Dell Precision 7670 & Dell Precision 7770 owner's thread

[MyPC8MyBrain]

@win32asmguy

confirmed wizard man!  \  o  /

 

 

https://www.3dmark.com/spy/34142631

 

https://www.3dmark.com/3dm/86172546

 

 

[win32asmguy]

33 minutes ago, MyPC8MyBrain said:

@win32asmguy

confirmed wizard man!  \  o  /

 

 

https://www.3dmark.com/spy/34142631

 

https://www.3dmark.com/3dm/86172546

 

 

 

Nice! I guess the next question is if Nvidia will be patching this work around in the next driver release.

 

It sounds like it may be similar to what MSI does with Overboost and Tongfang with their TDP unlocks. It may also be something related to Nvidia allowing higher TDP on the desktop 4090 card too.

[MyPC8MyBrain]

40 minutes ago, win32asmguy said:

I guess the next question is if Nvidia will be patching this work around in the next driver release

is that a dell or nvidia driver implementation?

i plan to try and track the driver itself, maybe there's an inf instruction or something we can edit ourselves,

 

Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\nvpcf

 

[MyPC8MyBrain]

setting above reg key "start" value to 4 achieves the same result without disabling in device manager

(seem to be some NVidia telemetry  service),

Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\nvpcf]
"Start"=dword:00000004

 

to make sure it stays off one can also edit "C:\Windows\INF\oem27.inf" line 49

"StartType      = 3" to "StartType      = 4"

 

[win32asmguy]

9 hours ago, MyPC8MyBrain said:

is that a dell or nvidia driver implementation?

i plan to try and track the driver itself, maybe there's an inf instruction or something we can edit ourselves,

 

I think it is power limits supplied by Dell within Nvidia's range for the product. The driver is supposed to enforce those limits via the PCF component.

 

I think this driver also has other bugs (noticed some artifacting even at stock clocks) so I reverted back to the Dell provided drivers for now as I do not want to damage the GPU.

[MyPC8MyBrain]

what sort of artifact did you see? (in a way you want to see artifacts to find your threshold to back off from)
few frames flickering is not a big deal i have seen that happen when i push memory clock speeds too far,
+1400Mhz is the upper limit here before i start to see some artifacts with higher memory clocks,

maybe the artifacts coming from the +138% power limit increase side?
here i can only push to plus +130% max as the slider stops after,

 

 
 

for the record, power supply during bench is pegged at 240w for the entire duration of the bench,
sometime it spiked slightly over 240w which tells me the 330w brick is now needed to exceed these limits safely and remain stable (running with 100% stock Dell paste still),

 

a small tidbit i noticed MSI native OC feature (top left corner) works fully when NPCF is disabled,

 

below is my best score thus far (i feel no need to further prove my machine),
needless to say this is staying for good here with me https://www.3dmark.com/spy/34146281

 

[MyPC8MyBrain]

 

5 hours ago, win32asmguy said:

The driver is supposed to enforce those limits via the PCF component

a longshot but viable... it is possible that wherever CPU-Z is reading its 115w limit value,
Nvidia PCF is also reading that value for threshold instead of the 150 value set in NVidia system info,

(despite my efforts i was unable to trace that illusive 115.0 W TDP (TGP) value source)

 

 

[win32asmguy]

48 minutes ago, MyPC8MyBrain said:

a longshot but viable... it is possible that wherever CPU-Z is reading its 115w limit value,
Nvidia PCF is also reading that value for threshold instead of the 150 value set in NVidia system info,

(despite my efforts i was unable to trace that illusive 115.0 W TDP (TGP) value source)

 

Before I reverted the 527.56 driver on the 7670 I checked that value and it was 90W. So it definitely represents Base TGP. GPU-Z also has the same value I have found:

 

[MyPC8MyBrain]

my gpu-z advance tab is showing a bunch of weird stats (prob because i am on hybrid mode)
but it also exposes a reg path with corresponding files in "C:\Windows\INF" only oem1.inf and oem35.inf are NVidia's,

