Question Why does TDP and PPT differ, on consumer CPUs? And what role does Core Performance Boost and Turbo Clocks have on TDP and wattage?

[VirtualLarry]

Serious question. I've got a 65W-rated TDP Ryzen R5 1600, in a rig, on a 65W-rated AMD stock heatsink. It's blue-screening, crashing, and the CPU temps just keep going up and up.

I updated HWMonitor, and it's showing a "Package Power" for the entire chip, at 82W or so. No wonder it's constantly overheating and crashing. 82W TDP CPU > 65W TDP Heatsink.

The worst part is, this is AFTER limiting the number of PrimeGrid threads, down from 12 to 9. That's right, I'm not even running the CPU at a full thread load.

Edit: Yes, I know that the obvious answer, is to "get a better heatsink", and that the "stock heatsink" for the 1600 was the 95W TDP model. Which, at the time, was stated that AMD wanted to give users the ability to OC on the stock heatsink. Now I know that was a lie, it's because AMD CPUs (at least, the 1600), are NOT able to stay within their stated rated specs.

Edit: A slight update, very important, actually. My original premise for this thread, was that I *thought* I was using a 65W TDP-rated AMD stock Wraith Stealth cooler with my Ryzen R5 1600 CPU, and it was crashing, at "stock BIOS" settings, which includes "Core Performance Boost" on "Auto", which defaults to enabled, to allow "Turbo Clocks" (the 1600 has an ACT of 3.4Ghz). I was initially placing the blame on AMD for the fact that HWMonitor reported the "Package Power" as something like 82W, which I thought was overcoming the 65W-rated heatsink. As it turned out, I actually was using a 95W Wraith Stealth (copper-cored) in the first place. Yet, it was still crashing due to overheating of the CPU. Part of this was due to the heat load of dual GPUs mining, and part of it was due to using a case that had NO vents on top, no fan mounts, no rad mounts, nothing but a solid steel top, and only a single 120mm exhaust out the rear, combined with the fact that my PCs are in desk cubbies. They are open to the front, and have dual 120mm intakes and vented fronts, but that still wasn't enough to prevent the CPUs from slowly creeping up in temp, passing 95C, and crashing/restarting.

Thus far, I have split the two GPUs up, one per PC (same case, same type cubby, same EVGA 650W G1+ 80Plus Gold PSUs), and disabled CPB on both of them (one has a 3600 CPU, one has a 1600 CPU), and then also in Wattman, set the Power Limit for the RX 5600XT (which was a refurb, both of them) to -20%. Thus far, overnight, they seem to have stabilized at under 90C on the CPU, and haven't crashed.

 

[VirtualLarry]

PrimeGrid is a workload which any CPU at default settings, with a cooler which is sized and operated as specified by the CPU vendor (which includes parameters such intake air temperature) needs to run stable and return correct results. (I don't know whether or not @VirtualLarry's setup complies with AMD's specs.)

AMD specifies the Thermal Design Power with 65 W, but at the same time configures the Package Power Tracking target higher than the TDP. This is the central problem. *Why* is AMD doing that? They should not. It's idiocy. My guess is they do it because their competitor does it (and goes about it even worse than AMD). And then there are BIOS vendors who love to make it all even much worse with their own settings.

By the way, while the firmware of EPYCs have separate values for TDP and PPT too, AMD sets these to the same value by default (in the SKUs which I am aware of).

Thank you Stephan. That's really the meat of the matter right there.

 

[VirtualLarry]

Running a heavy non-stop load (like PrimeGrid) and using a stock cooler is not what Intel or AMD has designed the stock coolers for. They are perfectly fine for how most people use their PCs for, and the "average joes" aren't pegging it at a constant100% doing things like mining for crypto or PrimeGrid. And for users who do that kind of thing, they buy an aftermarket cooler (AIO or air).

Exactly. I may run 100% 24/7, but I have AIO's or top of the line air on all 18 systems, and none in enclosed spaces and all with the side cover odd, and only one GPU in each. You (@VirtualLarry )violate most every one of those rules from what I have read.

