GTX260 Vs. GTX480 The Show Down!
  1 / 3    
[center][b][size=4][color="#008000"]Welcome back my friends to the show that never ends, we’re so glad you could attend, come inside, come inside![/color][/size] [/b][/center]
[center][img]http://farm5.static.flickr.com/4034/4692853348_ff49700a32_b.jpg[/img]

It might have taken me two months, a reload of Windows 7 and multiple PSU and GPU swap outs to get here but it's all been worth it! A while back I started posting requests for information on what types of data you would want to see on these cards in my Ryme Review's thread. The response I got was great and I hope I've covered most of the bases. I took all requests into consideration but ran into a few hickups along the way. I had to remove two games from the list as after the windows reload I was no longer able to access my save game files due to Games For Windows Live not allowing it. (/me mutters something under his breath) Those games were Bio Shock 2 and F.E.A.R. 2. Not really that big-a-deal as both of those games are fairly light on hardware and were only added to give an idea of DX9 performance while the focus of this review was on the more current DX10 and DX11 performance of these cards.

So that being said lets move on to the good stuff.

[b][u]Games Reviewed:[/u][/b]

Metro 2033

Heaven Benchmark 2.0

Dirt 2

Crysis

Batman Arkham Aslyum

Far Cry 2


[b][u]Test Bed Setup:[/u][/b]

EVGA 780i Socket 775 3-Way SLI MoBo
Intel Qx9770 4.0Ghz (1600mhz FSB) 1:1 with 800mhz memory
4 Gig OCZ ReaperX 4-4-4-12-2T 800mhz
4 Gig Mushkin Blackline 4-4-12-2T 800mhz
PC Power & Cooling T12W 1200w 100A Single 12v rail
1x EVGA GTX260 Core 216 Super Clock 65nm
1x EVGA GTX260 Core 216 Super Clock 55nm
2x EVGA GTX480 Reference Clock

1x 46inch Samsung 120hz HDTV at 1920x1080 (1808x1012 after desktop resize)

For the purpose of this review the only hardware change for all benchmarks is swapping the set of GTX260's for the set of GTX480's. No dedicated Physx card was used in any test other then during the single card tests the second GPU was set to process Physx. All listed frame rates are average frame rates with the coinciding benchmark being run 3 times when possible. The tests done with the GTX260's were run with the WHQL 197.45 drivers while the tests run with the GTX480's was run with the Beta 257.15 drivers. No changes were made in the Nvidia Control Panel at the driver level. Default profiles for all.

[u][b]Metro 2033[/b][/u]


This game is the new be all end all to hardware stress testing. Even my friends who are casual gamers have been telling me that it's the new Crysis and boy are they right. During all tests GPU usage was in the high 95% range with both sets of cards and during the DX11 Max Settings run both of the 480's were running 97%+ with over 1.2gig of video ram in use. With Metro 2033 there is no in game benchmark so I was forced to run a FRAPS benchmark. I started a new game and played until just before you go outside for the first time. For the most part this took about 7,000 frames give or take. Single and SLI tests for both sets of cards were run in DX10 with max settings. The DX11 test is noted in the graph.


[img]http://farm5.static.flickr.com/4012/4692220825_56a5b4ba78_b.jpg[/img]



[b][u]GTX260's[/u][/b]
As you can see the rumors are very [u]true[/u]. This game is a GPU killer for sure but looks absolutely amazing. A single GTX260 stood no chance at a playable frame rate. Adding a second card for SLI resulted in a 73.91% increase in average frame rate which is pretty amazing scaling for such a new game but still fell far short of that golden 60FPS for V-sync. I think Metro 2033 would be very playable in DX10 with a set of GTX260's if you were willing to sacrifice some settings to get there, but as you all know I hate to sacrifice anything. I want to see the game just as the Dev's intended it to be.

[u][b]GTX480's[/b][/u]
Right off the bat the single GTX480 comes flying out of the gate slamming down a 51.65% increase over the single 260. Poor guy, I almost feel bad for him now. Nahhh! So if we take just those numbers it's pretty safe to say that a single GTX480 was doing just about as well as SLI 260's. Kicking it up a notch SLI GTX480's unleashes a 41.43% over all increase in performance over the 260's and even grabs V-sync with 61.1 average FPS. Now that's all well and good, but what about the scaling ratio between the 480's that everyone has been talking about? Well 62.19% is the number and doesn't seem to be quite as high as I was expecting. I still think that we have to take into account that the drivers that ran the GTX260 tests were fully... fully... fully..mature while this is really only the second public set of drivers for the 480's as well as a completely new driver set. If all of that wasn't enough information to leave your head spinning here comes the DX11 numbers. Not being able to help myself I had to run one more test with full DX11 settings which resulted in a 32.81% performance drop from the DX10 numbers. That is a pretty heavy hit right there, but hopefully adding in a GTX260 as a Physx card will help to boost that frame rate a bit. I'll keep my fingers crossed and comment on the results of that in the full game review to come.