 

 

Computer\HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Control\Class\{4d36e968-e325-11ce-bfc1-08002be10318}\0000

[MyPC8MyBrain]

18 minutes ago, MyPC8MyBrain said:

my gpu-z advance tab is showing a bunch of weird stats (prob because i am on hybrid mode)

oops... was on the wrong tab (might be barking at the wrong tree here, 115 is just default  )

[MyPC8MyBrain]

paste this little script in a batch file to easily cycle "NVIDIA Platform Controllers and Framework" ON/OFF

mode con: cols=40 lines=4
@echo off
powershell -Command Get-PnpDevice "ACPI\NVDA0820\NPCF" | findstr /i "error" >nul && (powershell -Command pnputil /enable-device "ACPI\NVDA0820\NPCF" & echo ** NPCF Enabled **) 
if %errorlevel% equ 1 (powershell -Command pnputil /disable-device "ACPI\NVDA0820\NPCF" & echo -= NPCF Disabled =-) 
TIMEOUT 3 > nul

:: in case of DeviceId error or to use with a diffrent device, query for DeviceId with example below 
:: Get-PnpDevice -FriendlyName "*Friendly Name Goes Here*" | ft -Wrap -AutoSize Instanceid

 

[win32asmguy]

18 hours ago, MyPC8MyBrain said:

what sort of artifact did you see? (in a way you want to see artifacts to find your threshold to back off from)
few frames flickering is not a big deal i have seen that happen when i push memory clock speeds too far,
+1400Mhz is the upper limit here before i start to see some artifacts with higher memory clocks,

maybe the artifacts coming from the +138% power limit increase side?
here i can only push to plus +130% max as the slider stops after,

 

The artifact was a random box area would flicker periodically in games, even at default clocks and 90W TDP. On another machine with a 3080Ti I noticed checkerboard patterns which persisted to the desktop even after exiting the game, which did not go away until a reboot.

 

On the 7670, 90W x 138% = 125W vbios cap, hence why the slider goes farther. 

[MyPC8MyBrain]

1 hour ago, win32asmguy said:

The artifact was a random box area would flicker periodically in games

was the flickering white patches in random areas of the screen?
try to logon to ms-dos or a low level screen color stage like repair (hod shift when you click restart) and boot to safe mode option 6, reason for the above is to try and observe these flickering happen outside windows as low color level will induce these, during boot time it is also possible to observe these flickers, if you are able to observe them outside the game or windows it is likely not a driver issue but a hardware one, I had similar issue with one of my recent exchange units.

[Aaron44126]

I have also hit hardcore flickering/artifacting when running with graphics switching off which did persist after a reboot / into BIOS setup screens.  This was with nothing going on to "artificially" push the GPU harder.  I have a video of it that I took with my phone.  I will likely warranty this eventually but for now I am just running with graphics switching on (not causing me any trouble).

[win32asmguy]

49 minutes ago, MyPC8MyBrain said:

was the flickering white patches in random areas of the screen?
try to logon to ms-dos or a low level screen color stage like repair (hod shift when you click restart) and boot to safe mode option 6, reason for the above is to try and observe these flickering happen outside windows as low color level will induce these, during boot time it is also possible to observe these flickers, if you are able to observe them outside the game or windows it is likely not a driver issue but a hardware one, I had similar issue with one of my recent exchange units.

 

I will run it through an ePSA diagnostic test in dedicated mode to see if it occurs. I think in dedicated mode it was always artifacting regardless of driver loaded from Dell or Nvidia.

 

I also noticed the 7670 memory is the 14gbps type instead of 16gbps. Probably does not matter at 125W but interesting to note. I wonder if the gddr6 could be an inferior bin or it's just the same chips at a lower speed.

