Overclock.net banner

Skylake X Phantom Throttling

22K views 31 replies 16 participants last post by  coccosoids 
#1 ·
If anybody has been following the other Skylake X threads, you'll notice that there's been some talk about the "phantom throttling" on Skylake X. For the past week, there's been at least several people (including myself) that have been investigating the throttling issue. And I'd like to gather all the information in one place.

What is "Phantom Throttling"?

Phantom throttling is when the processor throttles the performance without a visible change in clock frequency. Overclocker der8auer best explains this here (though he doesn't actually use the term "phantom throttling").

What this means is that your processor can throttle in a way that is undetectable in CPUz because the clock speed stays the same - hence "phantom".

But even through the clock speed doesn't change, there are multiple (indirect) ways to detect it:
  • A decrease in performance.
  • A drop in temperatures.
  • A drop in power consumption.
The problem with phantom throttling is that it is hard to detect unless you're specifically looking for it. So if your a 7900X passes a stress-test at 5.0 GHz on air, you may think the chip is great. But in reality, it's probably phantom throttling and therefore the stress-test is invalid. Furthermore, when the phantom throttling kicks in, the performance drops off much more than normal (clock-speed) throttling.

What does it look like?

As mentioned, it is a drop in performance/temperature/power-consumption without a drop in clock frequency. In effect, it's a drop in IPC (instructions per cycle).

Here's a table of benchmarks showing the 7900X phantom throttling with the Gigabyte AORUS Gaming 7 (BIOS F7a) at stock settings under an AVX512 load.

(Source: http://www.numberworld.org/y-cruncher/news.html#2017_7_6 - There are also screenshots for all the benchmarks with detailed monitoring info in the source.)

Most of the benchmarks show perfect performance scaling from AVX2 -> AVX512 as well as increase in thread count. However, the AVX512 benchmarks start throttling at 10 threads/10 cores. The frequency is still 4.0 GHz as read by CPUz, HWMonitor, and CoreTemp. But the performance, temperature, and power draw drop off drastically. The performance loss is so severe that AVX512 becomes slower than than the AVX2!

What causes phantom throttling? (Updated: 9/24/2017)

There seems to be multiple causes of the throttling - some of which are more "phantom" than others.

Voltage droop on the CPU input voltage (VCCIN/VRIN) under load:

The stock VCCIN is 1.8v. But under a large load, it may drop to as low 1.60 - 1.65v. If it drops too far, the throttling kicks in. The exact cut-off is unknown, but it seems to be around the range of 1.68 - 1.70v.

Furthermore, the throttling isn't always a sudden cliff. And you may see the VCCIN go back above the thresholds. My suspicion is that there may be a feedback loop where the throttling (with the decreased CPU power draw) allows the VCCIN to go back above the threshold as which the throttle stops and the power-draw increases causing the VCCIN to droop back down - thus repeating the cycle.

So depending on where your hardware monitor probes the VCCIN during this feedback loop, it may catch it anywhere in the range of 1.65 - 1.80v. This makes it possible to read 1.75v+ even though you're clearly throttling.

This VCCIN throttle is a phantom throttle. Based on my conversations with an industry insider, the cause seems to be a frequency divider on the core that kicks in when certain triggers are hit. The reason why it is "phantom" is because the vast majority of (all?) the hardware monitors available (including CPUz) are not able to read this frequency divider. And therefore, they cannot correctly compute the actual CPU frequency.

In other words, the throttling is an actual drop in CPU frequency that's undetectable. There are no forms of architectural or IPC throttles as had been hypothesized before.

Thermal Limits:

There are numerous other limits such as current, wattage, etc... that may cause the CPU the throttle. These may or may not be phantom.

(Outdated) Original Answer:
So far multiple causes have been identified. And they vary by motherboard. One that seems to be common across all motherboards is thermal limitations. But the details of are still fuzzy.
In my personal emails with Silicon Lottery, he says that pretty much all the X299 motherboards have some sort of phantom throttling. But so far most of the information is about ASUS and Gigabyte. Silicon Lottery also says that ASUS boards are the least affected. But this was before der8auer found that the VCCIN vdroop was the problem with the Gigabyte boards.

