Precision 7560 & Precision 7760 owner's thread

[Aaron44126]

1 hour ago, Etern4l said:

Performance State The current performance state for the GPU. States range from P0 (maxi- mum performance) to P12 (minimum performance).

 

Yeah, on desktops or pro GPUs the scale might be wider than the ones I enumerated above.  Those are what I have observed on laptops and there is a big drop-off between P2 and P3 (at least on Ampere GPUs in the Precision).

 

You can force a certain power state with NVIDIA Inspector but I don't think that will help here.  I tried messing with it before; if Dell wants you in a low power state, it's going to put you in a low power state regardless of what you try to force.

 

And now thinking about it, I have experienced "oscillations" before where a combined CPU/GPU load would force the GPU to throttle on and off periodically.  I think the best solution is along the lines of what @Etern4l suggested up above.  Do what you need to do to lower the CPU power draw to the point where the system stops throttling the dGPU.  You might have to be more aggressive with it.

 

As another test, you could run a high-GPU low-CPU load for a while and confirm that this issue doesn't occur.  I'd suggest 3DMark Fire Strike, just run a single GPU test in a loop, windowed.  It should put the GPU up to 100% utilization and not even stop for loading between loops.  (Then maybe add a separate CPU load and see if the GPU starts power throttling.)

 

[Edit]

One more thing.  Do you have a 180W or 240W adapter attached to your system?  Do you have access to a 240W to test with, if you are using 180W?  I don't know if it will matter, but at least on Precision 7670, Dell is shipping 240W with systems with a high-power dGPU.

[Etern4l]

Seeing P2 under load. Great question about the power adapter, could explain the power limit.  That said, downclocking (and necessarily undervolting) the CPU should help in that case as well as it would free up the power for the starved GPU. On the other hand how much more power can that GPU take, if it's running at 74C under 50-60W load. I would just repaste both the CPU and GPU immediately (well, as soon as 7950/7958 arrived from ebuy7). 

[Aaron44126]

52 minutes ago, Etern4l said:

Seeing P2 under load.

 

You're on a desktop right?  I think that the power management behavior is going to be different.  I always see the GPU spending most time in P0.  But, the laptop GPUs have a way lower max power level.

 

I had another thought.  If we're on the right track, this seems to be a case of a "not ideal" (putting it nicely) implementation of power sharing between the CPU and GPU.  Dell's older systems would just power-starve the CPU if the GPU was under load, rather than the other way around, and I think that would be the preferred behavior for gaming (and most other high-GPU loads).

 

Anyway.  Do you see the "NVIDIA Platform Controllers and Frameworks" device in Device Manager?  I think it is under "Software devices".  That driver is at least partly responsible for handling NVIDIA Dynamic Boost (shifting power allocation between the CPU and GPU).  I wonder if the system behaves any differently if you just disable it.

[Etern4l]

9 minutes ago, Aaron44126 said:

 

You're on a desktop right?  I think that the power management behavior is going to be different.  I always see the GPU spending most time in P0. 

 

Yes. Could be, or could depend on the OS, driver version etc. The clocks and performance are as expected, but... P2. Still, would be consistent with your P2->P3 falloff observation.

[Aaron44126]

You can disable Dynamic Boost in the NVIDIA control panel... but not for Ampere and later GPUs?  Though, maybe a procedure like this would be helpful.

 

[Edit]

Now I am also very interested in disabling or bypassing Dynamic Boost on Linux because I wonder if this might be explaining a few odd things I have been seeing with more intense GPU loads.  There is documentation on it.  nvidia-settings -q DynamicBoostSupport indicates that dynamic boost is supported, but I do not see a nvidia-powerd service/process running.  (Using NVIDIA 525 proprietary drivers from the Ubuntu repos.)

[Ionising_Radiation]

2 hours ago, Aaron44126 said:

Do you have a 180W or 240W adapter attached to your system?

I have a 180 W adaptor. I need to buy a 240 W one... Bummer. 

