I was Lucky got 3 x79 boards in my house one with dual x79 and another 2 with single x79 they arent bad. The only complaint i got is bad fan pwm controllers but i sorted that with standalone pwm sensors.
That’s only true at stock mate as the k20x has much lower memory clocks as well as lower core clock and less rops depends what you’re doing the difference can be much more the 10 percent especially taking in most k40s do 1150 on stock voltage and the k20x won’t I’ve had three now all stop at 1037 at stock and then memory clock I’ve had that be much more of a limiter on the k20x than the core clock so if you get stuck with one that does move much on the memory it’s just alot slower epically for the price difference usually between the k40 and 20x depending where you are
Tried telling myself the k20 was enough. I should have known better. Ordered an m40…
It’s just like bikes- the correct number of bikes to own is always x=x+1
If you ordered from that guy in Hong Kong in ebay keep in mind he sells the gpu only no back fan and no power cable ; you need to get the power cable your self the tesla cards use a very specfic power cord if you try using the gpu with a normal pci power cable you will short the gpu. Guaranteed White smoke.
1 Like
Is it different than the k20x? The power looks the same to me. And no - I eBayed it from the US.
Edit: oh yeah- it is different. Thanks for the heads up! Looks like you just need to switch some pins around.
Second edit- no, it looks like it more than switch pins around.
Third and final: cable ordered!
Some use PCI-e some use a different connector, I believe it’s a CPU eps but I could be wrong
My k40 uses pci-e
1 Like
Yup when i first got my hands on my A100, i was so confused why the PCI wouldn’t fit properly, and after i realised its EPS, i was thankful i didn’t try jam the connector in and power the card.
1 Like
The m40 one I ordered looks like it marries parts of the 6-pin and 8-pin PSU leads (I think that’s EPS, but I’m not sure). In other weirdness my k20x has connects for both the 6 and 8 pin but I was only using the 6. Had both plugged in at one point and it did nothing. Don’t need both on that one, I think it’s either.
You should have a 2 8 pin pci power adapter to the actual plug that fits On the tesla
1 Like
Ah- right right right it is 2 8-pin to 1 8-pin. And correction- my k20x is using the 8-pin PCIe, not the 6-pin. When I plug the 6 in it makes no difference.
After reading the datasheet it looks like the 6 is only there if the 8 is unable to provide 150W. Might have something to do with older and/or server PSUs.
Tesla k80 a newer seem to all use eps power instead of pcie
You really should have the 6 pin plugged in as these cards do pull 235 watts stock mate a a 8 pin and the slot will get you to 225 which is enough but you won’t be using all of the vrm as each plugged on the gpu connects to different parts of the vrm and just isn’t good for them to run like that
1 Like
Hmm- when I run both there is no noticeable difference. Benchmarked the same. It may just be the k20x. Or me just not knowing WTF I’m doing.
No reason to run ECC, right? Also your clock numbers are the best. This is my first time being able to overclock so I wanted to screw around a lot. Settled on 3006, 956 I believe.
Okay- unsurprisingly I didn’t know WTF I was doing. I had vsync on globally and in Heaven and my old monitor is only 60 Hz. It wasn’t even pulling close to max of just the 8-pin (180W). Turned that off and it was chomping at 215W in FurMark. Also kicked my Heaven score up by A LOT. It all makes sense now. And did all this this without breaching 70C (I mean I did not run FurMark very long…)
Can you tell me step by step what to do to get this Tesla working, I’m gonna try to pass through it through the P620. Right now whenever i plug the Tesla in, on device manager it just has a yellow triangle code 10, but the quadro is working perfectly fine.
Right that makes sense now yeah and you should get 1037 and I think 3100 on the memory should be fine is pretty memory chocked if you run heaven and pause it on a scene and drop the memory back to stock you’ll see a big drop think around 10 to 15 fps if running heaven Max settings
1 Like
Okay so which driver you using
So what tesla you trying to run and for now just to make sure the tesla it working justifiable the quadro in device manager you’ll still get video out no problem just won’t load the drivers in that card