After all that, SLI is a wate of time

FatRakoon · 1 Feb 2007 at 13:56

Been working on knocking up an SLI setup, nothing serious, I just wanted to do it you know.

Anyway, I was half and half knocking up a cheap as chips SLI setup with the cheapest possible cards I can find, but thankfully, I was put off that idea, and I decided to not bother and just get a normal card instead, so I opted for a 6600GT as I dont need it for games, but I also needed to be able to play games should I want to!

Now, the card I got turned out to be an SLI capable Gigabyte 6600GT, then a short while later, I saw the exact same model, and so naturally, with wanting SLI a short while before this, I grabbed it.

Now, I have thrown the cards into an old A8N-SLI, along with a few other bits, and I never bothered a re-install or anything, I just wanted to make sure I could get SLI running first.

Anyway, so far, Im not overly impressed, and this morning I got hold of a proper SLI cable, and so I did a fresh install, and got the setupo all fully running and what did I find?

Its still crap.

And I do mean crap.

I have been using AquaMark3 for the benching, but also playign a couple of games, while displaying their FPS and Im not really finding it much faster at all!

Single 6600GT gives me 57K in AquaMark3 and with both cards in SLI mode, I get 66K??? - WTF? only 10K more... Thats like 20%???

Why only 20%?

Utter waste of time all of this.

Surely I should be getting more?

Anyway, SLI does give better frame rates, dont get me wrong... The Massive Overdraw section in AM3 drops to about 10-12FPS with a single card, and 33FPS in SLI mode, so it does make a difference, but overall, I feel that the people who told me not to bother have been right all along.

I also quickly ( If you can do that ) run 3DMARK03 and I got 12K

Does that sound right for 6600GT SLI?

melbourne720 · 1 Feb 2007 at 14:13

IMHO SLi is only ever worth it if (a) you get a really good price on the additional card or (b) you have a top of the range card (such as an 8800 GTX) and there isn’t any further you can upgrade!

I think you've just proved it.

Energize · 1 Feb 2007 at 14:24

FatRakoon said:
Anyway, SLI does give better frame rates, dont get me wrong... The Massive Overdraw section in AM3 drops to about 10-12FPS with a single card, and 33FPS in SLI mode, so it does make a difference, but overall, I feel that the people who told me not to bother have been right all along.

A 200% performance increase isn't bad. After all benchmarks don't mean anything for real world performance.

Cold Fusion · 1 Feb 2007 at 14:33

Energize said:
A 200% performance increase isn't bad. After all benchmarks don't mean anything for real world performance.

200%

lay-z-boy · 1 Feb 2007 at 14:42

Cold Fusion said:
200%

15fps=100%

30fps=200%

Mr Paul · 1 Feb 2007 at 14:50

I'll swap my X1800XT for your Atari Falcon

FatRakoon · 1 Feb 2007 at 15:08

Energize said:
A 200% performance increase isn't bad. After all benchmarks don't mean anything for real world performance.

It wasnt 200% increase...

It was 20%

200% yeah, that will be cool, and worth the effort... Even 50% would have been worth it, but 20%?

FatRakoon · 1 Feb 2007 at 15:09

Mr Paul said:
I'll swap my X1800XT for your Atari Falcon

A Falcon CT63 for just that?

You're a funny guy.

Energize · 1 Feb 2007 at 15:09

FatRakoon said:
It wasnt 200% increase...

It was 20%

200% yeah, that will be cool, and worth the effort... Even 50% would have been worth it, but 20%?

fatrakoon said:
he Massive Overdraw section in AM3 drops to about 10-12FPS with a single card, and 33FPS in SLI mode, so it does make a difference

100% increase of 11fps is 22fps
200% increase of 11fps is 33fps.

Deks · 1 Feb 2007 at 15:58

Energize said:
100% increase of 11fps is 22fps
200% increase of 11fps is 33fps.

I agree with energize.

A 20% improvement on 11 fps would increase the speed to 13.2 fps !

200% is one heck of a percentage improvement

Deks

jellybeard999 · 1 Feb 2007 at 16:13

Think he's saying 20% (near enough) on the 57k to 66k aquamark score

TBH aquamark is like using 3dmark01 to bench GPU, makes little difference, ATi 9800 to nVidia 7800 with the same CPU gives barely any benefit, but it clearly is a massive step up...

Try a game thats not CPU limited, but grapihics limited with SLI off and then on... something like FEAR (not R6 vegas, as it seems thats just poor coding tbh) and see if theres a benefit

I dont like SLI, but it will be an improvement...

IMO SLI is only worth it on a TOP card, if you NEED the FPS (ie 30" Dell monitor)

If you can manage with 1 card comfortably, its a waste, as you can save your cash, sell the old card, and get the next gen in 6-12months, for a much smaller price, and probably get better performance than the SLI would have been...

That, and benchers who are going for sponsorship etc, casual benchers should just aim for top score within a section (top 9800pro score etc).

FatRakoon · 1 Feb 2007 at 16:15

Oh, I see...

Well, that was one small section.. It was one routine where a sub blew up! - One routine in the whole of the benchmark, and this does not show an across the board average.

The entire benchmark other than that was about the same, just a tiny bit better ... Few areas that I look for, when the camera view circles around this or that, then the frame rate was often just 80FPS when singl card was 75 giving a lot lower, and this seemed to be about the same all the way through.

As I said, that section was the one and only part that was showing a big improvement.

Also I played a round of UT2004 and in the JunkYard ( Assault ) running up to get the powercore, with one card, Im talkign 110FPS which then drops to about 60-70FPS, but dual cards I was at 115-120 from the off and then it dropped to 65-75 meaning that there was next to no improvement at all?