 

1 hour ago, Etern4l said:

if it's running at 74C under 50-60W load

The repaste was performed by the Dell technician who replaced my RTX A4000 for the RTX 3080. I have just purchased 80×80mm of the Honeywell PTM7950 thermal pads, which should arrive in a couple of weeks. I thought Thermal Grizzly Kryonaut (or Conductonaut, or Carbonaut) or any of the various carbon pads were the rage a couple years ago back in NBR...

 

[Aaron44126]

3 minutes ago, Ionising_Radiation said:

I have a 180 W adaptor. I need to buy a 240 W one... Bummer.

 

I'm not sure if that will really help you.  The system might not draw more than 180W of power even if you hook up a 240W.  I just thought that it might be something to try.

 

I'm right now more interested in what happens if you try to disable Dynamic Boost...  It might stop the system from trying to steal power from the GPU if there is CPU load pressure.

[Ionising_Radiation]

2 minutes ago, Aaron44126 said:

I'm right now more interested in what happens if you try to disable Dynamic Boost

Forgot to quote this. I have it disabled, and am running Halo Infinite now... Let's see if I get into any games, and I'll see if anything has changed.

[Etern4l]

54 minutes ago, Ionising_Radiation said:

 I have just purchased 80×80mm of the Honeywell PTM7950 thermal pads, which should arrive in a couple of weeks. I thought Thermal Grizzly Kryonaut (or Conductonaut, or Carbonaut) or any of the various carbon pads were the rage a couple years ago back in NBR...

 

Cool, I'm sure you will be amazed by the results. Be careful with the application, in that it's easy to rip up the "pad", and bear in mind there might be some short curing time (a day or two). Kryonaut is really only good for desktops if at all, Conductonaut is LM, Carbonaut and the other carbon pads would be useless in this case.

[Aaron44126]

On 5/7/2023 at 10:25 AM, Ionising_Radiation said:

Let's see if I get into any games, and I'll see if anything has changed.

 

Any luck?

 

I thought about this some more.  If it really is a CPU/GPU power contention thing, I was wondering if it makes more sense to just lower the CPU power limit (PL1/PL2) rather than specifically downclocking each of the cores.  The BIOS or whatever is setting this off probably cares more about the power use than the clock speed.

 

I'm thinking something like...

• Run a combined CPU/GPU load (3DMark loop + Prime 95?) and observe that the GPU power level and performance fluctuates.
• ...Dial down PL1/PL2 and repeat tests until it stops happening.  (Maybe just lowering PL2 would be OK.  It would only be necessary to do this if you plan to put the GPU to use.)

I'm going to be doing this on my system soon.  I have found similar behavior with more intensive titles that I have been testing on Linux, and it definitely explains some odd behavior that I saw when trying to run games in the full 4K on Windows.  GPU performance is inconsistent and it is often dropping into lower power levels.  For instance, if I just run the benchmark in "Shadow of the Tomb Raider" with high resolution and graphics settings, I can see the GPU going from smooth to stuttering and back every few seconds.  I can see the GPU "PowerMizer" state drop down when it goes to stutter mode.  It happens more readily on the "ultra performance" thermal mode, but using the "optimized"/"balanced" mode doesn't fully stop it from happening.  It doesn't happen at all if I disable turbo boost; GPU performance is consistent in that case, but I should be possible to have the CPU run at a good bit higher than the base speed and still be OK.

 

Should be able to adjust PL1/PL2 with Throttlestop on Windows or powercap on Linux ...?

 

I'm not thinking that taking steps to turn off "NVIDIA Dynamic Boost" will fully fix the problem.  It doesn't look like that is even working on Ubuntu, the "nvidia-powerd" dynamic boost service is not engaged (even though the files to support it are present).  I'm thinking the Dell EC/BIOS is at the root of this right now.

 

...Just too much high-wattage stuff going on in a system that is not properly set up to handle it?  And really, the wattage isn't even that high.  😕

[Ionising_Radiation]

27 minutes ago, Aaron44126 said:

Any luck?

