Radeon HD 6990 Bandwidth Comparison Test 16x vs 8x vs 4x 3DMark 11 Linus Tech Tips

Linus Tech Tips ·Linus Tech Tips ·2011-05-08 · 826 words · ~4 min read
Floatplane YouTube

Transcript

JSON SRT VTT 68
0:00 So, as a follow-up to my recent video where I took two GTX 580s, threw them on
0:05 the Gigabyte G1 Gorilla motherboard, and I tested in dual 16x mode in SLI and
0:12 dual 8x bandwidth mode in SLI. I am
0:15 going to be taking an even higher an even more demanding card, the Radeon
0:21 HD6990, and I'm going to be running it in a single card configuration in a 16x
0:25 slot, an 8X slot, and finally a 4X slot
0:30 to find out what difference there is. In terms of performance, I'll be using the
0:34 3D Mark 11 extreme preset. Now, please
0:37 note that the extreme preset will not show as much of a difference as one of
0:42 the lower quality presets, uh, such as
0:45 the entry or performance presets. But what it does give us is a more realistic
0:51 real world scenario because most people are not going to be using a 6990 to run
0:57 Quake 3 at 1024 x 768. So, you're going
1:01 to be turning up the details, you're going to be turning up the res. So, that is the scenario that we will be looking
1:05 at. Something to bear in mind before I get too far into my testing here is that
1:10 the top slot at 16 and the second slot at 8X are both running off the PCI
1:15 Express controller integrated into the Sandy Bridge CPU on this board which is
1:20 clocked at 4.7 GHz by the way to eliminate any CPU bottlenecking or as
1:24 much as possible. The bottom slot actually runs off of the chipset at 4x
1:31 speed. So, it has an additional disadvantage that is not shared by the
1:36 top two slots. So, just bear that in mind. Take the results with a bit of a
1:39 grain of salt, but I don't have another platform handy where I can show you 16x,
1:43 8x, and 4x where they will all be equal.
1:47 So, my 16x result is a score of
1:51 3,330. So, I'll be back with 8x and 4x
1:55 results shortly. Now, this may look like a stupid configuration. I've got my 6T90
2:00 and a GTX 580 here. But the reason I've installed the GTX 580, which won't
2:04 affect my benchmark scores in any way, is to force these PCIe lanes to go from
2:08 16x to 8x8x. So, you can see here within
2:12 GPUZ, the 6990 reports as running at
2:16 16x. And while that may be true because
2:19 it's running internal Crossfire where each GPU is getting 16x bandwidth from
2:24 the uh from the lane spplitter that's on the PCB
2:29 itself. The actual lane splitter itself is only
2:33 getting if we look here at the GTX 580.
2:37 So you can see that the actual bandwidth to the slot is only 8x because it is
2:42 correctly split at this time. So my benchmark results are
2:48 3,312 3D marks. So we have seen
2:52 a difference in performance between 16x
2:56 and 8x
2:59 312. Okay, very small, very small. And I
3:03 will be back with PCIe 4X results very
3:08 shortly. So at the end of the day, this video becomes more about how if you
3:13 internally bottleneck a card um
3:17 according to the computational resources
3:20 located on the card that the amount of PCI Express bandwidth you supply it with
3:25 is actually going to make very little difference in terms of performance. Now,
3:30 something to bear in mind too about this particular testing setup today. You can
3:34 see I had to nudge my motherboard over a little bit to get the uh 6T90 to fit
3:39 into that bottom slot on my tech station. But so, I've got it in the PCIe
3:43 4x slot. Now, think about it this way.
3:47 Basically, I've got two GPUs. I have on a 6990, I pretty much have two 6970
3:54 GPUs. Okay? So, so bear with me here.
4:00 So, by the comparison I did the other day where I was comparing
4:05 16x6 and 8x8x SLI, this is basically running it
4:11 on this motherboard, this card in 16X mode is pretty much like running two
4:15 cards, two high-end single GPUs at
4:19 8x8x, 4x4x in the 8x slot, and then at
4:24 2x 2x in that 4x slot at at the very
4:28 bottom that that's essentially the amount of bandwidth we have available
4:31 per GPU and we still see very very
4:36 little fluctuation in terms of the overall score. So, what I may have to do
4:39 in the coming uh next little while is do
4:42 an updated video about this where I find a slightly less uh GPU speed
4:49 bottlenecked benchmark and uh demonstrate maybe more of a more of a
4:54 typical or worst case scenario for how much the PCI Express bandwidth affects
4:58 single card performance. So, thank you for checking out this little episode of
5:01 Linus Tech Tips. Don't forget to subscribe for more unboxings, reviews, and other computer videos.