9600GT SLI beats 3870X2 in a couple games (inq, expreview)

[Sr7]

http://www.theinquirer.net/gb/inquir...bble-25feb2008

Thought this was funny...

http://en.expreview.com/?p=273&page=16

They only tested 2 actual games and 3dmark. X2 still won 3dmark as expected, but in the games it lost by quite a margin.

[0maro]

Awesome results of 9600SLI.It's really good choice for most of us(mid-range) pc's users
Waiting for nVidia's triple SLI boost

[K404]

roughly the same price for the GPUs, but means buying an SLI mobo... I guess no situation is perfect!

[awdrifter]

That's very good results, I can't wait to see what the 9800GX2 can do.

[emlyn_dewar]

Yeah, I'd rather not be tied to an Nvidia chipset. I'd also like to see a site compare more than 2 games for the 9600SLI/3870 X2 test... Not really a good test tbh.

[Sr7]

roughly the same price for the GPUs, but means buying an SLI mobo... I guess no situation is perfect!

Right now on newegg, 2 9600GT's are $360.

A 3870x2 is minimum $419. That's $60 cheaper.

Also, the other point is that originally, with single GPU's, ATI was competing with NVIDIA's high end (XT vs. GTX).. high end vs. high end.

Then they were competing with their middle-high end (2900XT vs. 8800GTS). Now they're competing with their mid-range (3870 vs. 9600GT).

Anyone see a pattern?

[Sr7]

Yeah, I'd rather not be tied to an Nvidia chipset. I'd also like to see a site compare more than 2 games for the 9600SLI/3870 X2 test... Not really a good test tbh.

True, but its interesting considering those (Bioshock and CoD4) are two of the best multi-gpu scaling apps for AMD.

[zanzabar]

dont take that x2 v sli to seriously the knutered the x2 with a p5nt, its not like expreview to do that intentionally but with an x38 or 48 their 3dmark 06 is in the 17k range and not the 15k

[Sr7]

dont take that x2 v sli to seriously the knutered the x2 with a p5nt, its not like expreview to do that intentionally but with an x38 or 48 their 3dmark 06 is in the 17k range and not the 15k

That makes no sense though.. anything that slows down the x2 is gonna slow down the 9600gt's too...

[Eastcoasthandle]

They used a P5NT-Deluxe motherboard specifically designed to run SLI. They used a 780i chipset to benchmark a X2...come on. Ok to even things up only benchmark the GX2 with a X38 and use the same board to compare it to 2 X2. It makes as much sense with this test bed.

#1 The 3dmark score of the X2 is too low which suggests that COD4 and Bioshock results are lower then expected.
#2 Although they used Cat 8.2 it's not the CFX drivers that are supposed to be design for the X2.

[zanzabar]

ur #2 isnt really correct they used 1 card so it dosnt matter if its crossfire x or not they did 2gpu v 2gpu not 2card v 2card

That makes no sense though.. anything that slows down the x2 is gonna slow down the 9600gt's too...

not exactly, the 9600 probly only uses 8x pci-e1 bandwidth (per card) while the x2 uses 16x pci-e2 and can max out the slot so its much more dependent on the buss and having the bus with more lat and less bandwidth than what u would use with it normally will lower its scores by a good chunk, and the 780i has way more latency than teh x38 since it dosnt have pci-e2 on teh NB and since NV limits cards to 8x for non SLI cert (cant be proven per say but a bandwidth test would show it and its been done since the nf4)

[Sr7]

ur #2 isnt really correct they used 1 card so it dosnt matter if its crossfire x or not they did 2gpu v 2gpu not 2card v 2card




not exactly, the 9600 probly only uses 8x pci-e1 bandwidth (per card) while the x2 uses 16x pci-e2 and can max out the slot so its much more dependent on the buss and having the bus with more lat and less bandwidth than what u would use with it normally will lower its scores by a good chunk, and the 780i has way more latency than teh x38 since it dosnt have pci-e2 on teh NB and since NV limits cards to 8x for non SLI cert (cant be proven per say but a bandwidth test would show it and its been done since the nf4)

That's really not true though.. again, the same logic you use can be used against that argument:

If each 9600GT only needs 8x, then each 3870 GPU on the x2 only needs 8x. Therefore 16x is plenty. I thought there were rumors about PCIE2 benefiting the x2 which were proven to be false around launchtime... if this is the case, again, 16x is plenty.

Also, do you really believe the x2 runs at 8x just because it's on a 780i and you assume NVIDIA must be out to cheat?

[jimmyz]

The 3870 x2 is a pci-e 1.0 card not 2.0, the bridge chip is 1.0 no way to make it 2.0 without a 2.0 bridge. even on my x-38 it shows as a 1.0.