As of right now, the exact mechanism of the phantom throttling is unknown. Is the processor turning off execution units? Is it lying about the frequency? We currently do not know.

Assuming phantom throttling is intentionally designed by Intel, we also don't know why they would choose a phantom throttle over a normal throttle. Normal throttling (by lowering clock speed) has the advantage of allowing the vcore to be dropped to further cut back on thermals.

How do I fix it? (Updated: 9/24/2017)

Voltage Droop on VCCIN/RIN:

The fix here is simply to keep the VCCIN/VRIN from dropping below the threshold of 1.68 - 1.70v.

There are numerous ways to do this depending on the motherboard and BIOS. The obvious way is to simply increase it above 1.80v so that when it drops, it still stays above the threshold. Alternatively, you can use load-line calibration.

Thermal Limits:

These vary by motherboard. But there seem to be two limits that need to be increased:
  • Current Limit
  • Current Protection Limits
If the current limits are hit, the processor will throttle. If the current protection limits are hit, the system will simply shutdown. See the second post for the known motherboard-specific solutions.

Going back to the benchmark above, here's what it looks like without the throttling:



So now instead of phantom throttling, there is a bit of temperature throttling as it hits the TjMax of 95C. The temperature throttling is only slight and is visible as drops in clock frequency. Overall the performance is much better.

Fun Fact: The 7900X can pull 300W at only 4.0 GHz at stock voltages while running AVX512!

Questions or comments?

Disclaimer: I'm the author of the y-cruncher benchmark that's been referenced above.

Here's a dump of other places where I or someone else has mentioned the phantom throttling:
 
See less See more
2
#2 ·
Known Solutions each Motherboard Manufacturer:

Gigabyte:
  • Package Power Limit 1: 400W (prevent throttling due to over-current)
  • Package Power Limit 2: 400W (prevent throttling due to over-current)
  • CPU VCore Loadline Calibration = Medium (prevent the vdroop on VCCIN)
  • CPU VCore Current Protection = High (prevent the hard shutdown due to over-current)
ASUS:

The only information here is very old and needs to be updated.

der8auer explains that you need to disable SVID and set CPU current capability to 140%. Presumably this gets around all the over-current protections, but it doesn't say anything about the VCCIN vdroop.

Asus motherboards are supposed to have better LLC settings to counter the VCCIN vdroop.

(If you have a solution for a board/manufacturer that isn't listed here or if you see something that doesn't work, please post it.)
 
#3 ·
Mystical, as i understand, this is related to the TDP imposed by intel to keep temperatures and wattage at bay.
Im not seeing many reviews of the 7820X, the one im considering, im looking into the cpu to know what kind of wattage load i can expect under 4.5 - 4.6 around 1.2v.

Also, about the 7900x, this issue is only reported at P95 or other loads are suffering the same?
 
#4 ·
Quote:
Originally Posted by postem View Post

Mystical, as i understand, this is related to the TDP imposed by intel to keep temperatures and wattage at bay.
Im not seeing many reviews of the 7820X, the one im considering, im looking into the cpu to know what kind of wattage load i can expect under 4.5 - 4.6 around 1.2v.

Also, about the 7900x, this issue is only reported at P95 or other loads are suffering the same?
Unfortunately, I also don't have any information about the 7820X. Aside from obviously having 2 fewer cores, they're also not supposed to have the full AVX512 capability.

Personally, I haven't hit the throttling for non-AVX512 loads because I haven't done much testing above 4.2 GHz. But others are hitting it for all loads (prime95 or not) in the 4.7+ GHz range. der8auer mentions in his video hitting it on prime95 at 4.5GHz for the 7900X. In my emails with Silicon Lottery, he says he's been hitting a lot, but didn't give me any specifics. I assume these are also in the 4.5 GHz range given those are the clock speeds of the chips he sells.

If I had to guess, the 7820X won't hit it until the 4.8 GHz range - and only under something like prime95. AVX512 loads shouldn't be any worse since they're only supposed to run at half speed on the 7820X. But this has yet to be confirmed.
 