As far Larry's issue my recommendation would be to get an aftermarket cooler simply because running a CPU at 100% doing any kind of long term workload on the stock cooler worries me. The stock coolers are generally just BARELY capable of handling 100% loads for their rated TDP they are generally not designed to be running at that for long periods of time. Case airflow as well as ambient air temperature could easily cause issues in that scenario. Haven't had any issues with my Ryzen CPUs blue screening/crashing due to thermal limits though and I generally have them running a bit hotter than most because I would prefer the lower noise from not running my fans quite as high. Also want to say mine thermal throttle themselves fine I think, but at the same time I don't believe I run any of mine in a situation that actually causes them to thermal throttle so may be wrong on that. Just want to say I saw throttling under a stress test before I adjusted my fan profile a bit, but I can't 100% remember.

I agree with all of those, in principle. I DO believe that you need to use an AIO, at least 240mm, with any AM4 6C/12T or above, when running loads such as PrimeGrid.

Part of the problem though, is my case, both of my secondary rigs use a Rosewill Magnetar case, which has a solid steel top, NO top ventilation, and NO top-mount 240mm Rad.

I'm looking at upgrading my cases to something a little more cooling and cubby-friendly (top venting/cooling/rad. mounting), like a Rosewill Cullinan MX RGB.

As far as my (stock settings, I just checked) Ryzen R5 1600 not throttling properly, it's in a Gigabyte X370 Gaming ATX board, with BIOS F25.

I should, however, check vSoC setting on it, because on these Gigabyte boards, at least, my B450 AORUS PRO WIFI, it was set a bit high on "Auto", and IIRC that raised my temps a little. (My other secondary rig has the AORUS PRO WIFI ATX B450 board, with a Ryzen R5 3600, and a Gammaxx 400 Blue LED tower cooler. That one was overheating and crashing too.)


As an aside, is it possible for a GPU (or a pair of them), to somehow, be shorting out in a way that causes additional +12V draw, such that the mobo CPU VRMs would be wanting for voltage, thus driving up current, and thus CPU temps? Is that even theoretically possible?

Because I'm having the issue with both of my secondary rigs, when I move both RX 5600XT cards into the case. And as I tested, even pulling the case out of the cubby and removing the side panel, doesn't seem to mitigate the CPU overheating eventually, after a few hours, and then crashing/BSODing/rebooting.

 

[Hitman928]

@VirtualLarry , You absolutely do not need a 240mm AIO for any stock 6C AMD cpu. A decent after market air cooler is more than sufficient to keep thermals down even with 24/7 heavy load. If your system is crashing running stock in a non-confined space with the case side off, then there is something wrong with the system, it's not from running the stock cooler. Additionally, as I mentioned, AMD desktop CPUs BY DESIGN will use more than their rated TDP but ONLY when power and thermal limits allow for it and you can easily keep it from exceeding TDP by simply turning off XFR/PB which is what allows it to go over its TDP as a feature. Even with these on, if you reach thermal limits, the CPU will downclock itself to TDP limits or lower if necessary so there is something else wrong in the system.

 

[VirtualLarry]

Additionally, as I mentioned, AMD desktop CPUs BY DESIGN will use more than their rated TDP but ONLY when power and thermal limits allow for it and you can easily keep it from exceeding TDP by simply turning off XFR/PB which is what allows it to go over its TDP as a feature. Even with these on, if you reach thermal limits, the CPU will downclock itself to TDP limits or lower if necessary so there is something else wrong in the system.

I know that's true on the 3600, which has arguably far more advanced thermal and electrical monitoring embedded into the CPU, but the 1600 "AE" (old-school) too? Are you certain?

Edit: I thought that the NON-X Zen1 CPUs didn't have XFR/PB? I'll check the BIOS, just to be certain. If there is a way in the BIOS, to constrain the actual Package Power to 65W, that's what I'd like to do to mitigate the temps and crashes.

Edit: As I suspected, there were no XFR/PB settings in the BIOS, but there was one for "Core Performance Boost", I changed that from "Auto" to "Disabled". Also, I went into AMD CBS Settings, the bottom option, and then "cTDP", set that from "Auto" to "45", and then "CPU Temp" (throttle?), I set that to "80".

Now, Package Power is hovering around 67-70W, and temps are under 80C, for now. Maybe it will survive this way.

BTW, I noticed on the PC Health section, that Vsoc was 1.200 to 1.202V, which it's supposed to be 1.1V max. But there was no setting to adjust it! (This is a really stripped-down BIOS, as Gigabyte does, on their lower-end boards. No apparent way to even adjust Vsoc, even though the motherboard overvolts it at stock! Gah!)

