Core i9-13900K and i9-14900K is 253W.
That’s a lot, but it’s not enough for Intel, and the important thing is that this power limit is only a recommendation. Motherboard manufacturers have been increasing the limits or setting them to de facto infinite values (4096 W) in reality, to make processors perform better in multi-threaded benchmarks. That’s where the infamous extreme measurements in reviews, where, for example, CPUs consumed 300–400 W in Prime94, are coming from. In CPU reviews, however, the use of these boards skewed performance upwards everywhere, whereas if you set the CPU to “recommended parameters”, it would perform worse. For this reason, it seems very likely that Intel was more than comfortable with this practice, if it was not actually the CPU maker that quietly encouraged the manufacturers to do it.
This practice lasted until this year, when reports of unstable processors began to appear widely, and it was later confirmed that the problem is extremely serious, because instability is merely a manifestation of physical degradation and damage to the chips (basically gradual failure, or, if you prefer, accelerated aging), which until then had been quietly building up in people’s processsors. And this issue only manifests itself after some time, so the cases where users already had their CPUs unstable were just the tip of the iceberg, so to speak.
Intel has been gradually coming up with several measures to address or mitigate the problem, the first of which was to crack down on exactly those relaxed power limits that were increased above the recommended values. The company has never admitted that it was at fault for tolerating or allowing non-compliance, but it has negotiated (or dictated) with board manufacturers that from about May 2024 onwards, BIOSes (UEFI firmware) of boards will use the so-called “Intel Default Settings”, which meant return to the original recommended settings completely in line with the specifications. It also means that with these settings, processors have lower performance (especially in multi-threaded tasks) than they originally displayed in launch reviews (which were usually measured with boards that had the limits raised). We wrote more about Intel Default Settings in this article.
However, Intel then revealed that the real main cause of the spontaneous degradation of Raptor Lake CPUs was various deficiencies in their power regulation and CPU voltage management, which led to the CPU receiving dangerously high voltage fluctuations in the Vcore that gradually wore the silicon down, eventually making it unable to function correctly (Intel calls this Vmin shift, i.e. the CPU needs higher and higher voltages for achieving stability compared to the original out-of-the-factorystate).
Do you have to use Intel Default Settings forever with Raptor Lake, or was it just a temporary measure?
Since this high voltage is not directly related to the issue of high power consumption due to relaxed limits, it was unclear whether Intel Default Settings were still needed (i.e., whether this was just a temporary precaution at the time, intended as only a provisional measure until the true cause was discovered). But last week, Intel again included Default Settings among the four measures that are stated as a solution to protect a processor from the degrading and damage. This would imply that, after all, excessive unlocked power consumption can cause damage. Fortunately, Intel has now clarified some of these ambiguities. This is reported by Tom’s Hardware, who received the clarification from Intel.
Intel has confirmed that it is “recommended” for users to keep using Intel Default Settings. And at the same time, setting Intel Default Settings is one of the elements that, together with updating the microcode to version 0x12B or later, solves the cause of the whole chip degradation issue (but will not help already damaged processors – their degradation is irreversible, as Intel directly confirmed, the only solution is an RMA replacement or refund).
There is, however, one nuance. Users can change some of the sub-limits after selecting Intel Default Settings to get some or all of the lost performance back. But from Intel’s point of view, the only risk-free option is to change the power limits – so basically mainly that PL2 limit (PL1 too). So, you can set that to 300 W and, according to Intel, you won’t be putting the CPU at risk. Changing merely these “wattage” limits is still within the scope of usagee covered by the warranty (which is not invalidated by increasing the PL2 limit), according to Intel’s statement to Tom’s Hardware.
You can unlock PL2 safely, but the rest is hazardous
The Intel Default Settings did not only deal with PL2 and power limits, but also with other limits, such as current limits. In fact, before Intel started enforcing the recommended parameters again this May, the boards were relaxing a lot of other things along with power consumption – for example the current limit (IccMax), but also various VRM settings (Current Excursion Prevention). And changing these other limits (or protections) can already be dangerous. While you can safely change the PL2 and wattage limits according to Intel, you can’t change these other settings while staying safe. Changing them is already considered overclocking, and Intel mentions that they can potentially lead to voiding the warranty – unlocking these limits is no longer covered by the warranty. Of these potentially dangerous out-of-warranty settings, for example, increasing the IccMax current limit is explicitly mentioned as being one of them.
Other things that Intel Default Settings sets and you’d better not change are TjMax Offset (you probably shouldn’t increase the CPU’s temperature limit since it is in place to protect the CPU), IccMax_App, CEP, Thermal Velocity Boost settings, and even C-States (which should be kept enabled according to Intel).
It is not explicitly confirmed that increased IccMax (or TjMax Offset) is known to increase the risk of processor degradation, but it is probably a good idea to listen to Intel here. If you don’t want to go out of your way to play with manual overclocking, then you better not touch anything except the PL2. While we don’t think Intel would refuse to honor the warranty on a CPU that is demonstrably affected by that degradation anyway, even if the CPU had a record of having been overclocked (the CPU can get flagged as being overclocked by blowing an integrated eFuse), it’s probably not worth the risk.
It would be different for processors that do not have known problems with increased failure risk. But if you own a Raptor Lake and you have been using for a while before the patches were available, the processor probably already has some damage accumulated in it and may fail in the future due to it. So it’s preferable that you have the option of RMAing it in the future (by the way, you should absolutely make sure to keep the invoice for the same reason). Especially since Intel has gone out of its way to accommodate RMAs and extended the warranty to five years.
Read more: Intel extends warranty on problematic Raptor Lake processors
Specifications being non-binding was a mistake from the start
This otherwise means that if you switch the BIOS settings of the motherboard from the Intel Default Settings to some “Performance” or “Xtreme” (or otherwise named) preset from the manufacturer, which used to be often set in the default state before, you will actually void the warranty, in theory. This is because the CPU will notice that this profile has raised IccMax, for example, and will set the overclocked flag.
To be more precise, your CPU may already have this flag set from long ago, if you ever used such a motherboard (especially Z690, Z790 models) when it had such an overclocked settinga as default. This again shows how crazy the state of affairs on the Intel platform actually was before the company (very belatedly) started enforcing its “recommended settings” as the default this year. What to say about a situation where virtually all the better boards had some form of overclocking enabled by default, which strictly speaking violated the CPU warranty virtually the moment they was first run in the board, before the user even had a chance to change such warranty-voiding settings in the BIOS? Ideally in the future Intel should do it like AMD and just make the specs mandatory and by-spec behaviour to mandatorily be the default setting, explicitly keeping all alterations as an user-selectable actions.
That being said, we do not expect Intel will use this against the customer to deny RMA requests. You probably won’t be denied even with a processor that has a record of overclocking. Such thing could probably happen in isolated cases where you are unlucky enough to happen on an overzealous or uninformed support employee, which would probably be rare. But it wouldn’t be fun to deal with the situation if you’d caught such bad luck.
Source: Tom’s Hardware
English translation and edit by Jozef Dudáš
⠀
- Contents
- Power limits, Intel Default Settings and their role in the problem of degrading Intel processors