[b][u]Haven Benchmark 2.0[/u][/b]


I've had the Haven Benchmark since it came out and I honestly think that they should make some type of game out of it. I've used it to show friends the differences between SLI performance and it's done a fantastic job. BUT I honestly didn't realize what surprises were waiting for me once I got the 480's and cranked up tessellation for the first time. Yes, I watched all the videos but seeing it in person on a 46inch TV will make even the most modest gamers chin hit the floor. I think that for a GPU benchmark it is one of the best and most scalable around.


[img]http://farm5.static.flickr.com/4001/4692853326_ee13b2eaaf_b.jpg[/img]


[u][b]GTX260's[/b][/u]
This benchmark in DX10 with all settings maxed looks beautiful. The problem with things that look beautiful these days is that they are super demanding on our hardware. That fact and my quest for perfection in gaming is what lead me to SLI. Looking at the singe card results we can see that the GPU is clearly the bottleneck here and performance is to the point of unplayability. Adding in a second card leads to an increase of 91.12% which is absolutely outstanding. Performance increases like this should be the goal for every developer who is designing a game knowing that it will have to be played on a multi-GPU setup. I don't care if it's crossfire or SLI, this is a clear example of how it's done. Though the scaling is top notch the performance still leaves something to be desired when it comes to average frame rate. The good news here is that this is only a benchmark and is designed to be "to good" for current hardware so the fact that SLI GTX260's kicked out a sold 40FPS is pretty good in my book.

[u][b]GTX480's[/b][/u]
Dear lord. If these are the performance numbers I can come to expect from future titles for the next few years I am excited. When the single card test finished I had to do a double take to make sure I read it correctly. A single GTX480 out performed SLI GTX260 Core 216's by 22.98% and to me that is jaw dropping. Pop open the control panel and enable SLI and you shall receive a, dear lord, 117.60% increase from SLI 206's to SLI 480's. Now that ladies and gents is cranking that s#it up. Next up adding in a second GTX480 results in a 76.93% increase in performance under DX10. This also proves to me that I don't need to go I7 to keep up graphically. My Qx9770 at 4Ghz is plenty enough to feed these cards all the data they can crunch in this setup. The big hit comes when you add maxed out tessellation and comes down to a 66.97% decrease in performance. It almost seems like you chop off an entire card when running extreme tessellation. I wonder if next it will be dedicated tessellation/Physx cards. Could be? Both cards ran between 97-99% GPU usage and as a special little treat I give you this.

[img]http://farm5.static.flickr.com/4011/4694789334_c78ca394ea_b.jpg[/img]

I know it's a little tough to read those numbers after the screen shot has been down sized but EVGA Precision says: GPU1 95% at 91c and GPU2 95% at 90c. Those are the highest temps I have been able to makes these cards run. Those temps are after The max DX11 max Tessellation benchmark and for the performance recieved are not at all bad in my mind. For most of the games I've been playing ie* MW2 and The Witcher 80ish C is the norm. Note: I haven't created any custom fan profiles. This is all done at stock with no modifications what so ever. These cards from what I have experienced so far are right on par with my G80 8800GTX's.


[b][u]Dirt 2[/u][/b]


Dirt 2 is one of those games that really needs a high performance system to be played properly. I remember playing the first one on my set of 8600GT's and not be able to understand why in the world the game was so choppy! Obviously a few years have passed and my knowledge has grown to the point where I now understand that two mid range cards with small amounts of V-ram aren't going to do it. The strange thing about Dirt 2 is the fact that it runs in only two modes. DX9 or DX11. Why not DX10 and 11? I have no idea. But this game will put a graphics card and CPU through the ringer. While in game expect to see 70-90% CPU usage along with 80-90% GPU usage the whole way through. However this is where things start to get a little weird. Take a look at the numbers below.


[img]http://farm5.static.flickr.com/4045/4692853286_b1d554a331_b.jpg[/img]


[u][b]GTX260's[/b][/u]
When Dirt 2 came out it flat out didn't support SLI. It sucked majorly, but the game is very playable at 30FPS all the way down to 25FPS with no stutter or anything really. Once those magic drivers came out with an SLI profile I do have to admit I was a happy camper. Going from single card to SLI the gain in performance was 72.64% which again with mature drivers seems to be the standard scaling percentage across the board. The weird thing is what happened next.