Nothing yet; I spent much of yesterday fixing my home network and NAS, which continues today. I'll probably get onto it this evening once the Wi-Fi is back up for the rest of the family...

[Etern4l]

Power limiting and/or temp limiting are the (much) simpler approaches, and indeed might be worth trying fiirst. Controlling clocks is helpful if not necessary for effective voltage adjustment under power or temp constraints. Basically, limiting the clocks allows you to undervolt more. 

[Ionising_Radiation]

Back again. I want to note that the CPU draws around 50 W under gaming loads; the GPU takes up the bulk of the power draw (consistently 90 W). I have disabled the NVIDIA Platform Controllers and Framework device, and so far... no drops. 

 

[Etern4l]

Oh dear, the CPU is hitting 100C at 50W. Struggling for polite words here :(

[Ionising_Radiation]

11 hours ago, Etern4l said:

Oh dear, the CPU is hitting 100C at 50W. Struggling for polite words here 😞

There's nothing to say; I haven't bothered with custom thermal management besides undervolting until now. It's a pain in the neck to open up and fiddle with my notebook; I have increasingly wanted to get something that just works. My ambient temps aren't great, either: it's currently 32 °C. It's a shared heatpipe set-up; the GPU is also outputting 90 W, which makes the CPU struggle to cool itself, too. 

[Etern4l]

4 hours ago, Ionising_Radiation said:

There's nothing to say; I haven't bothered with custom thermal management besides undervolting until now. It's a pain in the neck to open up and fiddle with my notebook; I have increasingly wanted to get something that just works. My ambient temps aren't great, either: it's currently 32 °C. It's a shared heatpipe set-up; the GPU is also outputting 90 W, which makes the CPU struggle to cool itself, too. 

 

I mean, we can say something... For instance:  disappointing, not great.

 

In the screenshot you posted before the GPU was at 54W and CPU was still hitting 100C. But as actually my Alienware m15 R1 (2019) is not that much better, if at all. In pure CPU benches it can only do 56W before the CPU throttles (on a good standard paste). Nothing you can buy will just work (unless you accept significant compromises), but looks like the the current top gaming laptops would work much better, since manufacturers actually focus on performance there. Either way, you need to be able to tune effectively to get the most of any high end system, bar perhaps $50k HEDT machines. Took me months to work it out on the desktop side. 

[Ionising_Radiation]

5 hours ago, Etern4l said:

Took me months to work it out on the desktop side. 

This is not something I want to do nor spend time on. I used to be really interested in tinkering, but now I just want things to work, even if it's at 85-90% of the theoretical maximum performance.

I really dislike messing around with hardware now, and even this PTM9750 application, I'm only getting it because from what I hear, it's a one-time change. Hopefully temperatures come under control.

It's why I got a Precision in the first place (next-biz-day on-site warranty with accidental damage coverage), but these power drops are irritating. Next time I'll try HP's Zbook. I've always wanted those, their builds always looked slightly better than Dell's (they even had MXM until 2019 with the Zbook 15 G6), but they're just a bit too expensive. 

[Etern4l]

2 hours ago, Ionising_Radiation said:

This is not something I want to do nor spend time on. I used to be really interested in tinkering, but now I just want things to work, even if it's at 85-90% of the theoretical maximum performance.

I really dislike messing around with hardware now, and even this PTM9750 application, I'm only getting it because from what I hear, it's a one-time change. Hopefully temperatures come under control.

It's why I got a Precision in the first place (next-biz-day on-site warranty with accidental damage coverage), but these power drops are irritating. Next time I'll try HP's Zbook. I've always wanted those, their builds always looked slightly better than Dell's (they even had MXM until 2019 with the Zbook 15 G6), but they're just a bit too expensive. 

 

I hear you. I had very specific reasons to undergo the effort (and it's a different story on the desktop side)!, the effort was actually mostly around tuning software settings (of which there is a gazillion in the BIOS), but would definitely prefer not to have to do that in general. Sadly it looks like manufacturers just struggle to get things running well out the box. Arguably the behaviour you desire (slower but steady wins the race), should be something you just experience in balanced performance mode. 

