Nvidia took another stab at Quad SLI with its GeForce 9800 GX2. Can two 9800 GX2s beat three 8800 Ultras? We compare two high-end gaming systems to find out.
Quad SLI Vs. 3-Way SLI : Read more
Quad SLI Vs. 3-Way SLI : Read more
- Status
neodude007
Distinguished
- May 25, 2008
- 125
- 0
- 18,680
You need to test with newer games, nobody cares about Serious Sam, Prey, or Quake 4 - we want to see Crysis, UT3, CoD4, and other NEW games being benchmarked!
bloodymaze
Distinguished
- May 29, 2008
- 83
- 0
- 18,630
USE THE SAME OR SIMILAR COMPONENTS IF MAKING A COMPARISON!
This is very dumb to compare the two when you have:
QUAD: -- Two 1 TB in raid0
TRI: -- MTRON SSD HD
-- OF COURSE THE SSD IS GOING TO WHIP THE CRAP OUT OF THE CLASSIC HD.
This is very dumb to compare the two when you have:
QUAD: -- Two 1 TB in raid0
TRI: -- MTRON SSD HD
-- OF COURSE THE SSD IS GOING TO WHIP THE CRAP OUT OF THE CLASSIC HD.
spearhead
Distinguished
- Apr 22, 2008
- 120
- 0
- 18,680
yes drivers seem to be crap indeed. they should fix the drivers instead of introducing a product which isnt finished yet. both nvidia and ati need to learn from that. either make cards with drivers which gain an acceptable margin running either in crossfire or sli or just forget the whole thing and use those man hours of your technical personel on improving your next gen card instead so it wont have heat and power usage problems with its first gen. yes crossfire and sli can be good. but only if the drivers let it.
This is a truly ridiculous comparison - the two computers have almost nothing in common - you're comparing:
4ghz CPU vs 4.17ghz CPU
12MB L2 cache vs 8MB l2 cache
HDD vs SSD
790i vs 780i
DDR3 vs DDR2
Which makes any comparison of the SLI setup entirely meaningless, which is ignoring the biggest factor of all - the two SLI systems are completely different, you're comparing 3 G80s with higher memory bandwidth against 4 G92s with a much lower memory bandwidth, even if their core frequencies are higher. If you want to do an accurate comparison, use identical hardware for everything else, and compare 2 9800GX2's with 3 8800GT 512MBs, or perhaps 3 8800GTS 512MBs. Unfortunately it is impossible to match them exactly, as the 8800GTS has the same number of stream processors, but more texture units, whereas the 8800GT has an equal number of texture units, but fewer stream processors.
To sum up, this seems like it could be far more appropriately titled, it clearly has nothing to do with 4-way vs 3-way SLI, and is merely a comparison of two pre-built high performance systems.
4ghz CPU vs 4.17ghz CPU
12MB L2 cache vs 8MB l2 cache
HDD vs SSD
790i vs 780i
DDR3 vs DDR2
Which makes any comparison of the SLI setup entirely meaningless, which is ignoring the biggest factor of all - the two SLI systems are completely different, you're comparing 3 G80s with higher memory bandwidth against 4 G92s with a much lower memory bandwidth, even if their core frequencies are higher. If you want to do an accurate comparison, use identical hardware for everything else, and compare 2 9800GX2's with 3 8800GT 512MBs, or perhaps 3 8800GTS 512MBs. Unfortunately it is impossible to match them exactly, as the 8800GTS has the same number of stream processors, but more texture units, whereas the 8800GT has an equal number of texture units, but fewer stream processors.
To sum up, this seems like it could be far more appropriately titled, it clearly has nothing to do with 4-way vs 3-way SLI, and is merely a comparison of two pre-built high performance systems.
[citation][nom]lamno[/nom]You need to test with newer games, nobody cares about Serious Sam, Prey, or Quake 4 - we want to see Crysis, UT3, CoD4, and other NEW games being benchmarked![/citation]
I agree....Crysis is probably the only game right now where a reasonable rig might need 3-4 video cards to run it smoothly at max detail....to omit it from benchmarks is rather silly IMHO.
I agree....Crysis is probably the only game right now where a reasonable rig might need 3-4 video cards to run it smoothly at max detail....to omit it from benchmarks is rather silly IMHO.
hughyhunter
Distinguished
- Nov 20, 2007
- 865
- 0
- 18,980
seems like the 3dmark06 scores are pretty low huh? I score in the 19k with only two 8800GT... of course the CPU is mamothly overclocked! 
Similar components please... different chips, clocked at different speeds, different motherboards, different chipsets, different harddrives, 2x9800GX2 vs 3x8800Ultra????!!! WTF.
THG... your reviews are usually good but... this is awful.