[Sr7]

The 3870 x2 is a pci-e 1.0 card not 2.0, the bridge chip is 1.0 no way to make it 2.0 without a 2.0 bridge. even on my x-38 it shows as a 1.0.

Thank you for reminding me.. I knew there was some reason those rumors were proven false, but I couldn't remember what it was. Yes, it's PCIE1.1

[K404]

Right now on newegg, 2 9600GT's are $360.

A 3870x2 is minimum $419. That's $60 cheaper.

Also, the other point is that originally, with single GPU's, ATI was competing with NVIDIA's high end (XT vs. GTX).. high end vs. high end.

Then they were competing with their middle-high end (2900XT vs. 8800GTS). Now they're competing with their mid-range (3870 vs. 9600GT).

Anyone see a pattern?

Well in the UK the 3870X2 is £235 and the 9600GT is £116

What pattern are you referring to? It looks like your stating that ATI are competing with nV on all levels, even if they arent winning on all fronts. Have I missed your point?

[Eastcoasthandle]

Not sure why some have to argue, the 3dmark scores on the X2 are lower then normal. This alone lets anyone know that something is wrong with the results of the X2 on both COD4 and Bioshock. Besides, they used a SLI motherboard.

[Sr7]

Well in the UK the 3870X2 is £235 and the 9600GT is £116

What pattern are you referring to? It looks like your stating that ATI are competing with nV on all levels, even if they arent winning on all fronts. Have I missed your point?

I'm saying that with each generation, their "high end" single GPU solution (the best their engineering can offer) is pit against a lower and lower SKU that NVIDIA has. First they fought at the high end. They couldn't compete more recently, so they fought the mid-high end. Now they're fighting the mid-range.

[Sr7]

Not sure why some have to argue, the 3dmark scores on the X2 are lower then normal. This alone lets anyone know that something is wrong with the results of the X2 on both COD4 and Bioshock. Besides, they used a SLI motherboard.

So whatever is *wrong* is only affecting ATI?

Logic Fail!

What's wrong with using an SLI motherboard? Does the chipset have an "Auto-slowdown-ATI" feature in the BIOS?

[emlyn_dewar]

I would have thought the severly lacklustre variety of games would also have raised questions.

What's wrong with using an SLI motherboard? Does the chipset have an "Auto-slowdown-ATI" feature in the BIOS?

*Dons foil hat*

[ATI-RaWpOwEr]

That makes no sense though.. anything that slows down the x2 is gonna slow down the 9600gt's too...

dude the x2 on x38 which has an intel chipset performs better than the p5nt which has an nvidia chipset its the same for the 9600gt only the other way around

[Sr7]

dude the x2 on x38 which has an intel chipset performs better than the p5nt which has an nvidia chipset its the same for the 9600gt only the other way around

So you're saying that changing *only* chipsets gets you an extra 2-3k in 3dmark06 on ATI?

Hmm.. that's news to me.

If this were a common problem, you wouldn't see sites benching on NVIDIA chipsets...

[K404]

I dont think its fair to say the 3870X2 is fighting nVidias mid-range, not single card to single card anyway.

If we try to ignore the fact it took ATI a LONG time to compete, with the X2 there is NO REAL reason to buy an 8800GTX or Ultra for gaming- the X2 wins at the same price (GTX,) or lower (ultra.) Most gamers will consider performance for the money spent as well.

Think this is heading OT though

[AbelJemka]

The fact the X2 score are too low is questionning the veracity of this test.
Moreover 2 benchmarks prove what? I can do the same, benchmark Oblivion and HL2 and say the X2 is better than 9600GT SLI...

[Eastcoasthandle]

So whatever is *wrong* is only affecting ATI?

Logic Fail!

What's wrong with using an SLI motherboard? Does the chipset have an "Auto-slowdown-ATI" feature in the BIOS?

1. You don't use a SLI board to bench a ati card, common sense tells you that.
2. As you can see, the results are much lower then what others have reported using X38 and other motherboards
3. There are only 2 games benchmark using an X2 vs 9600 SLI. They didn't show any results using QW:ET, HL2: EP2, UT3, Crysis, Call of Juareez, WiC and Lost Planet which was also benchmarked.
You really aren't making much of a point here and looking silly arguing it. Specially when 3870 CF can beat a X2 in a few games.

[ATI-RaWpOwEr]

So whatever is *wrong* is only affecting ATI?

Logic Fail!

What's wrong with using an SLI motherboard? Does the chipset have an "Auto-slowdown-ATI" feature in the BIOS?

the ati cards are more efficient on intel chipsets than nvidia ones thats just how things are