[Aaron44126]

52 minutes ago, Ionising_Radiation said:

This is not something I want to do nor spend time on. I used to be really interested in tinkering, but now I just want things to work, even if it's at 85-90% of the theoretical maximum performance.

 

I'm in much the same boat.  (...Says the guy who has recently thrown in the towel with Windows and is currently spending way too much time "tinkering" with Linux.)  Talking more about the hardware/thermal tuning side.  For instance, I haven't messed with undervolting, I don't really have the appetite to figure that out given that it looks like it is on the way out (for business-focused systems).  I did repaste my 7770 when it was new, but since I had the cooler replaced last month I've just been using factory paste.  It is "good enough" and I'd rather just spend time using the system than fine-tuning that.  I'm already annoyed that I have to spend time running tests to figure out where to set the CPU power limit to avoid this GPU throttling behavior.

 

So...  I'm also probably done with Dell.  There are a lot of things I like about the Precision 7770, but it has been sort of a hassle to get consistent gaming performance out of it (less because of thermals and more because of power management getting in its own way, as we are seeing here), and also it is also definitely a step down from older Precision laptops in terms of how easy it is to swap out components, should that be needed.  For example, the keyboard is no longer able to be swapped out without removing it from the bottom (requires a full disassembly), and the display panel is held in its enclosure by adhesive, not screws, making replacement of that an exercise in tedium to get it aligned just right.  They also "broke" manual fan control.

 

Well, that's for personal systems anyway.  At work, we are a Dell shop, so I'll be continuing my line of Precision laptops for work probably forever.

 

I've looked at HP ZBook and I'm not really sold there.  They seem to gimp the GPU power even more than Dell does.  I also really care about fan behavior.  They have a "never turn off the fans" BIOS option, awesome, but it turns out that it runs the fan a a consistent 2000+ RPM which is too noisy for an idle workload.  The price is consistently high, and because less people have them, there is less of an aftermarket parts market.

 

I'm paying close attention to Framework Laptop 16 and really am interested to see the full specs reveal on that.  I really like the idea of being able to have a long-term system that can be upgraded piecemeal, like you can do in a desktop.  (Shouldn't all computers be that way?)

 

Since I got the 7770 less than a year ago, it will be some years yet before I'm ready to switch again.  Hopefully by then Framework has a satisfactory option.  I really just want to see a high-DPI 120Hz display, a reasonably high-end NVIDIA GPU, and "not a downgrade" in terms of RAM or storage from what I have now.  (Bonus if they release an 18" version.)

[Etern4l]

Just to add to the previous post. I think Intel's high end CPUs just destroyed the thin and light dream - both the laptop and desktop variants put out too much heat for any reasonable cooling solution to handle. I've read somewhere that the heat flux generated by the 13900K is similar to that of a Bunsen burner lol

 

Now, that wouldn't necessarily be a huge issue if manufacturers just provided suitably sized laptops, with beefy cooling solutions. Alas, Dell and some others have painted themselves into a corner by trying to follow Apple's thin and light design. Apple cleverly side-stepped much of this by switching to power-efficient ARM/RISC CPUs, whereas Dell continues  putting Intel gas burners into thin throttlebooks on the high end. Don't get me wrong, I like Intel high-end CPUs - perform marvelously once tamed, but they can be incredibly frustrating if not configured and cooled properly, and most laptops fail to cool a 60-90W CPU (as per here).

[Ionising_Radiation]

10 hours ago, Aaron44126 said:

I also really care about fan behavior

I'm okay with this. I live at the Equator, it's bloody hot all the time, and air conditioning is a LOT louder than any laptop fan I've seen. Plus, I almost always have headphones on... 

 

I do want to see reviews for the Fury G10, though. The build, display options, keyboard, track pad buttons, tool-less access, vapour chamber, all make for a much more attractive package than the Precision. There's always a return window if I don't like it... 

