M4700 Owners Thread

GPU is by default at P0, max power state, with multiple displays connected. Am not sure if CPU load triggers both fans, but don't care really, the fan noise doesn't bother me when the machine is being put to use/under heavy load. What is agonizing is hearing the fans cycle on-off when doing light load tasks, or when doing nothing at all -- in these scenarios I feel like chucking the machine out the window.

Thinkpad users have it worse with the tighter chassis and single fan -- Google "thinkpad fan noise unbearable"

Dell Precision series seems to offer the best mix of portability, performance, and cooling capacity. i7 power management is not an issue, works flawlessly as far as I can tell; it's the Nvidia chip that's the cause of all the idle noise.

 

Bad news for Windows users, just booted into Windows 7 on 2ndary drive and, while an improvement, my vBIOS tweaks only delay the inevitable steady march to 57c fan trigger point -- light load usage results in fan trigger once every 2 to 3 minutes vs. every minute, better, but problem not at all solved.

For kicks, I underclocked core clock to 270Mhz with MSI Afterburner and that delayed the inevitable another 30 seconds, and then click-whirrrr fan cycle all over again.

Not sure why on Linux the GPU stays at or below 50c under light load, but am not complaining, have suffered plenty with Nvidia chips and fan noise since switching to Linux 3 years ago.

Anyway, if you do take the vBIOS route, obviously backup existing vBIOS first so you can revert back to working version in the event that you brick the chip with overly aggressive clock settings (thankfully we have integrated Intel chip to save the day if/when the *hit hits the fan). Make sure Optimus is turned OFF in the BIOS, don't want to flash the Intel chip by accident ;-)

You're probably going to have to play how-low-can-you-go limbo with KBT clock settings if you want a silent machine under Windows with multi-monitor setup. I'm at 405/405 .8375 with P0 state and that doesn't quite do the job. KBT allows undervolt one more notch to .8250, so if MSI Afterburner can underclock to 270Mhz at .8375, should be able to "safely" go with 270/405 .8250 in KBT -- getting all the other settings to match up is the issue, not sure what ratio of clock speed to voltage is in the 10-odd fields to fill out

 

One thing I noticed was that some tweak/monitoring software force on either a higher performance state or stock GPU clock.

I can't remember exactly, but I think SiSoft Sandra was overriding the MSI driver level underclock and forced stock clock when I was monitoring temperatures with it. I think some other monitoring software did the same thing.
(yes, just tested -> starting Sandra hardware monitor makes Core clock jump from 465MHz to 670MHz overriding the MSI underclocking, and does not fall back until Sandra is turned off completely, not just the monitoring app)

Another thing I vaguely remember is that the MSI underclocking is not fully consistent. It certainly underclocks directly, but if you reboot and start MSI monitor then you seem to have stock core clock until MSI has fully started, and after a few seconds it drops to the underclock level. But the memory is underclocked from the start (?)
This happens even when using "Apply overclocking at system startup" and then reboot. Dunno if this is just a bug in the monitoring software, or if they are the actual clocks you see.
EDIT: Seems like this is a monitoring thing, normal behavior, GPU-Z is also showing the same. I.e. first showing stock GPU clock which then drops directly to underclock.

Also, the Windows power profile settings have the ability to override some settings, e.g. forcing a higher performance state for avoiding latency when throttling. But I doubt they would override a vBIOS underclock...

 

I don't understand why in Windows the GPU climbs to 57c fan threshold over and over and over despite the vBIOS mod. The march to 57c is slower than before the underclock, but it happens regardless, and cyclically, so even if all applications are closed with a 99% idle system, the GPU heat-up-click-whirrrr-cool-down cycle will occur repeatedly -- noise for nothing, how pointless.

Hopefully the underclock continues to keep the machine quiet on the Linux side of the fence, but am worried about when ambient temps rise (within 7 degrees of the dreaded 57c as it stands, summer is still a couple of months away).

 

Knew I saw this somewhere, the GPU 57c fan threshold problem, right here in M4700 thread a couple of months ago:
http://forum.notebookreview.com/dell-latitude-vostro-precision/682758-m4700-owners-thread-60.html#post9069034

Am curious to know if any Windows users with multi-monitor setups have been able to get around the fan cycling problem via software underclocking alone.

 

Think I found part of the problem...
If this is not a Sandra bug, there actually seems to be two temperature sensors on the GPU card: GPU temp and Board temp

You can see these in SiSoft Sandra -> Tools -> Monitor Environment -> 2nd Temperature tab.

What MSI and GPU-Z are reporting as GPU temp is what Sandra is reporting as Board temp. And the other temp is reported as GPU1 temp. And even if the Board temp is low, it seems that the GPU temp is still rising and triggering the GPU fan.

When the Board temp is still under 50ºC the GPU1 temp is over 60ºC. It seems that the GPU fan is triggered when the GPU1 temp reaches 66-68ºC.

This is probably also the reason I can get apparenty different temperature ranges for fan cycling in MSI/GPUz when playing with external fan. Currently MSI/GPUZ is showing cycling between 41...47ºC, but at the same time Sandra shows GPU1 temperature goes between 49º...68ºC.

EDIT: And if this is the case, then it may be that your Linux drivers do not see this other temperature if the fan control does not react on it?! Try to check that out some way in Linux as it might fry something on the GPU board.

 

Well using the LVDS cable for FHD screen Dell part # NV9R0 and the Samsung LTN156HT01 Display upgraded my display to FHD.

Screen color is quite good.

Also did a re-paste with AS5 to the GPU and CPU while things were opened up.

Will post some temps in a bit once things have settled.

 

57c is the fan threshold trigger point, whether that's accurate or not I don't know, but it should be, nvidia-smi let's you query the GPU temp directly, so assume that that number is not wildly off the mark (i.e. don't smell anything burning here). Displayed max temp in both Linux and Windows during light load is 57c, that's the magic number it seems.

The reality is that the machine gets noisy over & over & over with nothing running in a multi-monitor setup, particularly on Windows despite underclocking at hardware level, and I assume it's worse without the vBIOS mod on K1000M/K2000M (which have very limited software under/overclock capabilities).

I think power management between motherboard BIOS, OS, and Nvidia chip has problems -- have seen this on Linux where after a suspend/resume the GPU gets caught in gpu-heat-up-fan-on-cool-down loop, endlessly, but from a cold system start GPU can stay under 57c all day long (i.e. fans never turn on).

Unlikely that Dell or Nvidia will do anything about it the problem goes beyond Nvidia's forced power mode for multi-monitor setups.

 

Nice, please do post your temps when ready, my repaste had marginal results. On that note, were your CPU and GPU heatsinks single or dual pipe?

 

mine were single pipe

is there a dual pipe option ?