PCI-e x16, x8 and x4 tests

[theoldtimer]

I ran some PCI-e tests using the Heaven Benchmark a few months ago with my 4870x2’s and was surprised by some of the results (PCI-e x16 compared to x8). I thought I would expand on the test using a better GFX card plus test PCI-e x4.

After running these tests I’m not quite sure what to think. I would really like for someone with a 5970 and a CHIVF to run these tests to compare with my results. Maybe my results are flawed?

Each test was run three times then the average was calculated and used for the result. The PCI-e slots used for the tests were slot 1=X16, slot 2=X8 and Slot 4=X4. PCI-e slot 1 being the slot closest to the processor socket. I added 3DMark06 and Vantage tests in post #10.

System spec.
AMD Thuban 1090T@ 4.2GHz 14x300 (no special reason for these settings)
MB= ASUS Crosshair IV Formula bios 801
MEM= Corsair GT 2000 CL8 4GB
GFX= ASUS 5970@923/1203 MHz (air cooling)
Windows 7 64-bit
Chilled water cooling (CPU only)

I wasn’t trying to get the highest score in the test I was only trying to test and compare the PCI-e x16, x8 and x4 under a reasonable load. I didn’t use GPU-z in the SS’s because my system becomes unstable when I use GPU-z and two CPU-Z’s?



PCI-e x16


PCI-e x8

PCI-e x4


Edited by the oldtimer to replace the incorrect x8 SS with the correct x8 SS

[BeepBeep2]

Definately something wrong there...

[theoldtimer]

Definately something wrong there...

But what is wrong? I figured the x8 might score close to the x16 but I thought the x4 score would fall off a bunch. Oh well maybe someone with a 5970 and CHIVF will run this test and compare?

I have 3D06 and Vantage PCI-e tests that I'll post maybe tomorrow.

[BeepBeep2]

Your screenshots for the x8 and x4 Heaven tests...

x8 has 61.4 FPS and x4 has 70.2

[theoldtimer]

Your screenshots for the x8 and x4 Heaven tests...

x8 has 61.4 FPS and x4 has 70.2

Somehow that SS of the x8 ended up in the wrong place. I had one series of x8 tests earlier in which the second GPU wasn't clocked up during the tests, that x8 being one of them. All the scores in the graph are correct though.

[Leeghoofd]

Why are you shocked beep ? CPU bottleneck ?... we have seen differences between 8X and 16X with GTX 480 on Gulftown when benching... though I tink for gaming it will be pretty similar.

[FlanK3r]

and how it look in real games or benchmarks (AVP benchmark, Crysis bench etc) ? Thx

[theoldtimer]

Why are you shocked beep ? CPU bottleneck ?... we have seen differences between 8X and 16X with GTX 480 on Gulftown when benching... though I tink for gaming it will be pretty similar.

I didn't know what to expect going in but as I viewed the results I began rechecking my hardware then rerunning the tests. I kinda had a feeling that the x8 test scores may be close to the x16's (less than 1 FPS average difference) but the x4 only behind by 4 FPS, I wasn't expecting that. With the exception of the resolution the Benchmark setting are pretty much maxed out.

and how it look in real games or benchmarks (AVP benchmark, Crysis bench etc) ? Thx

I ran some 3D06 and Vantage tests but I haven't made it to any game tests yet. I'll try to get the 06 and vantage tests together sometime today then post them here.

[CobraXP]

@theoldtimer - nice testing. Here is an article I saw awhile back
http://www.techpowerup.com/reviews/A...xpress_Scaling

I have a crossfire setup that is 16x/4x I do not have any weird issues while playing CoD4, L4D 1 and 2, Trackmania, MW2 or most other games I play that support Crossfire.

[theoldtimer]

Thanks CobraXP, yeah techpowerup ran a nice test of the PCI-e 2.0 X16, x8, x4 and x1.

I’m more interested in X16 compared to X8 and X4 in this test. Plus I’m testing with more GFX power.

I did finally get enough variation in the GFX scores between the x16, x8 and x4 2.0 slots to compare in a graph. 3DMark06 didn’t really show much of a difference in the GFX scores @default or max settings. 3DMark Vantage did show a difference in the GFX score both in the Performance and custom settings. Of course different systems hardware or clocks may change the results so your result may vary?

Each test was run at least three times (some four or five times) then each set of scores was averaged. That averaged score was entered into the graph.

System spec.
AMD Thuban 1090T@ 4.2GHz 14x300 (no special reason for these settings)
MB= ASUS Crosshair IV Formula bios 801
MEM= Corsair GT 2000 CL8 4GB
GFX= ASUS 5970@923/1203 MHz (air cooling)
Windows 7 64-bit
Chilled water cooling (CPU only)

This SS shows the system clocks used in 3DMark06 and Vantage.



SS of 3DMark06 custom settings


3DMark06 graph @ the above settings showing no real difference in the PCI-e tests



SS of the custom settings used in Vantage



Vantage graph @the above settings and @ Performance settings, there is a noticeable difference between the scores based on the PCI-e slot tested (x16, x8 or x4). Blue= Custom settings, Red= Performance settings

[slaveondope]

Looks good to me.
I loss ~400 gpu points in Vantage on one 5850 from 16x to 8x

[FlanK3r]

not ut much diference, 8x link looks ok for using

[theoldtimer]

Looks good to me.
I loss ~400 gpu points in Vantage on one 5850 from 16x to 8x

I'll have to test one 5870 to see how it does but I wouldn't expect much if any difference between the x16 and x8 using a single GPU.

not ut much diference, 8x link looks ok for using

I don't have any concerns about using the x8 for a single GPU card that I own.

[crazydiamond]

my 25,3xx single 5850 run was on 8x, i can run it on 16x but dont think there will be any difference at all

[theoldtimer]

If you guys want to add any PCI-e comparision test results to this thread it's fine with me. Fact is I would welcome some result to compare, games or benchmarks.

[CobraXP]

@ t.o.t. - way to go bro. ty for taking the time to do all that testing. Very interesting Vantage scores.

[theoldtimer]

@ t.o.t. - way to go bro. ty for taking the time to do all that testing. Very interesting Vantage scores.

Yeah, I have many more test results that I didn't post. I tried to make sure I was getting consistent results that are repeatable (within reason). Then post only those results.