 

Furthermore, It looks like NVIDIA Ada is significantly more efficient than Ampere is. 

 

10 hours ago, Aaron44126 said:

Framework Laptop 16

Problem with this (and most other start-ups) is that warranty and support is nearly non-existent where I live, and after having enterprise quality support the past 5 years, I can't go back to waiting weeks for component replacement. Dell literally has a regional distribution centre in Singapore, can't get better than that. I've gotten motherboard and GPU replacements within a day of calling support.

 

I'll concede, the Framework certainly looks very interesting, but I still think it's too early to adopt them for me... I'd wait another 5 years or so, at least, let them get some enterprise contracts. 

 

8 hours ago, Etern4l said:

power-efficient ARM/RISC CPUs

I think this is much more an Intel problem than it is an x86 vs ARM problem. AMD has managed very good power efficiency at the 35 – 55 W range with Zen 4, whereas Intel's P cores need to draw > 60 W to reach their efficiency peaks (but then at these power draws, laptops begin to throttle). 

[Aaron44126]

22 minutes ago, Ionising_Radiation said:

Problem with this (and most other start-ups) is that warranty and support is nearly non-existent where I live, and after having enterprise quality support the past 5 years, I can't go back to waiting weeks for component replacement. Dell literally has a regional distribution centre in Singapore, can't get better than that. I've gotten motherboard and GPU replacements within a day of calling support.

 

Fully understandable.  I've thought about this too.  I'd probably keep a spare (low-end) motherboard around in case I need to swap that around.  I'm not sure what else could break that would be a total emergency.  And I'm not looking to switch as soon as Framework 16 comes out.  I'll be watching them for a while to see how generational upgrades are going, while I ride out my time on the 7770.  But yeah, being in the U.S. is probably best-case in terms of getting parts quickly from them.

 

(Though, my experience with Dell warranty parts replacement hasn't been great lately.  I waited almost a month to have my Precision 7770 display replaced, and about three weeks for the heatsink/cooler.  I'm no longer considering "next business day" to be a thing that I can count on.)

 

9 hours ago, Etern4l said:

Apple cleverly side-stepped much of this by switching to power-efficient ARM/RISC CPUs

 

I really do wish someone else would put out a high-performance laptop-first power-efficient CPU.  Intel's mobile CPUs are basically repackaged desktop dies with the power limits dialed down a little bit.

[Etern4l]

Intel CPUs tend to be massively overvolted out of the box. I'm not sure why that is. Poor manufacturing tolerances to increase yield, which they compensate for with voltage? Tight reliability requirements??The race with AMD is partially a blessing and partially to blame, as Intel has no choice but to crank up clocks and voltages in order to stay competitive in benchmarks. Whatever the case may be, in practice Intel's high end  CPUs get massively more efficient after undervolting, and even more so if people are willing to sacrifice a ratio point or two and a few % of performance. Sadly, no OEM or mobo manufacturer offers this kind of optimisation out of the box. The video covers 13900K, but surely relevant to laptops as well:

 

 

[Aaron44126]

Sigh.  Case in point.

 

(This game isn't that graphically intense, but I'm usually seeing a pull in the 50-80W range from the GPU.  It's an emulator so it is pretty CPU-heavy.  It had to be running at around 10 FPS for a few moments before I managed to grab the screen shot, I was full screen so the indicator wasn't shown...)

 

Haven't taken the time to do testing on the CPU power limit yet.  Maybe this weekend.

 

@Ionising_Radiation Still good with the NVIDIA platform controller thing disabled?  Have you had time to play any more?

[Ionising_Radiation]

48 minutes ago, Aaron44126 said:

Still good with the NVIDIA platform controller thing disabled?  Have you had time to play any more?

Yep; no drops so far. I've had several games in Infinite now, and I used to see it every other game or so, but it seems to have gone away... 

 

49 minutes ago, Aaron44126 said:

 

Also... That's KDE Plasma! You switched from Ubuntu?