M4700 Owners Thread

Have Windows 7 installed on 2ndary SSD in my M4700; Installed MSI afterburner to test out underclocking and thermal reduction.

Does...pretty much nothing as far as I can tell, perhaps a slight delay (maybe 30 seconds) before GPU fan kicks in vs. stock clocks, temps climb up to 57c gpu-fan click-whirrr threshold as always, even when doing nothing but watch the GPU temp pointlessly climb (with CPU steady at 40c).

Conclusion: software solution won't work to reduce thermal temps since Nvidia has forced max performance in vBIOS when multi-monitor setup is detected.

Brute force is therefore required; that or tolerate GPU fan kicking on ever other minute while doing GPU intensive tasks like typing an email, working in a text editor, or staring out the window.

Downloaded nvFlash, GPU-Z, and a recently launched Kepler BIOS mod "Kepler BIOS Tweaker", v1.25 is the latest version available, AFAIK.

Have never flashed video BIOS before, so uncharted waters here.

Here are my stock voltages:
P0 > .8875
P8 > .8375

The minimum voltage allowed via KBT interface is .8250

In my Google searches I have not been able to find a thread that covers the relationship between voltage and core/mem clocks. Since we cannot tweak K1000M/K2000M Kepler card voltage, it's not possible to software benchmark before flashing the vBIOS.

Anyone have some hard numbers to work with re: voltage and clock speeds? Would be nice to have a defined ratio to work with so I can proceed with at least some degree of "confidence" ;-)

 

Chipset
USB 3
Lan
Wifi
Web cam
GPU both the Intel and NVIDIA in that order or ATI
I usually do the sound drivers as well.
Depending on the config you may have to install smart card, Intel Drive, and TPM drivers etc. they usually show up as missing drivers after a windows install.

 

Hello. I recently bought refurbished precision m4700 laptop and it has a strange problem with the screen (panel LG Philips LP156WF3 - IPS 10bit-color). Gradients on the screen spilt into strips and dots (even in windows welcome logo and bios). Changing color gamut in premier color application affect their position in the image, switching to 40hz refresh rate too. Latest quadro drivers are installed. External 30-bit color monitor connected via display port works great. I know about tint problem in this panel (my screen has a very small tint and this isn't critical for me), but this blinking lines and dots are really annoying. I performed the ‘LCD BIST’ (Built in Self-test) (holding "D" button while powering on) but no lines, dots, whatsever during its passage. Bios is updated to the last version. I've tried to boot livecd and got the same problems (gradients are affected most, cursor's shadow for expamle). Is it some software issue or i need to change this panel ? What do you think?

 

I bit the bullet and did the Nvidia vBIOS flash, had no choice, the on/off GPU fan was driving me nuts.

Here's the deal, as per above:
multi-monitor setup = max performance at hardware level = heat = noise. Nvidia has made the decision to enforce max performance mode in multi-monitor setups. No way around it at software level (software downclock attempts in Windows did next to nothing for me), and on Linux we don't even have that option, completely FUBAR'd complements of Nvidia.

If you're fed up, grab latest versions of NVflash, GPU-Z, and KBT (Kepler BIOS Tweaker). You'll need a bootable DOS-enabled USB key, vBIOS extracted from GPU-Z, modified vBIOS from KBT, and NVflash to write the new vBIOS.

First time doing this so I took a conservative approach. In KBT I set performance (P0) clocks and voltage to match that of powersave (P8) across the board (i.e. in each tab). Once done, save the modified vBIOS.

NVflash guide here:
[GUIDE] NVIDIA VBIOS flashing

After flashing the vBIOS the silence is disturbing, in a very good way -- all together now, shut up Nvidia ;-) (Linus Torvolds had other choice words for them).

Clocks are now pinned across the board on my K1000M at 405MHz, vs 850MHz+ previously. This might not fly if you're a gamer, perfect for me, I just need the card to drive multiple displays without the laptop GPU fan attempting to achieve liftoff every minute.

GPU temps now hover around 50c, well below gpu-fan-blast 57c threshold. Am tempted to undervolt and underclock P0 and P8 states further, but will hold off for now all's quiet on the western front :thumbsup:

 

Thanks dave-p. I followed the Dell Re-Image guide too, but found it to be a bit confusing some times, as well as a few drivers missing from Dell's driver's page that were mentioned in the guide... Driver Genius topped it off I guess...

 

Congrats, great work! :thumbsup:
Do you get artefacts with HD video on external display with those clocks?
Does HD video trigger fan cycling?
What happens under CPU load?
I.e. if CPU temp tells the fans to kick in, is the total noise same as it was earlier?
(what I haven't completely understood is on what level the CPU and GPU fans are independent of each other.)

I was in contact with AMD about firepro M4000 and if I understand correctly now it is Dell who designs the cooling and chooses the fan and its RPM etc. It's hard to get good answers from Dell and the manufacturers so that I don't yet understand the fundamental technical problem that the fan control is not trimmed to modulate down to a low RPM instead of on/off cycling.