#5 ·
Good, i dont have nor pretend to use AVX loads.
P95 is already pretty damn hammering on cpu, if i record well, SL uses the asus bench for stability tests and it envolve encoding with avx calls.
This cpu will be used for gaming and a bunch of vms, in normal usage it should stay pretty stable.
 
#6 ·
Chrome browser uses AVX. I don't like where the technology is going towards restriction.

Thanks for posting all this information.
 
#8 ·
Quote:
Originally Posted by Mysticial View Post

Alright. It looks like the 7800X has been confirmed to have the full-throughput AVX512. So presumably, the 7820X will have it too:

Source: http://www.pcgameshardware.de/Skylake-X-Codename-266252/News/Core-i7-7800X-AVX512-Durchsatz-1232713/

This goes against all the pre-launch reviews about the 7800X and 7820X not having the full AVX512.
Wasn't expecting this. Intel is all about crippling features to upsell costly CPUs like they done with pci lanes. You confirmed it with flops output?
 
#9 ·
Quote:
Originally Posted by postem View Post

Wasn't expecting this. Intel is all about crippling features to upsell costly CPUs like they done with pci lanes. You confirmed it with flops output?
Correct. The guy confirmed with me that it was running at 3.5 GHz under AVX512.

The benchmark shows: 677.76 GFlops

Theoretical limit (half-throughput AVX512): (1 FMA/cycle) * (2 Flops/FMA) * (8 DP/inst for AVX512) * (6 cores) * (3.5 GHz) = 336 GFlops
Theoretical limit (full-throughput AVX512): (2 FMA/cycle) * (2 Flops/FMA) * (8 DP/inst for AVX512) * (6 cores) * (3.5 GHz) = 672 GFlops
 
#10 ·
Very informative. Thanks!

I'm also very curious how this affects the 7820X. It seems to be in a thermal sweet spot compared to the 7900X, and that fits me just fine! I'm too cheap for a 7900X anyway
tongue.gif
.

Question about AVX512.. Is it at all possible that production units of the 7820X are binned 7900X where two of the cores didn't pass, and therefore have both FMAs? Is it only some of them?
 
#11 ·
Quote:
Originally Posted by Urtie View Post

Very informative. Thanks!

I'm also very curious how this affects the 7820X. It seems to be in a thermal sweet spot compared to the 7900X, and that fits me just fine! I'm too cheap for a 7900X anyway
tongue.gif
.
I don't know anything about the 7820X. So people with those chips will have to do their own testing.

I didn't manage to throttle the 7900X under 8 thread/8 cores at stock. I might've come close to it though. And I imagine it might be possible with 16 threads/8cores, but I haven't tried disabling cores to simulate the 7820X. But seeing as how all the chips have the full-AVX512, it's not something to just blow off - even if nothing out there uses it yet. In any case, the AVX512 might be irrelevant here. As other people have reported the phantom throttling with AVX and normal code - just at higher frequencies.

Quote:
Question about AVX512.. Is it at all possible that production units of the 7820X are binned 7900X where two of the cores didn't pass, and therefore have both FMAs? Is it only some of them?
Yes. All the 7800X's, 7820X's, and 7900X's come off of the same assembly line with the same 10-core die design (the LCC die). Intel then bins them depending on which cores work, how high they clock, and supposedly which ones had a working dedicated 512-bit FMA.

It seems that the binning on 512-bit FMA never happened at all. And if Intel initially wanted to disable them on the 7800X and 7820X for market segmentation, it was purely for market segmentation and not because of yields (i.e. faulty 512-bit FMA).

And with pressure from AMD, Intel is turning them all on. (at least that's my hypothesis for now)
 
#12 ·
Hi, Mysticial. Have you figured out how TJ-Max Offset works in Gigabyte BIOS? Elmor pointed out the problem before but we are still having no solution to increase overheat temp limit properly. My situation is that runnung my 7900X @locked 4.3GHz AVX with 3 offset, My CPU still got throttled at 4.0GHz with Prime95.
 
#13 ·
Quote:
Originally Posted by CuewarsTaner View Post

Hi, Mysticial. Have you figured out how TJ-Max Offset works in Gigabyte BIOS? Elmor pointed out the problem before but we are still having no solution to increase overheat temp limit properly. My situation is that runnung my 7900X @locked 4.3GHz AVX with 3 offset, My CPU still got throttled at 4.0GHz with Prime95.
I haven't played with the TjMax offset because I don't intend to let my chip go above 95C. So the temperature throttling is something that I actually want.