FatRakoon · 1 Feb 2007 at 16:24

jellybeard999 said:
Think he's saying 20% (near enough) on the 57k to 66k aquamark score

Yeah, thats right.

jellybeard999 said:
TBH aquamark is like using 3dmark01 to bench GPU, makes little difference, ATi 9800 to nVidia 7800 with the same CPU gives barely any benefit, but it clearly is a massive step up...

WHAT? Since when?

AquaMark will give a pretty reliable result for a specific card, no matter what the CPU is?

For example, the RAD9800 gives me 49K whether I run it on an XP17, an x2 4400 or a P4 ( Dont have an LGA Board with AGP otherwise, Id happily try it out on a Conroe and the result will still be 49K

I have found this to be the same since I first started using it and its been the same thign until this day.

If im wrong on that, then Im wrong, but I have been using that program since it came out and its been pretty much exactly spot on every time.

jellybeard999 said:
Try a game thats not CPU limited, but grapihics limited with SLI off and then on...

Yes, Im going to go through some games later on, and see where I get benefit and see where I dont.

jellybeard999 said:
If you can manage with 1 card comfortably, its a waste, as you can save your cash, sell the old card, and get the next gen in 6-12months, for a much smaller price, and probably get better performance than the SLI would have been...

Oh I know.

Dont get me wrong, I dont actually need SLI... Its not even on a PC that I am using in any serious way, I only knocked it up purely to play with SLI and no other reason... When I have finished plonking about with it, I will chuck in the Opteron, and my daughter can have it as she is still stuck on a Single core ClawHammer & Rad9800AGP where the boys have got dual core.

Rroff · 1 Feb 2007 at 16:57

sorry you found SLI a waste of time, I don't know if its the architecture but I had similiar issues with a set of 6800 cards in SLI awhile back, at the time I put it down to driver immaturity but I've seen a couple of people recently trying it with 6800 cards and latest drivers without much joy...

However SLI is not a complete waste of time with a couple of 7x00 series cards on a motherboard that supports SLI properly, with the latest drivers. you will get around double your framerate in the large majority of games...
contrary to what I said previously about some games only getting 1.4-1.6x the framerate of a single card - I found my nvapps.xml file was corrupt, replacing it with the proper file I found all those games now run at double framerate.

Wile E. Coyote · 1 Feb 2007 at 18:46

FatRakoon said:
I also quickly ( If you can do that ) run 3DMARK03 and I got 12K

Does that sound right for 6600GT SLI?

That sounds about right if its stock, I got my old desktop with 6600gt in SLI and a 3000 winnie to 14.5k overclocking the cards

Richdog · 1 Feb 2007 at 18:52

FatRakoon said:
Its still crap.

And I do mean crap.

I have been using AquaMark3 for the benching, but also playign a couple of games, while displaying their FPS and Im not really finding it much faster at all!

Single 6600GT gives me 57K in AquaMark3 and with both cards in SLI mode, I get 66K??? - WTF? only 10K more... Thats like 20%???

Why only 20%?

Utter waste of time all of this.

Surely I should be getting more?

I'm sorry... are you actually telling us that you used a 5 year old benchmark that by todays standards looks like absolute carp and isn't remotely going to stress an SLI rig... to test an SLI rig?

Wow... sorry but that's really silly, you should know better than that.

Do the obvious thing and use 3DMARK 2005 and 2006 if you want to stress the GPU's...

kinggost · 1 Feb 2007 at 21:25

Cant say they the best cards to sli

FatRakoon · 1 Feb 2007 at 23:24

Richdog said:
I'm sorry... are you actually telling us that you used a 5 year old benchmark that by todays standards looks like absolute carp and isn't remotely going to stress an SLI rig... to test an SLI rig?

Wow... sorry but that's really silly, you should know better than that.

Do the obvious thing and use 3DMARK 2005 and 2006 if you want to stress the GPU's...

No, Its not just for stressing the GPU's but more of a comparison with other cards.

Its a very quick method that I have found to be perfectly reliable and it gives me a very good comparison of the true speeds and I know where I stand with it.

Sure, I will run newer benches if I wanted to really go for it, but there is nothing wrong at all with using AquaMark, no matter how old it is.

Its a tried and tested benchmark that I find vastly superior to 3DMark.

Richdog · 1 Feb 2007 at 23:34

FatRakoon said:
No, Its not just for stressing the GPU's but more of a comparison with other cards.

Its a very quick method that I have found to be perfectly reliable and it gives me a very good comparison of the true speeds and I know where I stand with it.

Sure, I will run newer benches if I wanted to really go for it, but there is nothing wrong at all with using AquaMark, no matter how old it is.

Its a tried and tested benchmark that I find vastly superior to 3DMark.

To say it is superior to 3DMARk 05/06 for testing GPU horesepower is nonsense... no offence intended.

It's not an accurate test of modern GPU speed by any stretch of the imagination... I am struggling to see how you can't see that for yourself if you use some common sense. It's a 5 year old benchmark with what is now a primitive game engine... and relies on overall system performance a lot.

How you can come to a definitive conclusion based on using SLI on a single old benchmark that isn't optimised for it I have no idea. It's an extremely CPU limited benchmark... test using the proper programs, or preferably ACTUAL GAMES that will utilise it before you deem it "a complete waste".

FirebarUK · 1 Feb 2007 at 23:53

The 6XXX series aren't particularly useful in SLI tbh. Definately 7XXX or 8XXX series for SLI - seems like the tech is better built for it.