[b][u]GTX480's[/u][/b]
A single GTX480 performed right on Q spitting out a 12.48% lead over SLI GTX260's. Then I got confused. SLI GTX480's performed with a 4.62% loss over a single card! Woah wait a second what? I ran the test 3 times getting exactly the same result. Low GPU usage on both cards and absolutely no performance improvement from SLI. Maybe they forgot the Dirt 2 SLI profile in the 257.15 drivers? No idea, but this is something Nvidia needs to take a look at ASAP. I even double checked to make sure I had forced DX11 and Tessellation off properly and I had. I'll keep my eyes open and see if anyone else has experienced the same thing.


[b][u]Crysis[/u][/b]


Regardless of what anyone tells you the first benchmark they look at in reviews is still Crysis. This may be on the verge of changing with Metro 2033 but at this very moment it's still true. Even my girl knows about Crysis and could probably tell you the story line in detail because I've played through it so many times in front of her testing new hardware. These tests were run with all settings Very High with 8xQAA Ha, but that's nothing you all don't already know so lets get right down to it. Mrs. Vanna White, the numbers please.


[img]http://farm5.static.flickr.com/4033/4692220785_3dc68a59f3_b.jpg[/img]


[u][b]GTX260's[/b][/u]
These cards love this game. A single card running above 30FPS with 8xQAA is pretty amazing, but still nowhere near the magic 60FPS of V-sync. Adding in a second card for SLI will net a 25.62% gain in performance. This is a major drop in scaling and I just don't understand why. I have to think at this point Crysis is no longer the focus for the driver team, but a more then 50% loss in scaling when compared to other games? Do drivers really make THAT much of a difference? If I still had the cards I would have loved to pop back to say the 180.** drivers or maybe even further back to see just how much of an improvement they make.

[u][b]GTX480[/b][/u]
Again we see similar results to Dirt 2 with a single GTX480 showing a disappointing loss of 25.27% over SLI GTX260 SLI. My thinking at the end of this test was, Woah if SLI scales well I could be seeing Crysis at 60FPS for the first time even though a single GTX260 out performed a single GTX480 with a decent amount of AA. Sadly this was not the case. Adding in a second card only resulted in a 6.06% increase over GTX260 SLI. What in the world is going on here? From single card to SLI only produced a 41.93% performance increase. This is better then the 260's but still no where near that 75%+ scaling we are seeing in more recent titles.


[b][u]Batman Arkham Asylum[/u][/b]


Batman was picked for the review due the issues it caused with Physx. Ok, well not only that, but the fact that it was an awesome game but still was demanding on hardware. Most of you have read Chrisray's thread dedicated to Batman AA and Physx which gave a literal ton of information when it came to SLI performance with Physx. The thing we all learned (and most of us learned it the hard way) was that when in SLI the primary card does all Physx calculations while the second card has to wait for the first to go back to rendering. This causes extreme frame rate drops and can even make the game unplayable on max Physx settings. The solution to this was disabling SLI and allowing your second card to be a dedicated Physx card (PPU) and should resolve the problem. Good work around, but we still want SLI+Physx so the name of the game is can Physx be scaled across two cards? Obviously we are still waiting on an answer to that, but in the mean time new hardware is still coming out. What Nvidia has done with the new architecture is increased the speed at which the GPU can switch between cuda and render modes. Again, a nice work around and does offer large performance gains but as you will see below there is still a need for a co-op or dedicated PPU for top notch performance.


[img]http://farm5.static.flickr.com/4054/4692853270_2352128737_b.jpg[/img]


[b][u]GTX260's[/u][/b]
The GTX260's performed just as was mentioned in Chrisray's Preview. A single card beat out SLI by 19.56% which is a large enough margin to make or break a games average frame rate and even brought it close to 60FPS for V-sync. Still for the amount of Physx in this game there really shouldn't have been this large of a performance hit. That is just my opinion, but from what I've seen it's shared by a lot of the community. Don't get me wrong, I loved the effects and want to see them in more titles, but more needs to be done to do it efficiently or spread the load amongst GUP's and CPU.

[u][b]GTX480's[/b][/u]
A nice performance gain was seen here with a single card. One GTX480 lead a single GTX260 by a margin of 34.54%. That is a good number, but I honestly would like to see more. Again I think this is due to driver maturity or the lack there of in this case. SLI performance again was an upset leading to a loss over a single card of 1.36%. Even with Physx enabled there should be no loss in performance due to SLI.


[u][b]Far Cry 2[/b][/u]


Far Cry 2 is one of those games that got lost in hype. I hold Ubisoft solely responsible for this and hope that they learned there lesson 100 times over. The game it's self looked very good and had some amazing new features, but the game they built around those features was repetitive and mostly just got annoying after small lengths of time. it does have a few redeeming qualities though. I can still go back and drive or walk around endlessly looking at the environment and the animals and foliage and enjoy myself. It also has an in-game benchmark that works well.


