Modding Quadro 4000, advice welcome.

Howdy folks,
I've just installed a ProlimaTech MK-13 GPU heatsink onto my Quadro 4000. With a couple of Noctua NF-P12 to keep air flowing, the MK-13's passive sized grill is now actively cooled.

So here's why I did the upgrade while kissing my remaining 1 years worth of warranty goodbye. I hate the fan's sound...it is just the right pitch to make my blown ears pop crackle and ring, same thing happens on jet planes, I figure it was the rock concerts from my youth. When it came time to repair a rattling rear fan (turned out to be a wad of lint) I took the down time to sub with a set Noctuas and mutilate my Quadro 4K to suit my needs.

The Quadro's shell comes off easy enough, several tiny screws and it practically pops off. Come to think of it, that is probably not a good sign. There was a lot of goo on the GPU and contact plate, too much in my opinion, this could have been the problem all along. But being committed, the MK-13 went on without a hitch. Idle temperature went from 58°C to 43­°C once the system was powered up again...not bad at all, but more importantly, blessedly quite.

Now, I've run some CUDA encodings and 3D modelling apps without much a-do but things seem oddly slower (or just quiter?). I am a little weary about testing a FPS game yet since the GPU fan is reading 0 RPM while PWM is sending 36%...so if the card is in limp mode, I don't want to break it.

If ya'll have some advice, I'd be glad to hear from ya'll.

Before then after.


Edited by Saibok - 6/27/12 at 9:57am

I couldn't find an adapter to use with the case's 12V power to run my two 120mm PCI blow fans so I got an mCubed miniNG so far. I had originally wanted to keep it cheapish and run SpeedFan for dynamic fan duty but after 3 crashes with some registry fragmentation, I just wanted something safe. The miniNG has two individual channels of control, both having a thermal probe as well as several different fan spin modes each. That sounded really nice after SpeedFan's psychotic break. Here is the little bugger.

A note if you buy this thing, you might get one without labels. The one I got from the web had a solid black cover that obscured all the nice tags on written on the pcb, odd that. I'm forgetful so labels are a must.

Now the Noctua fans at full blast sounds pretty quiet, barely a breeze really...but I wanted grave yard quite. The Noctuas at 35% on analog power is dead quiet for most productivity work and merely a woosh given some stink and that would be my goal using the miniNG. Of the many settings, I'd settled on "target temperature" mode in analog set to roughly 30°C via a tiny potentiometer of which the packaging includes a philips screw for (nice). Range is CCW 20°C, mid 45’C, full clockwise 70°C.

Rough testing.
Old set up with stock Quadro fan. Very loud, a bit like a thumb sized turbine at full power.
58°C - 70°C
New setup with miniNG and MK-13.
43°C - 53°C
Computer and distant traffic with windows open are about equal and I'm in the suburbs. So I guess I'm fairly pleased. If anyone hates their NVIDIA Quadro 4000's whine and your warranty is up, you can give this a shot. If I had more time, I'd try to rig up a PWM post off the Quadro but fed by the PSU's 12V or something. Maybe some other time.

The Quadro's stock heatsink.


Here is the contact surface of the top heatsink casing. It's sad to see a $700+ card has an inaccessible vent prone to clogging. Judging by the design, using an air can with the thin pipe shoved as far up the exhaust could help...but it wouldn't be thorough enough.


I guess the high temperatures were caused by this blockage. Not why I did the mod but...I should see how I could help those who just want a better way to clean this silly packaging.


So here is the case, I've read in some other forums that the Quadro card should be disassembled and the thermal paste replaced. Unless you like re applying thermal paste and have cpu and ceramic adhesive styles handy, I'd just remove the thin cover plate. There are 8 with one black one that I almost missed...oh tricky camouflaged screw (circled in red). Use a T5 sized 6 pointed star screw.


And there you are, full access to the fins, vents, and fan (not in picture)...for any that stumble on to this post, hope it helps.

@ pm40elys40: Yeah, loved your build man! Since my case's job was to host the Quadro 4000, not much contorting was necessary.

If ya'll want to see necessity being the mother of invention, check out PM's build. His is a true mod and the whole case's work is extraordinary. But staying to the topic, his Quadro 4000 work starts here. Check it out!

You'll need 16 for the VRAM. There are 8 on the front of the board and 8 at the rear that will all need to be cooled. I used Enzotech BCC9 Memory Ramsinks that come in packages of 8.

I didn't buy enough so Prolimatech's junk freebee VRAM sinks had to do until I get more.

By the way, I think forum member pm40elys40 said something about cooling the power capacitors or some such by keeping the OEM fan on the Quadro...since I've got two huge 120mm fans blowing directly onto the whole board, I should be OK. YMMV. Hope that answers your question.

Nah, Accelero looked neat but I wanted something dead silent. If the Artic Cooling folks tell you that Accelero will fit the Quadro, go for it. For me, I wanted control on the fans and had plenty of space to spare. Oh, and I hate the sound of tiny fans. The Accelero has three of the little bastards.

By the by, looks like the Accelero Xtreme comes with a bazillion heat sinks that will be great on the Quadro...but might want to count them to make sure you got enough of the square VRAM ones, you'll need 16 for the Quadro and Accelero packaging has 12 of the right sized ones.
Edited by Saibok - 6/29/12 at 9:00am

Got my Furmark Scores: Fail!

Below is an image of the side by side scores using Furmark 1.9.2 in the 1080p benchmark, first with then without the stock fan installed. When installed (right side) the fan hovered around 2000 RPM during the entire test. I have to conclude no throttling and driver compensation is happening, at least as far as Furmark is concerned. If you guys have other benches I should throw at my rather modest system, please give us a holler as I can easily unplug and replug the stock fan for comparo.


Note: since I don't know how to use HTML in a post, I've put the data in an image to compare side by side. There is so little difference, it almost felt pointless but that is probably the point.

EDIT: added both benchmark sessions together.
Edited by Saibok - 6/29/12 at 8:03pm