So if the fan control is not a GPU manufacturer issue and thus cannot be influenced in the GPU driver either (otherwise than heat generation by e.g. underclocking), then the fan control must be on the Dell BIOS level, right?
It is also Dell who programs the Windows power options (Dell utilities) where we get some very limited options to influence the fan operation.

Is it then the number of performance states (P0 and P8) that results in fan on/off cycling with the K1000M?
For PWM operation you actually have to give some lowest on-RPM and if this is set quite high for P8 then this could be the problem(?)

I guess I have to make a decision soon, I was planning to either send back the M4700 for change to firepro M4000 or refund and get another notebook.

One question still:
Does anyone know if the Firepro M4000 with external displays can be tweaked to such values that the fan does not kick in or that it modulates to constant non-disturbing low RPM?
(AMD didn't have an answer as they say it is Dell who does the cooling)

 

No artefacts with HD video, GPU temp maxed at 53c, so fans remained turned off -- haven't heard the fans in 12 hours.

To my knowledge fan control is set by Dell, not Intel/Nvidia/ATI, and is certainly not accessible on Linux end of things -- if you want to test uncharted waters, Google "Dell Precision BIOS engineering mode" (there's a key combo one can use on boot to access low level system details like hard wired fan speeds).

I'm actually not bothered by the fans, particularly when compiling applications from source, where I expect fan noise, lots of it -- am happy to see this beast chew through code that took ages on previous system. The problem is, and it's the one eveyone is bothered by, doing nothing and having the fans turn on-off, on-off, on-off for no apparent reason.

Nvidia has said that it's a hardware limitation; that max performance mode is required to drive multiple displays. Closer to the truth is the easiest thing to do to support all hardware was to require max performance and take it out of the end user's hands.

Based on what I'm seeing so far, you can underclock/undervolt with a multi-display setup and it will work just fine (for high end gaming, no, but you'd be going the other way and overclocking if that was the case). My P0 performance clocks are now 405Mhz/405Mhz vs. 850Mhz/900Mhz, and undervolted from .875 to .837, matching exactly what KBT displayed for P8 state in a dual monitor setup (which is the high end of P8 range, so much for power saving).

It's still unclear to me the relationship between boost state and voltage. I'm going to boot back into Windows today and see if boost state (which was greyed out/fixed to 850Mhz) has changed. Also, at times power state would toggle between P0 and P8 states with stock clocks/voltage; that's not happening now since P0 and P8 states are identical. I'm thinking about setting P8 state to .825 on voltage and 135Mhz/405Mhz to get a further temperature reduction -- could try that with P0 state as well, but then the poor thing might freeze to death

 

Oh, by the way.... I'm not so sure that 57ºC is the threshold. When I've experimented with external fan helping the GPU I've often managed to get the GPU temp cycling between appr. 42...52ºC instead of the 47...57ºC.
This might be because of some other temp sensor which is not visible in software triggering the fan, or there could be a time delay so that the counter starts at say 52ºC but the fan kicks in later after a certain amount of time above this temp, and it just happens that the temp has then reached appr. 57ºC.

You could test this by blocking the exhaust a bit a see where you find a temperature that triggers the fan.

Good to check because otherwise you might still be close to the threshold so that with slightly elevated amb temps the cycling starts again.

 

Good thing about HD video, because with audio/music production you often need to run HD video for audio sync, or reference listening to audio with video etc. Would be quite cumbersome to flash GPU firmware each time to switch...

The CPU thingy is important for audio/music work. You normally have quite high CPU load with near 0% GPU load, while sitting near the computer for mixing or editing. So if the GPU with external displays is per default at P8 state (which it will be even when underclocked) and the CPU triggers BOTH fans, then this will still be annoying as you get the same whirrrrr noise level when the GPU fan starts. Because the underclocking does not change the fan RPM for the P8 state as this is defined in the Dell BIOS.

 

I have a status bar (equivalent of Windows task bar) that I pipe in some system details like CPU and GPU temps. The delay is 3 seconds -- what I've noticed is the GPU fan kicking in at 57c, but the trigger point may in fact be @60c.

Regardless, pre-vBIOS flash the GPU fan was guaranteed to turn on, particluarly after resuming from suspend (for some reason GPU fan would cycle on-off every minute); from cold start I could get a few hours of pure quiet under light load, so the Nvidia chip does at times with stock clocks stay in P8 state. It's a lousy feeling to see the GPU temp steadily rise to 57c trigger point with zero programs open while the CPU remains at 40c; that's a defect, IMO, and is what forced my hand to flash the vBIOS.

Moving forward, re: ambient temps you might be right -- am considering going 135/405 on P8, but for now am going to play it safe with what I have, was quite a relief to see the success message post-vBIOS-flash in DOS prompt