P45 vs X48 Crossfire test - X48, Wolfdale + 4870 Crossfire user needed!

Ok, I got my 4870 crossfire rig set up as per my signature. Really pleased with it. Anyway, I am looking for somebody with an X48 or x38 board with the same setup as me so we can try to properly test whether there is much difference between 8x\8x or full 16x\16x pci express lanes when running crossfire. Basically the idea is to run some tests with the same settings as each other and see what the results are if anyone is up for it. I know there will be differences because of different boards\chipsets etc anyway but at least we would have some rough idea how P45 crossfire compares to X48.

Let me know if anyone is up for it, in the mean time, a quick test before i go to bed:

3dmark 06 @ default settings (1280x1024):

http://pix.nofrag.com/1/7/d/d53666bf...590dd05f6f.bmp

Not a wolfdale, but gives you an idea

http://www.tweaktown.com/articles/14...nce/index.html

well, i dont have a 4850 or 4870, however i did notice a 4000pt difference in 3dmark01 using 2900pros switching between a maximus (x38) and p5q-e (p45)
the p5q-e is now collecting dust because of this :(

You unfortunately didn't tell us anything about your configuration unless it is precisely as in your sig. From my experience, the difference between 8x 8x PCIe2.0 and 16x 16x PCIe2.0 CrossFire is negligible. Decent 3D06 score for a Wolfdale, I imagine that'd be near if not higher than 24k with a Quad. Impressive SM3.0 score though, these cards eat that portion of the benchmark for lunch :up:

Quote:

---

Originally Posted by NeedMoMegaHurtZ

Not a wolfdale, but gives you an idea

http://www.tweaktown.com/articles/14...nce/index.html

---

That is the exact test I am hoping to disprove, as many people have stated, it does not make any sense.

Quote:

---

Originally Posted by [XC] gomeler

You unfortunately didn't tell us anything about your configuration unless it is precisely as in your sig. From my experience, the difference between 8x 8x PCIe2.0 and 16x 16x PCIe2.0 CrossFire is negligible. Decent 3D06 score for a Wolfdale, I imagine that'd be near if not higher than 24k with a Quad. Impressive SM3.0 score though, these cards eat that portion of the benchmark for lunch :up:

---

Yes - setup exactly as in my sig mate. Will change it to match other setups if anyone has got any tests to compare. The test I posted was not really the best to use especially with it being low res, but it was just a quick test, I will get some high res tests done soon.

Hope someone will help testing this. Cant be true Tweaktowns test (whether right or wrong) is going to be the undisputed truth everywhere, without some more testing on the subject.
Even if there is variance between motherboards, which makes testing difficult, some of the results on tweaktown differs way too much from each other to be purely explained by motherboard variance imo.

Quote:

---

Originally Posted by ZenEffect

well, i dont have a 4850 or 4870, however i did notice a 4000pt difference in 3dmark01 using 2900pros switching between a maximus (x38) and p5q-e (p45)
the p5q-e is now collecting dust because of this :(

---

Thats kinda scary. 2900s using up the bandwidth sounds uhm, weird, doesnt it? Maybe something else entirely is bugged in P45 crossfire.

Actual game tests is more interesting than 3dmark though.

Damn, a few days away from buying a P5Q-E, would sleep more easy knowing whats false and true about all this :D

Im one of the people bashing that test, I dont see how the tweak town test can be true, maybe for ultra high end cards like a 9800GX2 I think a 8x 2.0 could bottleneck but no way from a 4850, also there are several other articles out there that prove the tweaktown one wrong.

One that I saved is this one: http://www.bit-tech.net/hardware/200...has-arrived/13 check out the crossfire results vs the other mobo's wich include 16x/16x x38/x48 mobos.

In addition I have done my own testing by manually lowering my slot speeds in bios, I didnt see a 3dmark06 bottleneck until 1x 2.0 speed. (was 2x on a 1.0 mobo my score got cut with a 9600gt, raising it to 4x gave me the full normal score for a 9600gt in 3dmark06)

4,4 e3110 and 675/1075 4850 Xfire
vista 64, 8.6 hotfix drivers, x38 maximus formula with rampage formula bios 0308, 20,336 in 06