[img]http://farm5.static.flickr.com/4067/4692220813_32b06d2eac_b.jpg[/img]


[u][b]GTX260's[/b][/u]
I originally played this game with 3-Way SLI 8800GTX's (happy Morph?) and ran into many driver issues, but this seems to all have been worked out by the time I went back and played around a bit for this review. Of course with each of these games I had to play an hour or two before running the benchmark tests just for fun. So, a single GTX260 is still not enough to play this game to it's full potential. Adding a second GTX260 raised performance by 58.74%, wait a second! Were getting better! Still not at the 70%+ mark but much better then the last two games. I hope this carries over to the 480's as well.

[u][b]GTX480's[/b][/u]
A single GTX480 pushes over that 60FPS number and keeps right on going. Adding a second GTX480 we see a gain of 0.27%. That is just sad. Again, what is going on here Nvidia? I had to hop out of the game to make sure that I didn't make a mistake while enabling SLI. Everything was fine, so I went back and ran the bench again and got the exact same results. Again I feel this has to be driver related. The games in this review that failed horribly really make me wish that I had waited to do this review with more mature drivers. But this is the truth and things need to be brought to the attention of those who can fix them.


[u][b]Closing[/b][/u]


In closing I would like to once again thank Nvidia for hosting the CLUB SLI dinner at PAX. I can personally tell you that all of the forum members, the mods and everyone who was present had a great time and we can't wait to do it again next year.

As for the cards, I am greatly impressed with the performance in newer titles and will be enjoying the features as soon as I proof read and post this. I'm excited to get into Metro 2033 and BFBC2 and I know that my hardware is more then capable of Cranking That S#it up and serving the performance I want and demand from my investment. I do think that some time needs to be spent perfecting drivers to be able to maintain the balance of performance in new titles while still keeping performance for older, yet not old, PC games. I was sad to see my GTX260's gone from my case, but those heat pipes look wicked mean and ready to go. One topic I didn't cover was noise. I've seen people complain that the fans on these cards are loud and distracting, but I honestly don't notice. I can say that the fans in the 480's don't make any more noticeable noise then the 260's did. The fact of the matter is, if you are playing with high end equipment that needs high end cooling, how can you complain about some fan noise? That makes no sense to me. Turn up your 7.1 Dolby souround sound or crank up your headset and quit your whining.

[img]http://farm5.static.flickr.com/4060/4692853404_6b7bcfecd8_b.jpg[/img]

I'll post an updated pic once I slap in the 55nm GTX260 Core 216 for dedicated Physx. Metro 2033 here I come.

[/center]
Welcome back my friends to the show that never ends, we’re so glad you could attend, come inside, come inside!


Image



It might have taken me two months, a reload of Windows 7 and multiple PSU and GPU swap outs to get here but it's all been worth it! A while back I started posting requests for information on what types of data you would want to see on these cards in my Ryme Review's thread. The response I got was great and I hope I've covered most of the bases. I took all requests into consideration but ran into a few hickups along the way. I had to remove two games from the list as after the windows reload I was no longer able to access my save game files due to Games For Windows Live not allowing it. (/me mutters something under his breath) Those games were Bio Shock 2 and F.E.A.R. 2. Not really that big-a-deal as both of those games are fairly light on hardware and were only added to give an idea of DX9 performance while the focus of this review was on the more current DX10 and DX11 performance of these cards.



So that being said lets move on to the good stuff.



Games Reviewed:



Metro 2033



Heaven Benchmark 2.0



Dirt 2



Crysis



Batman Arkham Aslyum



Far Cry 2





Test Bed Setup:



EVGA 780i Socket 775 3-Way SLI MoBo

Intel Qx9770 4.0Ghz (1600mhz FSB) 1:1 with 800mhz memory

4 Gig OCZ ReaperX 4-4-4-12-2T 800mhz

4 Gig Mushkin Blackline 4-4-12-2T 800mhz

PC Power & Cooling T12W 1200w 100A Single 12v rail

1x EVGA GTX260 Core 216 Super Clock 65nm

1x EVGA GTX260 Core 216 Super Clock 55nm

2x EVGA GTX480 Reference Clock



1x 46inch Samsung 120hz HDTV at 1920x1080 (1808x1012 after desktop resize)



For the purpose of this review the only hardware change for all benchmarks is swapping the set of GTX260's for the set of GTX480's. No dedicated Physx card was used in any test other then during the single card tests the second GPU was set to process Physx. All listed frame rates are average frame rates with the coinciding benchmark being run 3 times when possible. The tests done with the GTX260's were run with the WHQL 197.45 drivers while the tests run with the GTX480's was run with the Beta 257.15 drivers. No changes were made in the Nvidia Control Panel at the driver level. Default profiles for all.