My 7900X doesn't temperature throttle under Prime95 Small FFTs AVX @ 4.0 GHz. The temps stay under 90C. While I didn't delid my chip, I am running a 360 AIO.
 
#16 ·
Thought I'd post an update to this. As of BIOS F7I for the Gigabyte 7, the VRIN/VCCIN vdroop issue has been fixed. So you no longer need to increase it or turn on load-line calibration.

I presume that the latest BIOS for all the other Gigabyte boards have this fix as well.
 
#17 ·
Has anyone collected any data on this for MSI motherboards? I'm having a weird issue with my 7820x and Gaming Pro Carbon. It seems to score well and increase linearly with clocks in xtu, cinebench and firestrike but my time spy and 3d11 physics are down (10800 and 16000 respectively). I haven't tested all clocks in these, just in fire strike, so maybe I need to do this to see if it plateaus. My stable daily clock is 4700 1.22v, 3000 cache and 4000-16-16-16-28 1T on the rams at 1.45v.

Any help would be great
 
#18 ·
Quote:
Originally Posted by OZrevhead View Post

Has anyone collected any data on this for MSI motherboards? I'm having a weird issue with my 7820x and Gaming Pro Carbon. It seems to score well and increase linearly with clocks in xtu, cinebench and firestrike but my time spy and 3d11 physics are down (10800 and 16000 respectively). I haven't tested all clocks in these, just in fire strike, so maybe I need to do this to see if it plateaus. My stable daily clock is 4700 1.22v, 3000 cache and 4000-16-16-16-28 1T on the rams at 1.45v.

Any help would be great
Not that I'm aware of. But if you want to investigate for yourself, there are two ways to do it:

Start from a low clock (3.5 GHz) and slowly work your way up.
Each time you increase, record your benchmark scores, temperatures, and power draw. If they plateau, then you're hitting bottlenecks. But if the performance drops when you pass a certain speed, then you're throttling. And if the CPU frequency doesn't drop with it, then it's phantom throttling.

The other method (the one that I used to discover this) was to pick a static frequency that is very high. (For non-AVX stuff, you can try 4.8 GHz.)
Start by benching on 1 thread only, then 2 threads, then 3, etc... Again each time you step up, record the benchmark scores, temperatures, and power draw. Just as before, If you see a sudden drop in anything when you go past a certain # of cores loaded, then you're throttling.

In the case of the Gigabyte boards, the drops that indicate phantom throttling were extremely large. Something like 2x drop in performance and power draw simply by increasing the CPU frequency by 100 MHz.

Based on my experience with the Gigabyte board, the phantom throttling kicks in when you pull more than 260 - 280W. That kind of power draw is easy to do on the 7900X. But it'll probably be a lot harder on the 7820X. I haven't seen anyone phantom throttle on anything other than the 7900X so far.
 
#19 ·
Does anyone have settings for the X299 MSI Gaming Pro Carbon?

Temps are good but 'Power limit throttling' is happening according to Intel XTU. This is at 4.7Ghz with -7 set for AVX. vCore is 1.2v.

Quote:
Originally Posted by OZrevhead View Post

Has anyone collected any data on this for MSI motherboards? I'm having a weird issue with my 7820x and Gaming Pro Carbon. It seems to score well and increase linearly with clocks in xtu, cinebench and firestrike but my time spy and 3d11 physics are down (10800 and 16000 respectively). I haven't tested all clocks in these, just in fire strike, so maybe I need to do this to see if it plateaus. My stable daily clock is 4700 1.22v, 3000 cache and 4000-16-16-16-28 1T on the rams at 1.45v.

Any help would be great
EDIT:

It seems that the 'Turbo Boost Power Max' setting in Intel XTU was set to 270W. My CPU is just jumping over that so I raised it to 400+ watts and now it looks like the power limit throttling has gone!

I can't find any MSI BIOS settings that look like the same thing, so until someone can answer that, Intel XTU will have to do the job!
 