 

[coercitiv]

AMD specifies the Thermal Design Power with 65 W, but at the same time configures the Package Power Tracking target higher than the TDP. This is the central problem. *Why* is AMD doing that? They should not. It's idiocy. My guess is they do it because their competitor does it (and goes about it even worse than AMD). And then there are BIOS vendors who love to make it all even much worse with their own settings.

That's not the central problem and it's not idiocy. AMD has a rather smart system set in place to squeeze performance out of these CPUs. This "idiotic" power management you're talking about helped someone like me run 1600X without active cooling on a Scythe Ninja heatsink, using only airflow from low rpm case fans. Imagine that, my 95W TDP CPU with stock settings managed to keep temps at a constant 75C while gradually throttling down from 95W+ under Prime95 to 70-75W where it achieved thermal stability. I didn't even touch one knob and this idiotic CPU found the exact thermal limit of my cooler setup and extracted maximum performance while keeping temps safe. Stupid stupid AMD CPU, should have stayed at 95W.

Ask @VirtualLarry if he's ever had stability and thermal issues with this system before, you might be surprised. Ask yourself why his system is crashing when others are running just fine. I had mine running heavy CPU & dGPU workloads 24/7 with no stability issues, even without direct airflow on the heatsink.

Are you certain?

Yes.

 

[Hitman928]

I know that's true on the 3600, which has arguably far more advanced thermal and electrical monitoring embedded into the CPU, but the 1600 "AE" (old-school) too? Are you certain?

Yes, thermal throttling has existed and worked in AMD CPUs for over a decade now, it's not new technology. The newer gen CPUs improvements in thermal and current monitoring allow for more precise hotspot monitoring, more frequency options, and quicker response times for boost algorithms, but it's not anything new in terms of thermal protection.

 

[Thunder 57]

I DO believe that you need to use an AIO, at least 240mm, with any AM4 6C/12T or above, when running loads such as PrimeGrid.

Bullocks.

 

[VirtualLarry]

Ask yourself why his system is crashing when others are running just fine.

Ask yourself if you've ever run PrimeGrid, one of the LLR tasks, on your rig.

 

[amrnuke]

WTH is this supposed to mean? I'm not overclocking. CPUs aren't tires. I was running it at stock, I expected it not to "blow out" (using a tire analogy), but it was, repeatedly.

Have you run a Ryzen R5 1600 on wraith stealth? For more mundane, "normal" tasks, like web browsing, sure, it works fine.

But try running PrimeGrid, 12 threads of pure AVX2 torture for the CPU. The Package Power is over 80W, and termps on a 65W TDP heatsink, will just go up, and up, and up.

This is exactly my point.

You're running a track day ("pure AVX2 torture for the CPU") with daily-driver equipment (a 1600 with Wraith Stealth "for more mundane, normal tasks").

If you're unhappy with how the CPU and cooler are performing in PrimeGrid then return them.

 

[Hitman928]

I know that's true on the 3600, which has arguably far more advanced thermal and electrical monitoring embedded into the CPU, but the 1600 "AE" (old-school) too? Are you certain?

Edit: I thought that the NON-X Zen1 CPUs didn't have XFR/PB? I'll check the BIOS, just to be certain. If there is a way in the BIOS, to constrain the actual Package Power to 65W, that's what I'd like to do to mitigate the temps and crashes.

Edit: As I suspected, there were no XFR/PB settings in the BIOS, but there was one for "Core Performance Boost", I changed that from "Auto" to "Disabled". Also, I went into AMD CBS Settings, the bottom option, and then "cTDP", set that from "Auto" to "45", and then "CPU Temp" (throttle?), I set that to "80".

Now, Package Power is hovering around 67-70W, and temps are under 80C, for now. Maybe it will survive this way.

BTW, I noticed on the PC Health section, that Vsoc was 1.200 to 1.202V, which it's supposed to be 1.1V max. But there was no setting to adjust it! (This is a really stripped-down BIOS, as Gigabyte does, on their lower-end boards. No apparent way to even adjust Vsoc, even though the motherboard overvolts it at stock! Gah!)

AMD has changed the marketing of their boost algorithm over the last few years, but core performance boost is what you want to turn off which looks like it had the desired effect. I don't think the 1600 supports cTDP so I don't think changing that will have any effect.

 

[TheELF]

Serious question. I've got a 65W-rated TDP Ryzen R5 1600, in a rig, on a 65W-rated AMD stock heatsink. It's blue-screening, crashing, and the CPU temps just keep going up and up.