Metro 2033





This game is the new be all end all to hardware stress testing. Even my friends who are casual gamers have been telling me that it's the new Crysis and boy are they right. During all tests GPU usage was in the high 95% range with both sets of cards and during the DX11 Max Settings run both of the 480's were running 97%+ with over 1.2gig of video ram in use. With Metro 2033 there is no in game benchmark so I was forced to run a FRAPS benchmark. I started a new game and played until just before you go outside for the first time. For the most part this took about 7,000 frames give or take. Single and SLI tests for both sets of cards were run in DX10 with max settings. The DX11 test is noted in the graph.





Image







GTX260's

As you can see the rumors are very true. This game is a GPU killer for sure but looks absolutely amazing. A single GTX260 stood no chance at a playable frame rate. Adding a second card for SLI resulted in a 73.91% increase in average frame rate which is pretty amazing scaling for such a new game but still fell far short of that golden 60FPS for V-sync. I think Metro 2033 would be very playable in DX10 with a set of GTX260's if you were willing to sacrifice some settings to get there, but as you all know I hate to sacrifice anything. I want to see the game just as the Dev's intended it to be.



GTX480's

Right off the bat the single GTX480 comes flying out of the gate slamming down a 51.65% increase over the single 260. Poor guy, I almost feel bad for him now. Nahhh! So if we take just those numbers it's pretty safe to say that a single GTX480 was doing just about as well as SLI 260's. Kicking it up a notch SLI GTX480's unleashes a 41.43% over all increase in performance over the 260's and even grabs V-sync with 61.1 average FPS. Now that's all well and good, but what about the scaling ratio between the 480's that everyone has been talking about? Well 62.19% is the number and doesn't seem to be quite as high as I was expecting. I still think that we have to take into account that the drivers that ran the GTX260 tests were fully... fully... fully..mature while this is really only the second public set of drivers for the 480's as well as a completely new driver set. If all of that wasn't enough information to leave your head spinning here comes the DX11 numbers. Not being able to help myself I had to run one more test with full DX11 settings which resulted in a 32.81% performance drop from the DX10 numbers. That is a pretty heavy hit right there, but hopefully adding in a GTX260 as a Physx card will help to boost that frame rate a bit. I'll keep my fingers crossed and comment on the results of that in the full game review to come.





Haven Benchmark 2.0





I've had the Haven Benchmark since it came out and I honestly think that they should make some type of game out of it. I've used it to show friends the differences between SLI performance and it's done a fantastic job. BUT I honestly didn't realize what surprises were waiting for me once I got the 480's and cranked up tessellation for the first time. Yes, I watched all the videos but seeing it in person on a 46inch TV will make even the most modest gamers chin hit the floor. I think that for a GPU benchmark it is one of the best and most scalable around.





Image





GTX260's

This benchmark in DX10 with all settings maxed looks beautiful. The problem with things that look beautiful these days is that they are super demanding on our hardware. That fact and my quest for perfection in gaming is what lead me to SLI. Looking at the singe card results we can see that the GPU is clearly the bottleneck here and performance is to the point of unplayability. Adding in a second card leads to an increase of 91.12% which is absolutely outstanding. Performance increases like this should be the goal for every developer who is designing a game knowing that it will have to be played on a multi-GPU setup. I don't care if it's crossfire or SLI, this is a clear example of how it's done. Though the scaling is top notch the performance still leaves something to be desired when it comes to average frame rate. The good news here is that this is only a benchmark and is designed to be "to good" for current hardware so the fact that SLI GTX260's kicked out a sold 40FPS is pretty good in my book.



GTX480's

Dear lord. If these are the performance numbers I can come to expect from future titles for the next few years I am excited. When the single card test finished I had to do a double take to make sure I read it correctly. A single GTX480 out performed SLI GTX260 Core 216's by 22.98% and to me that is jaw dropping. Pop open the control panel and enable SLI and you shall receive a, dear lord, 117.60% increase from SLI 206's to SLI 480's. Now that ladies and gents is cranking that s#it up. Next up adding in a second GTX480 results in a 76.93% increase in performance under DX10. This also proves to me that I don't need to go I7 to keep up graphically. My Qx9770 at 4Ghz is plenty enough to feed these cards all the data they can crunch in this setup. The big hit comes when you add maxed out tessellation and comes down to a 66.97% decrease in performance. It almost seems like you chop off an entire card when running extreme tessellation. I wonder if next it will be dedicated tessellation/Physx cards. Could be? Both cards ran between 97-99% GPU usage and as a special little treat I give you this.