THG... your reviews are usually good but... this is awful.
wh3resmycar
Distinguished
- Sep 12, 2007
- 2,508
- 3
- 21,015
yeah, i wonder why toms still hasnt use cod4,ut3,crysis as a standard for 3d benchmarks. are they waiting for cod5? ut4? crysis:warhead? lol
razor512
Distinguished
- Jun 16, 2007
- 2,135
- 71
- 19,890
Everyone should just get rid of sli,
SLI is the most inefficient invention in computer history.
the more cards you add, the lower the performance increase becomes.
when you do 2 cards in sli, you get about a 30-40% improvement oner a single card,
when you do 3 day SLI, you only get about a 5% improvement
the performance overhead of SLI exponentially increases as more cards are added so your getting less and less performance per dollar out of your investment
the reason why videocard companies are doing this s because people keep buying the crap. if you accept crap then you will get crap
when you compare the speed improvement when going from a geforce 5900, to a geforce 6800, the improvement is like 3-4 times more
when you go from a 6800 to a 7800, the improvement is about 1.5 times faster
when each new card, the level of improvement becomes more and more minute
and instead of making better cards, nvidia and ati are seeing that they can make more money by selling 2 slow expensive cards in CF or SLI, instead of making 1 fast card
SLI is just too inefficient, while it helps, it is smarter to just buy 1 card, then wait a month or 2 then but the next gen card that would be faster than of you wet SLI now and it would have cost less
it is like if you get a motherboard that supports 2 physical CPU's if you put 2 single core CPU's on that motherboard and do a benchmark , then remove those CPU's and put 1 dual core cpu instead, the dual core will benchmark higher, even if it is just going from 2 single core amd 64 3800+ to 1 dual core 3800+x2
SLI is the most inefficient invention in computer history.
the more cards you add, the lower the performance increase becomes.
when you do 2 cards in sli, you get about a 30-40% improvement oner a single card,
when you do 3 day SLI, you only get about a 5% improvement
the performance overhead of SLI exponentially increases as more cards are added so your getting less and less performance per dollar out of your investment
the reason why videocard companies are doing this s because people keep buying the crap. if you accept crap then you will get crap
when you compare the speed improvement when going from a geforce 5900, to a geforce 6800, the improvement is like 3-4 times more
when you go from a 6800 to a 7800, the improvement is about 1.5 times faster
when each new card, the level of improvement becomes more and more minute
and instead of making better cards, nvidia and ati are seeing that they can make more money by selling 2 slow expensive cards in CF or SLI, instead of making 1 fast card
SLI is just too inefficient, while it helps, it is smarter to just buy 1 card, then wait a month or 2 then but the next gen card that would be faster than of you wet SLI now and it would have cost less
it is like if you get a motherboard that supports 2 physical CPU's if you put 2 single core CPU's on that motherboard and do a benchmark , then remove those CPU's and put 1 dual core cpu instead, the dual core will benchmark higher, even if it is just going from 2 single core amd 64 3800+ to 1 dual core 3800+x2
customisbetter
Distinguished
- Apr 13, 2008
- 1,054
- 0
- 19,290
this is by far the stupidest review i have ever encountered. TWO DIFFERENT SYSTEMS??? come on guys...
Luscious
Distinguished
- Apr 5, 2006
- 525
- 0
- 18,980
This article has a few evident flaws:
1. The two boxes had similar but still different hardware. You should have tested just one system and simply do a "swap" with the video cards to eliminate other performance factors.
2. Crysis was notably missing from your benchmark - a game most people would especially like to see perform on exactly a 3/4-way GPU setup.
3. The HD playback results deserve some further explanation. An 8% to 31% difference seems odd to me.
Keep working on it Tom's - you'll get there!
1. The two boxes had similar but still different hardware. You should have tested just one system and simply do a "swap" with the video cards to eliminate other performance factors.
2. Crysis was notably missing from your benchmark - a game most people would especially like to see perform on exactly a 3/4-way GPU setup.
3. The HD playback results deserve some further explanation. An 8% to 31% difference seems odd to me.
Keep working on it Tom's - you'll get there!
mr roboto
Distinguished
- Oct 22, 2007
- 151
- 0
- 18,680
Another article showing systems that .001 percent of people actually own. Seriously how many people own 2 GX2's? If you do I feel sorry that you wasted your money according to what Nvidia wants you to think. Woohoo! a 15% increase for an extra $500! I'll stick to my single 8800GTX that still owns most games today. I agree completely that Nvidia needs to optimize SLI and even single card drivers first. Seems that we get new drivers quarterly now instead of monthly or even bi-monthly.