I updated HWMonitor, and it's showing a "Package Power" for the entire chip, at 82W or so. No wonder it's constantly overheating and crashing. 82W TDP CPU > 65W TDP Heatsink.

The worst part is, this is AFTER limiting the number of PrimeGrid threads, down from 12 to 9. That's right, I'm not even running the CPU at a full thread load.

Edit: Yes, I know that the obvious answer, is to "get a better heatsink", and that the "stock heatsink" for the 1600 was the 95W TDP model. Which, at the time, was stated that AMD wanted to give users the ability to OC on the stock heatsink. Now I know that was a lie, it's because AMD CPUs (at least, the 1600), are NOT able to stay within their stated rated specs.

They are not lying, stock settings are supposed to cut off the CPU at 65W for long term reducing your performance to keep cool, if your system doesn't do that then it's due to the mobo/bios.
You are not supposed to run a power virus at max clock speeds indefinitely at TDP.
Your system hangs because the fail safe of the CPU itself gets triggered and shuts down the CPU to prevent damage.

 

[blckgrffn]

As others have pointed out, the BIOS does all sorts of craziness depending on nearly endless variables so if you are having issues disabling XMP, auto boosts, etc. are a must.

Even a Hyper 212 or a used Wraith Spire or Prism should be able to take the full brunt of what you are trying to do. $20 and problem solved. The premium wraith coolers may not be silent but at least they can get the job done and are securely mounted.

Failing that, there are a one of coolers between $40 and $60 that will take a that CPU at crazy settings without an issue (Scythe, beQuiet! and more). I was looking into air coolers more recently and I didn't realize how stacked that price range really was.

 

[Leeea]

Same reason Intel, Nvidia, and everyone else does. Even more annoying, they do each do it differently, so they cannot be compared to each other.

Perhaps TDP is not a good metric to look at?

"It's blue-screening, crashing," ...
I have not seen a stock cpu crash when over heating. They just throttle back. With yours I believe that starts at 95C. It is possible to run a Ryzen without a heatsink at all*.

If your overclocking, well then it is your fault.
Otherwise, odds are your problem is not being caused by the cpu.


*https://www.overclock3d.net/news/cp...bile_cpus_can_run_crysis_without_a_cpu_cooler
(or https://www.google.com/search?q=running+ryzen+without+heatsink&oq=running+ryzen+without+heatsink )

 

[VirtualLarry]

"It's blue-screening, crashing," ...
I have not seen a stock cpu crash when over heating. They just throttle back.

If your overclocking, well then it is your fault.
Otherwise, odds are your problem is not being caused by the cpu.

The problem here is, the CPU is TRYING to throttle back (or should be, it was mostly pinned at 95C when I went to check it just now and adjust settings), but I've got two 180W GPUs dumping heat into the case, which only has a 120mm rear exhaust, and no top ventilation.

So the CPU throttling, is fighting not only itself, but the GPUs' thermals too, and it wasn't winning that battle (BSODs, black-screen, crash, reboot, etc.)

NOT overclocking, just bone stock on this Gigabyte X370 Gaming mobo (which does, apparently, overvolt Vsoc to 1.200V). I AM running XMP, on some GSkill DDR4-3000 (nothing too extravagant).

I wasn't having issues with these rigs, running more mundane GPUs, until I put in the pair of RX 5600XT triple-fan cards (Navi10 GPU chip).

 

[VirtualLarry]

AMD has changed the marketing of their boost algorithm over the last few years, but core performance boost is what you want to turn off which looks like it had the desired effect. I don't think the 1600 supports cTDP so I don't think changing that will have any effect.

I'm not sure if cTDP is having any effect or not. (Perhaps I shouldn't have changed them both at once.) Although, "Core Power" is below 45W, but Package Power is not. Perhaps cTDP only affect Core Power? I am running XMP 1.35V DDR4-3000 2x8GB RAM too, and the mobo is overvolting Vsoc to 1.200V, but I can't seem to change it. (It actually seems like my entire voltage adjustment page is missing out of the M.I.T. settings page. Is there a "magic hotkey" that I need to hit? Gigabyte seems to love their "Magic Hotkeys".)

 

[Leeea]

The problem here is, the CPU is TRYING to throttle back (or should be, it was mostly pinned at 95C when I went to check it just now and adjust settings), but I've got two 180W GPUs dumping heat into the case, which only has a 120mm rear exhaust, and no top ventilation.

AMD cpu fails when heat gun is pointed at it. This is not news.


