![]() ![]() But those are for dedicated gaming and I wouldn't run Plex on this servers (to ensure Plex streaming won't be interrupted by people playing games or needing to reboot a Windows machine). For gaming I have dedicated PCs (one with GTX 1080 and another with RTX 2080). It was fairly easy to get the card working with Passthrough. 2 stream limit of Nvidia's consumer line, not having to fool the VM into thinking the VM isn't a hypervisor, and not having to deal with connecting a monitor or dummy monitor. Stability in a server that is shared with many other VMs is also a high priority for me, so it was also appealing not having to do a driver hack to exceed the max. Yes, price is more than a consumer card, but other considerations (1 slot, lower power, etc.) were a bigger factor in selecting this card. That's a big difference for my general purpose ESXi server that's not a dedicated game machine. ![]() By comparison, the 1060 max power is 120W. ![]() So we have practically more than a factor 3 increase in FP64 performance/price ratio. Notice the Titan Black was from the Kepler generation an hasn't been on sale for a long time. Max power consumption of the P2200 is 75W (from spec) and 73W (as tested by STH). Machine learning performance (125 FP16 TFLOPS) is extremely impressive, but the 6,9 FP64 (double-precision) TFLOPS are also a huge deal. I only had 1 slot available, and I wanted to keep power usage low and the system cool. The server where my Plex VM runs is a 24x7 server that runs many other VMs. I could have tested more streams, but was running out of clients and interest in testing since I can see the GPU was doing its intended work. I tested with a little more than half a dozen simultaneous transcodes, and CPU utilization was very low. The newer P2200 is equal or better than the P2000 in various metrics. You could buy one of those Eurocom MXM card + adapters - but you'll end up paying double and more for something more kludgey. I hope there is a vendor out there who would put a Geforce GTX 1050 mobile on a PCIe x16 SFF card (like what happened almost 20 years ago with S3 SavageMX laptop GPUs that found its way into budget gaming cards) - this Quadro P2200 looks like a 1050 Mobile with some tweaked drivers and some nVidia artificial restriction ridiculousness lifted. But then it's just a shopping exercise, not an engineering challenge. It's so easy to say that if power/slot count isn't a factor, there are tons of things that would do a better job and cheaper. I don't see any AMD cards out there that will come close to the performance of this card, and only use up 75 Watts TDP. I mean, just look at what you couldn't do with a Radeon Pro WX4100. Click to expand.No, you don't need a Quadro per-se - but I would not discount buying this particular Quadro - for an application in a very specific thermal/size envelope (single slot card that fits in a small form factor, with a thin fan and doesn't require an external power connector), this is actually rather competitive. Based on 1,978 user benchmarks for the Nvidia Quadro K6000 and the Quadro P6000, we rank them both on effective speed and value for money against the best 701 GPUs. ![]()
0 Comments
Leave a Reply. |