Image



I know it's a little tough to read those numbers after the screen shot has been down sized but EVGA Precision says: GPU1 95% at 91c and GPU2 95% at 90c. Those are the highest temps I have been able to makes these cards run. Those temps are after The max DX11 max Tessellation benchmark and for the performance recieved are not at all bad in my mind. For most of the games I've been playing ie* MW2 and The Witcher 80ish C is the norm. Note: I haven't created any custom fan profiles. This is all done at stock with no modifications what so ever. These cards from what I have experienced so far are right on par with my G80 8800GTX's.





Dirt 2





Dirt 2 is one of those games that really needs a high performance system to be played properly. I remember playing the first one on my set of 8600GT's and not be able to understand why in the world the game was so choppy! Obviously a few years have passed and my knowledge has grown to the point where I now understand that two mid range cards with small amounts of V-ram aren't going to do it. The strange thing about Dirt 2 is the fact that it runs in only two modes. DX9 or DX11. Why not DX10 and 11? I have no idea. But this game will put a graphics card and CPU through the ringer. While in game expect to see 70-90% CPU usage along with 80-90% GPU usage the whole way through. However this is where things start to get a little weird. Take a look at the numbers below.





Image





GTX260's

When Dirt 2 came out it flat out didn't support SLI. It sucked majorly, but the game is very playable at 30FPS all the way down to 25FPS with no stutter or anything really. Once those magic drivers came out with an SLI profile I do have to admit I was a happy camper. Going from single card to SLI the gain in performance was 72.64% which again with mature drivers seems to be the standard scaling percentage across the board. The weird thing is what happened next.



GTX480's

A single GTX480 performed right on Q spitting out a 12.48% lead over SLI GTX260's. Then I got confused. SLI GTX480's performed with a 4.62% loss over a single card! Woah wait a second what? I ran the test 3 times getting exactly the same result. Low GPU usage on both cards and absolutely no performance improvement from SLI. Maybe they forgot the Dirt 2 SLI profile in the 257.15 drivers? No idea, but this is something Nvidia needs to take a look at ASAP. I even double checked to make sure I had forced DX11 and Tessellation off properly and I had. I'll keep my eyes open and see if anyone else has experienced the same thing.





Crysis





Regardless of what anyone tells you the first benchmark they look at in reviews is still Crysis. This may be on the verge of changing with Metro 2033 but at this very moment it's still true. Even my girl knows about Crysis and could probably tell you the story line in detail because I've played through it so many times in front of her testing new hardware. These tests were run with all settings Very High with 8xQAA Ha, but that's nothing you all don't already know so lets get right down to it. Mrs. Vanna White, the numbers please.





Image





GTX260's

These cards love this game. A single card running above 30FPS with 8xQAA is pretty amazing, but still nowhere near the magic 60FPS of V-sync. Adding in a second card for SLI will net a 25.62% gain in performance. This is a major drop in scaling and I just don't understand why. I have to think at this point Crysis is no longer the focus for the driver team, but a more then 50% loss in scaling when compared to other games? Do drivers really make THAT much of a difference? If I still had the cards I would have loved to pop back to say the 180.** drivers or maybe even further back to see just how much of an improvement they make.



GTX480

Again we see similar results to Dirt 2 with a single GTX480 showing a disappointing loss of 25.27% over SLI GTX260 SLI. My thinking at the end of this test was, Woah if SLI scales well I could be seeing Crysis at 60FPS for the first time even though a single GTX260 out performed a single GTX480 with a decent amount of AA. Sadly this was not the case. Adding in a second card only resulted in a 6.06% increase over GTX260 SLI. What in the world is going on here? From single card to SLI only produced a 41.93% performance increase. This is better then the 260's but still no where near that 75%+ scaling we are seeing in more recent titles.





Batman Arkham Asylum





Batman was picked for the review due the issues it caused with Physx. Ok, well not only that, but the fact that it was an awesome game but still was demanding on hardware. Most of you have read Chrisray's thread dedicated to Batman AA and Physx which gave a literal ton of information when it came to SLI performance with Physx. The thing we all learned (and most of us learned it the hard way) was that when in SLI the primary card does all Physx calculations while the second card has to wait for the first to go back to rendering. This causes extreme frame rate drops and can even make the game unplayable on max Physx settings. The solution to this was disabling SLI and allowing your second card to be a dedicated Physx card (PPU) and should resolve the problem. Good work around, but we still want SLI+Physx so the name of the game is can Physx be scaled across two cards? Obviously we are still waiting on an answer to that, but in the mean time new hardware is still coming out. What Nvidia has done with the new architecture is increased the speed at which the GPU can switch between cuda and render modes. Again, a nice work around and does offer large performance gains but as you will see below there is still a need for a co-op or dedicated PPU for top notch performance.