Write some articles that people can relate to. I especially liked the review of the ATI 3850 AGP solution that tested using the single core CPU. It's at least a much more common setup than this boutique stuff.
Write some articles that people can relate to. I especially liked the review of the ATI 3850 AGP solution that tested using the single core CPU. It's at least a much more common setup than this boutique stuff.
prey released july 11 2006
quake 4 released october 18 2005
serious sam 2 released october 11 2005
supreme commander febuary 16 2007
Warhammer Mark of Chaos November 14 2006
3dmark06 well idk its release date but i'm guessing 2006
its june of 08 and I would expect todays top hardware to handle TODAYS not yesterdays tough games. I love you guys at toms but this review sucked. Im not saying those games aren't graphically demanding but everyone of those games was playable on both systems. I'm going to have a real hard time noticing the difference in 53fps to 60 fps much less 266 to 290. I would notice the difference in say 20fps and 30 fps in crysis. I'm not really big on synthetics but I think that 3dmark vantage would be more suitable choice. If I am going to spend that type of chash I don't want to have to upgrade next month to max out my games so the question isn't how does it play 2-3 year old games and programs its how does it play todays stuff and tommorrows.
I want to know about age of conan, crysis, cod4, ut3, and tons of other games you probably would never do. frontlines, solar empire, rainbow 6 vegas 2, turok, etc. etc. etc. I don't expect 20 games reveiwed either but newer stuff for this type of technology. I don't think nvidia designed tri or quad sli for quake 4, prey, or warhammer.
So next time name the article comparison of Mach V vs Deluge L3 and include games that might use this type of hardware better.
quake 4 released october 18 2005
serious sam 2 released october 11 2005
supreme commander febuary 16 2007
Warhammer Mark of Chaos November 14 2006
3dmark06 well idk its release date but i'm guessing 2006
its june of 08 and I would expect todays top hardware to handle TODAYS not yesterdays tough games. I love you guys at toms but this review sucked. Im not saying those games aren't graphically demanding but everyone of those games was playable on both systems. I'm going to have a real hard time noticing the difference in 53fps to 60 fps much less 266 to 290. I would notice the difference in say 20fps and 30 fps in crysis. I'm not really big on synthetics but I think that 3dmark vantage would be more suitable choice. If I am going to spend that type of chash I don't want to have to upgrade next month to max out my games so the question isn't how does it play 2-3 year old games and programs its how does it play todays stuff and tommorrows.
I want to know about age of conan, crysis, cod4, ut3, and tons of other games you probably would never do. frontlines, solar empire, rainbow 6 vegas 2, turok, etc. etc. etc. I don't expect 20 games reveiwed either but newer stuff for this type of technology. I don't think nvidia designed tri or quad sli for quake 4, prey, or warhammer.
So next time name the article comparison of Mach V vs Deluge L3 and include games that might use this type of hardware better.
robwright
Distinguished
- Feb 16, 2006
- 1,129
- 7
- 19,285
We would have preferred to test two systems that were more similar, but we had to go with what Puget and Falcon NW graciously delivered us. As for swapping the video cards, we're reluctant to monkey around with a system builder's product too much, especially when they come with agreements that prevent any tinkering or swapping.
razor512
Distinguished
- Jun 16, 2007
- 2,135
- 71
- 19,890
[citation][nom]RobWright[/nom]We would have preferred to test two systems that were more similar, but we had to go with what Puget and Falcon NW graciously delivered us. As for swapping the video cards, we're reluctant to monkey around with a system builder's product too much, especially when they come with agreements that prevent any tinkering or swapping. [/citation]
there no way for them to know that you did some swapping as long as you swap it back,
just do all of the needed swapping then after that post the results but use 2 different names, like you can have all of the test done on the Falcon NW but when writing up the article, just have 1 of the pc's listed as Falcon NW and have the other listed as a commodore 64 or some other random crap
and we will just assume that both were done on the same pc since we know that the commodore 64 will not be too good for running crysis or 3dmark
so thats all you need to do,
both tests done on the same pc, but the 2 results listed under 2 different PC's
that way everyone is happy about the article, and Falcon is also happy because you didn't tinker with their pc
there no way for them to know that you did some swapping as long as you swap it back,
just do all of the needed swapping then after that post the results but use 2 different names, like you can have all of the test done on the Falcon NW but when writing up the article, just have 1 of the pc's listed as Falcon NW and have the other listed as a commodore 64 or some other random crap
and we will just assume that both were done on the same pc since we know that the commodore 64 will not be too good for running crysis or 3dmark
so thats all you need to do,
both tests done on the same pc, but the 2 results listed under 2 different PC's
that way everyone is happy about the article, and Falcon is also happy because you didn't tinker with their pc
- Status
Facebook
Twitter
Instagram