#20 ·
Quote:
Originally Posted by The_Rocker View Post

Does anyone have settings for the X299 MSI Gaming Pro Carbon?

Temps are good but 'Power limit throttling' is happening according to Intel XTU. This is at 4.7Ghz with -7 set for AVX. vCore is 1.2v.
EDIT:

It seems that the 'Turbo Boost Power Max' setting in Intel XTU was set to 270W. My CPU is just jumping over that so I raised it to 400+ watts and now it looks like the power limit throttling has gone!

I can't find any MSI BIOS settings that look like the same thing, so until someone can answer that, Intel XTU will have to do the job!
By now, it's pretty clear that the phantom throttling in particular is caused by the vdroop in the VCCIN. So on all motherboards, the fix is to either increase VCCIN so it stays above 1.75v at all times or to enable the right LLC to compensate for it.

But as you've noticed, there's a lot of other types of (non-phantom) throttling as well. (I count 17 of them in Intel's technical docs.) But getting around those is unfortunately motherboard-specific since each mobo/BIOS gives slightly different names to all the various throttling limits.
 
#21 ·
Quote:
Originally Posted by OZrevhead View Post

Has anyone collected any data on this for MSI motherboards? I'm having a weird issue with my 7820x and Gaming Pro Carbon. It seems to score well and increase linearly with clocks in xtu, cinebench and firestrike but my time spy and 3d11 physics are down (10800 and 16000 respectively). I haven't tested all clocks in these, just in fire strike, so maybe I need to do this to see if it plateaus. My stable daily clock is 4700 1.22v, 3000 cache and 4000-16-16-16-28 1T on the rams at 1.45v.

Any help would be great
Quote:
Originally Posted by Mysticial View Post

By now, it's pretty clear that the phantom throttling in particular is caused by the vdroop in the VCCIN. So on all motherboards, the fix is to either increase VCCIN so it stays above 1.75v at all times or to enable the right LLC to compensate for it.

But as you've noticed, there's a lot of other types of (non-phantom) throttling as well. (I count 17 of them in Intel's technical docs.) But getting around those is unfortunately motherboard-specific since each mobo/BIOS gives slightly different names to all the various throttling limits.
Well ive been running AVX prime95 for the past 4 hours and my VCCIN is reading 1.808v. Maybe this means this MSI board isn't suffering the same issue?

I don't notice any drop in temps or power consumption anyway.
 
#22 ·
I've noticed the "Vccin tracker" setting in the new Asus bios really works. No change in ycruncher bbp time with vccin 1.8v and LLC1 (lowest). At this setting vccin droops to 1.6v at load.
rolleyes.gif
 
#24 ·
Quote:
Originally Posted by Nautilus View Post

I did everything suggested in the thread, yet my phantom throttling still remains. x264 stress test causes my 4500mhz overclock to go down as much as 4200mhz on random cores.

I tried with VRIN as high as 1.95 still no go.

7900x and gigabyte aorus gaming 9.
Since you can see the clock drop, it's not the phantom throttling. You're throttling for other reasons. Are you hitting Tj.Max?

HWInfo64 will be able to catch most throttling reasons. But it doesn't detect phantom throttling itself. So you can try using that to see what you're throttling on.
 
#25 ·
Quote:
Originally Posted by Mysticial View Post

Since you can see the clock drop, it's not the phantom throttling. You're throttling for other reasons. Are you hitting Tj.Max?

HWInfo64 will be able to catch most throttling reasons. But it doesn't detect phantom throttling itself. So you can try using that to see what you're throttling on.


Thank you. I think I found it. Says: Max Turbo Limit. Uhmm but it doesn't make any sense because i disabled all the Turbo.
 
#26 ·
Help I'm literally having the same problem as this guy above me... Everything was just fine for me for a while I had my 7900x @ 4.7 1.275v and it was all great and suddenly today all of my scores took a poop and I look in hwinfo and I see that? I have no clue what to do... My mobo is the Asus X299 deluxe and I have a custom loop with two 360 rads and there is zero thermal issues. Delidded CPU and my temps never went above 75. I've looked everywhere and cannot find a thing so any help would be GREATLY appreciated! Thank you!
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top