Image





GTX260's

The GTX260's performed just as was mentioned in Chrisray's Preview. A single card beat out SLI by 19.56% which is a large enough margin to make or break a games average frame rate and even brought it close to 60FPS for V-sync. Still for the amount of Physx in this game there really shouldn't have been this large of a performance hit. That is just my opinion, but from what I've seen it's shared by a lot of the community. Don't get me wrong, I loved the effects and want to see them in more titles, but more needs to be done to do it efficiently or spread the load amongst GUP's and CPU.



GTX480's

A nice performance gain was seen here with a single card. One GTX480 lead a single GTX260 by a margin of 34.54%. That is a good number, but I honestly would like to see more. Again I think this is due to driver maturity or the lack there of in this case. SLI performance again was an upset leading to a loss over a single card of 1.36%. Even with Physx enabled there should be no loss in performance due to SLI.





Far Cry 2





Far Cry 2 is one of those games that got lost in hype. I hold Ubisoft solely responsible for this and hope that they learned there lesson 100 times over. The game it's self looked very good and had some amazing new features, but the game they built around those features was repetitive and mostly just got annoying after small lengths of time. it does have a few redeeming qualities though. I can still go back and drive or walk around endlessly looking at the environment and the animals and foliage and enjoy myself. It also has an in-game benchmark that works well.





Image





GTX260's

I originally played this game with 3-Way SLI 8800GTX's (happy Morph?) and ran into many driver issues, but this seems to all have been worked out by the time I went back and played around a bit for this review. Of course with each of these games I had to play an hour or two before running the benchmark tests just for fun. So, a single GTX260 is still not enough to play this game to it's full potential. Adding a second GTX260 raised performance by 58.74%, wait a second! Were getting better! Still not at the 70%+ mark but much better then the last two games. I hope this carries over to the 480's as well.



GTX480's

A single GTX480 pushes over that 60FPS number and keeps right on going. Adding a second GTX480 we see a gain of 0.27%. That is just sad. Again, what is going on here Nvidia? I had to hop out of the game to make sure that I didn't make a mistake while enabling SLI. Everything was fine, so I went back and ran the bench again and got the exact same results. Again I feel this has to be driver related. The games in this review that failed horribly really make me wish that I had waited to do this review with more mature drivers. But this is the truth and things need to be brought to the attention of those who can fix them.





Closing





In closing I would like to once again thank Nvidia for hosting the CLUB SLI dinner at PAX. I can personally tell you that all of the forum members, the mods and everyone who was present had a great time and we can't wait to do it again next year.



As for the cards, I am greatly impressed with the performance in newer titles and will be enjoying the features as soon as I proof read and post this. I'm excited to get into Metro 2033 and BFBC2 and I know that my hardware is more then capable of Cranking That S#it up and serving the performance I want and demand from my investment. I do think that some time needs to be spent perfecting drivers to be able to maintain the balance of performance in new titles while still keeping performance for older, yet not old, PC games. I was sad to see my GTX260's gone from my case, but those heat pipes look wicked mean and ready to go. One topic I didn't cover was noise. I've seen people complain that the fans on these cards are loud and distracting, but I honestly don't notice. I can say that the fans in the 480's don't make any more noticeable noise then the 260's did. The fact of the matter is, if you are playing with high end equipment that needs high end cooling, how can you complain about some fan noise? That makes no sense to me. Turn up your 7.1 Dolby souround sound or crank up your headset and quit your whining.



Image



I'll post an updated pic once I slap in the 55nm GTX260 Core 216 for dedicated Physx. Metro 2033 here I come.



Image

#1
Posted 06/13/2010 03:00 AM   
[quote name='420Ryme' post='1072612' date='Jun 12 2010, 11:00 PM']...The fact of the matter is, if you are playing with high end equipment that needs high end cooling, how can you complain about some fan noise? That makes no sense to me. Turn up your 7.1 Dolby [i][b]sour[/b][/i]ound sound or crank up your headset and quit your whining...[/quote]
Or actually use an E.L.P. soundtrack in the background since they've been referenced.

Very well done and clearly written...
[quote name='420Ryme' post='1072612' date='Jun 12 2010, 11:00 PM']...The fact of the matter is, if you are playing with high end equipment that needs high end cooling, how can you complain about some fan noise? That makes no sense to me. Turn up your 7.1 Dolby souround sound or crank up your headset and quit your whining...