Actually, it sounds like you have power supply failure to me. Electrolyte capacitors are far less tolerate to heat then cpus. Either your on mainboard or your power supply itself. If interior case temperature is anywhere near 95 degrees C (which is 203 F), you should just go buy a computer from Best Buy.

 

[VirtualLarry]

Actually, it sounds like you have power supply failure to me. Electrolyte capacitors are far less tolerate to heat then cpus. Either your on mainboard or your power supply itself. If interior case temperature is anywhere near 95 degrees C (which is 203 F), you should just go buy a computer from Best Buy.

Get real. I never said that the case temps were 95C. If they were, I would be getting some bacon to cook on my case. CPU core temps of 95C (silicon hotspots), are necessarily going to be a lot higher than the actual air temp inside the case, which is again going to be higher than ambient temps in the room.

 

[Velgen]

Wouldn't be surprised if XMP was raising CPU temp by 2-4C not sure how much it would raise package power draw, but for the performance gain from being at 3000MHz vs 2133 it is probably worth it. Seems like the big issue is those GPUs dumping all that heat in there and there isn't enough exhaust to get it out of the system. Seems like you would need blower style cards in that case if you wanted to run dual GPUs or simply get a case that can better handle that kind of heat load. Oh and just remembered if you do plan to put it in a cubby make sure to cut out fan holes and mount some fans to circulate air through it so long as you circulate enough air through the cubby should be fine putting it there.

 

[coercitiv]

but I've got two 180W GPUs dumping heat into the case, which only has a 120mm rear exhaust, and no top ventilation

here comes the truth...

 

[Hitman928]

I'm not sure if cTDP is having any effect or not. (Perhaps I shouldn't have changed them both at once.) Although, "Core Power" is below 45W, but Package Power is not. Perhaps cTDP only affect Core Power? I am running XMP 1.35V DDR4-3000 2x8GB RAM too, and the mobo is overvolting Vsoc to 1.200V, but I can't seem to change it. (It actually seems like my entire voltage adjustment page is missing out of the M.I.T. settings page. Is there a "magic hotkey" that I need to hit? Gigabyte seems to love their "Magic Hotkeys".)

TDP is package power, not core power. If all this started after putting in beefier GPUs then I would start looking at it being a PSU, GPU, or possibly a motherboard issue.

Edit: Could even be memory if the memory modules are heating up too high due to the added GPU power not being exhausted fast enough.

 

[VirtualLarry]

Wouldn't be surprised if XMP was raising CPU temp by 2-4C not sure how much it would raise package power draw, but for the performance gain from being at 3000MHz vs 2133 it is probably worth it. Seems like the big issue is those GPUs dumping all that heat in there and there isn't enough exhaust to get it out of the system. Seems like you would need blower style cards in that case if you wanted to run dual GPUs or simply get a case that can better handle that kind of heat load.

That would indeed seem to be the primary issue, but even taking the case out of the cubby and taking the side off, the CPU was still overheating, which kind of puzzled me. I felt the heatsink, it was really warm, and wasn't loose, and the fan was spinning.

I'm willing to guess that the Vsoc overvoltage from 1.1V (max recommended) to 1.200V (Gigabyte BIOS default on "Auto"), is probably pushing package power higher too. I could probably lower my temps by 3-5C if I could get that down.

 

[UsandThem]

you should just go buy a computer from Best Buy.

Bwahahaha

here comes the truth...

Yup.

If you have two GPUs and a CPU crunching non-stop, and have one 120mm case fan exhaust along with a dinky CPU cooler, you're going to have a bad time.

 

[VirtualLarry]

here comes the truth...

But if AMD's thermal-throttling worked as well as Intel's, then it shouldn't be crashing. Even with two GPUs in the case. (GPU chip vendor is same as CPU vendor, GPU card vendor is same as mobo vendor. Don't they test these things together?)

 

[Velgen]

Personally I don't think Intel's thermal throttling would handle that situation particularly well either really from my experience of using Intel's stock heatsink as a temporary solution.

 

[Hitman928]

But if AMD's thermal-throttling worked as well as Intel's, then it shouldn't be crashing.

Have you actually monitored the CPU frequency during this time to see if it's not reducing frequency as temperatures near thermal limit? I still highly doubt this is a CPU thermal issue in terms of crashing.

Even with two GPUs in the case. (GPU chip vendor is same as CPU vendor, GPU card vendor is same as mobo vendor. Don't they test these things together?)

No, that's just silly.