[MyPC8MyBrain]

1 hour ago, win32asmguy said:

I will run it through an ePSA diagnostic test

i believe the diagnostics test is over a white background, that will make it hard to spot white flickering that's why i suggested to find a low res mode like safe mode where its black and little to no drivers involved until late in the load, the goal is to spot it outside of windows to rule out any driver involvement,

 

in my case out of the box on first boot It was noticeable, it wasn't bad or persistent but occasionally small random small rectangular sections flickered quickly no matter if I was in windows or booting or in safe mode, that was a replacement for the first 7770 unit which got immediately replaced to the unit i have now, 3 rounds of 7770 replacements before finally I got a unit that was behaving "relatively" nice without displaying any major issues out of the box (for the record all 7770 units arrived with bios 1.6.3),

[Jack]

Hey Guys, I ordered an heatsink replacement and it got delivered. It is the Sunon Fan. I want to put it on but I also repasted with Liquid Metal and it works just fine (22000 on cinebench single run) so I want to wait for a while at least until the Liquid Metal dry up. However, dell says I have to send the defect part back in 10 days. I was wandering if anyone have encountered this before? Are there any consequences if I don’t return back. Thank you! 

[jorgeregula]

The method with disabling nvidia platform does not seem to work for an RTX A4500. The power limit slider in afterburner is still greyed out / locked. 😞

 

In GPU-Z I have a similar entry when it comes to power limit, and it also shows as adjustable -99% and +39%. So it should somehow be possible?

[developer]

I'll try it 3 hours later when I'll at home immediately and I have a3000 12gb.

 

Hope this will work because I need that 130w 🙂

 

@jorgeregula do you know whih version of your bios? maybe it is related with it. and nvidia driver is also important I guess.

[Aaron44126]

6 hours ago, Jack said:

I was wandering if anyone have encountered this before? Are there any consequences if I don’t return back. Thank you! 

 

This is normal for warranty claims, they generally want to have the defective part back.  You could potentially be charged for it if you do not return it.  (I've never actually heard of anyone wanting to hang onto it for an extended period of time.....)

[1610ftw]

2 hours ago, jorgeregula said:

The method with disabling nvidia platform does not seem to work for an RTX A4500. The power limit slider in afterburner is still greyed out / locked. 😞

 

In GPU-Z I have a similar entry when it comes to power limit, and it also shows as adjustable -99% and +39%. So it should somehow be possible?

 

You should be able to make the adjustment after installing 527.56.

 

For some reason the drivers for the professional cards do not allow for that adjustment but it should not be an issue to install the one for the regular cards and then you will be good to go to get full 3070 Ti / A4500 performance.

[developer]

Hello everyone,

 

After installing driver 527.56 and disabling Nvidia Platform controllers and framework, I can recah up to 130W with A3000 12gb.

 

Now I can give another chance to 7670 🙂

[Aaron44126]

Great work with pushing the GPU power limit higher.  I added a note about this procedure in the OP.

[jorgeregula]

After installing the geforce drivers instead of professional drivers the power limit indeed unlocks.

 

Setting it to +39% gives 125W GPU power usage instead of 102W on an RTX A4500.

 

Ran an extra 3dmark run, graphics score went from 9700 to 10800, so that's great. However, now the CPU score has dropped from 13800 to 10500, and it seems to be thermally limiting. I guess my only options are undervolting / limiting boost to avoid this?

 

Total system power usage shows as 235W.

[Aaron44126]

10 minutes ago, jorgeregula said:

Ran an extra 3dmark run, graphics score went from 9700 to 10800, so that's great. However, now the CPU score has dropped from 13800 to 10500, and it seems to be thermally limiting.

 

Not surprised that pushing the graphics power level higher also restricts the CPU a bit.  (...This is quite possibly why Dell put the GPU power limit where they did.)  At least you have the option to give more power priority to the GPU when you need to run a more GPU-heavy workload.

 

Undervolting will help out some with CPU performance, but I'd say that would be the case even if you weren't messing with the GPU power levels.