Titan X (Pascal) PCIe 3.0x8

[dueprocessofflaw]

Anyone have some experience running a Titan X Pascal on PCIe 3.0 x8 as opposed to x16 (and any numbers regarding relative performance @2160p-- I know it will work, the question is whether there's a measurable negative impact).

I've done a fair amount of research online, and it seems that, for the most part, there's practically no drop off in fps when testing less powerful cards (i.e. Titan X Maxwell, GTX 1080); however, I've heard rumors of latency issues and GP102 possibly saturating PCIe 3.0, and I can't find any information regarding:
(I) the impact on frame times (not avg fps);
(II) tests utilizing a Titan X Pascal.

I'm running a Titan X Pascal on a Z97 chipset (Asus Maximus VII Impact M-ITX, w. i7-4790k OC'd w. HT to 4.7ghz core @1.22v / 4.4ghz cache @ 1.2v).
Currently the card is set up to use all 16 lanes, but I'd like to get a Samsung 960 nvme m.2 drive, which would occupy 4 lanes of PCIe 3.0 and force me to run my GPU on PCIe 3.0 x8.

I'm hesitant to swap my chipset out for an x99, etc., as:
(I) my CPU appears to be a high percentile chip based on current temps / clock speeds / voltages, that still hasn't hit its ceiling;
(II) new mobo / CPU / RAM gets expensive (only went with Titan as SLI isn't an option for M ITX).

Anyone with a Titan X and access to more advanced benchmarks have any feedback on the effects of running the card on 8 lanes?
I can test FPS myself, but I don't have the software to get a good look at fps spikes, frame times, etc.

FYI: (if anyone runs a comparison) w. custom fan curve, TDP increase, marginal OC, the Titan is sitting at 2,100mhz core and 11,000mhz mem consistently (120% TDP, +225 core / +500 mem).

Thanks!

 

[leo2kp]

I can confirm that the GTX 1080 runs just fine in that configuration (except that I'm running PCI-E 2.0 x16, so it's the same as 3.0 x8). But for your specific card, I can't say for sure but I wouldn't expect there to be any issues.

 

[cub_fanatic]

Since you have the Titan X already, you can find out pretty easily if it saturates PCIe 3.0 x8. Just go into your BIOS and tell it to run the PCIe in gen 2 mode instead of gen 3. That will run your PCIe slot @ PCIe 2.0 x16 which is essentially the same as PCIe 3.0 x8. Get a free benchmarking tool like 3DMark and run the demo version of Firestrike and Time Spy (for DX12) while it is in gen 2 mode. Take note of the scores and the frame rates. Then, reboot and set it back to gen 3 and re-run those same tests again. If there is a significant difference in the scores and frame rates, then you'll at least have a good idea how much performance you may or may not be losing in PCIe 3.0 x8.

 

[dueprocessofflaw]

Does Firestrike let you measure frame times? Based on other reports, fps differences are minimal. I'm more worried about consistent delivery

 

[cub_fanatic]

Does Firestrike let you measure frame times? Based on other reports, fps differences are minimal. I'm more worried about consistent delivery

No. It will just tell you if you are losing any performance due to it running on PCIe 3.0 x8. You'll get a score that you can use to compare to other systems and your own system running the same GPU on PCIe 3.0 x16. To measure frame time, you actually need a frame grabber, a second PC, a second monitor, FCAT software as well as a few other special gadgets. Frame grabbers are physical pieces of hardware. They cost anywhere from $300 to over $1000. Have a look at this article explaining what a typical frame time measuring setup might look like: http://www.guru3d.com/articles-pages/fcat-benchmarking-review,1.html

Basically, it isn't just a piece of software that you can buy and install. It is also mentioned in this article that measuring frametime isn't really necessary on single GPU setups, it is used mostly to test SLI and Crossfire setups...

Now realistically for single graphics cards, measuring frametime (or as what like to call, frame experience) FRAPS is sufficient. However in multi-GPU setups FRAPs just does not detect all information and as such a lot simply does not show up, in the created charts that many websites including ourselves have been demonstrating.

But, if you want to find out frame time on a Pascal Titan X anyway, unfortunately, I doubt that any of the reviewers that have tested a single Pascal Titan X on one of these setups did so on PCIe 3.0 x8 just for the hell of it. So, unless you are willing to invest in probably another grand in hardware just to generate a couple graphs on frame time, your next best option is to just use something like 3DMark or FRAPS while running your favorite demanding game.

Considering what you are trying to do here, that money would be better spent on a better Z97 motherboard that has a PLX chip so that you can get more lanes of PCIe without having to upgrade everything else. Something like the Asus Z97-WS could run your Titan X on PCIe 3.0 x16 and a M.2 SSD at the full PCIe 3.0 x4 and still have 12 lanes of PCIe 3.0 left over.