Or actually use an E.L.P. soundtrack in the background since they've been referenced.



Very well done and clearly written...

Intel Siler DX79SI Desktop Extreme | Intel Core i7-3820 Sandy Bridge-Extreme | DangerDen M6 and Koolance MVR-40s w/Black Ice Stealths | 32 GB Mushkin PC3-12800LV | NVIDIA GTX 660 Ti SLI | PNY GTX 470 | 24 GB RAMDisk (C:\Temp\Temp) | 120 GB Intel Cherryville SSDs (OS and UserData)| 530 GB Western Digital VelociRaptor SATA 2 RAID0 (C:\Games\) | 60 GB G2 SSDs (XP Pro and Linux) | 3 TB Western Digital USB-3 MyBook (Archive) | LG BP40NS20 USB ODD | LG IPS236 Monitor | LogiTech X-530 Speakers | Plantronics GameCom 780 Headphones | Cooler Master UCP 1100 | Cooler Master HAF XB | Windows 7 Pro x64 SP1

Stock is Extreme now

#2
Posted 06/13/2010 03:22 AM   
I have been waiting for this review since the release of the 400 series so i could make the decision of upgrading or not.

Thank's for the time and effort you have put into this review.
I have been waiting for this review since the release of the 400 series so i could make the decision of upgrading or not.



Thank's for the time and effort you have put into this review.

#3
Posted 06/13/2010 03:44 AM   
I notice you didn't put what resolution you were playing all of these games in, however, I did see you mention a 46" TV during the Heaven Benchmark section. I'm assuming you ran every thing in 1920x1080?
I notice you didn't put what resolution you were playing all of these games in, however, I did see you mention a 46" TV during the Heaven Benchmark section. I'm assuming you ran every thing in 1920x1080?

#4
Posted 06/13/2010 03:53 AM   
[quote name='Impatient' post='1072634' date='Jun 12 2010, 11:44 PM']I have been waiting for this review since the release of the 400 series so i could make the decision of upgrading or not.

Thank's for the time and effort you have put into this review.[/quote]

No problem! I would definitely say go for it. As drivers mature things can only go up from here. Like I mentioned a few times the GTX260's were on very mature drivers and produced results that showed that.

@Neff - yes the resolution is 1920x1080 but with desktop resize it's 1808x1012. I'll add that to the system specs at the top. Thanks for reminding me!
[quote name='Impatient' post='1072634' date='Jun 12 2010, 11:44 PM']I have been waiting for this review since the release of the 400 series so i could make the decision of upgrading or not.



Thank's for the time and effort you have put into this review.



No problem! I would definitely say go for it. As drivers mature things can only go up from here. Like I mentioned a few times the GTX260's were on very mature drivers and produced results that showed that.



@Neff - yes the resolution is 1920x1080 but with desktop resize it's 1808x1012. I'll add that to the system specs at the top. Thanks for reminding me!

Image

#5
Posted 06/13/2010 04:44 AM   
Thanks for the quick response 420Ryme! As an owner of 2x GTX 260 in SLI, I was very happy to see this article. I was also surprised (pleasantly) that SLI 260s could compete with one 480, in a few things at least. Since I don't have a spare grand lying around, I won't be upgrading to SLI 480s any time soon. :( Hopefully the drivers for the 480 are able to bring out its potential. I am looking forward to upgrading to a setup that supports DX 11 and can rock Tessellation in the future. /banana.gif' class='bbc_emoticon' alt=':banana:' />
Thanks for the quick response 420Ryme! As an owner of 2x GTX 260 in SLI, I was very happy to see this article. I was also surprised (pleasantly) that SLI 260s could compete with one 480, in a few things at least. Since I don't have a spare grand lying around, I won't be upgrading to SLI 480s any time soon. :( Hopefully the drivers for the 480 are able to bring out its potential. I am looking forward to upgrading to a setup that supports DX 11 and can rock Tessellation in the future. /banana.gif' class='bbc_emoticon' alt=':banana:' />

#6
Posted 06/13/2010 04:55 AM   
The tessellation in the Heaven benchmark is just amazing. Looking at the rocks on the walkways and buildings along with the roves of houses and the round shapes that are no longer made up of poly's is just astonishing. The funny part about it is from playing games for so long the old way being tricked by fake geometry then seeing this almost makes you feel dumb for not noticing before. As I play current games now I've started to notice more and more the little tricks that game devs use to fool the eye and make you suspend belief. Tessellation truly brings the scene to life.
The tessellation in the Heaven benchmark is just amazing. Looking at the rocks on the walkways and buildings along with the roves of houses and the round shapes that are no longer made up of poly's is just astonishing. The funny part about it is from playing games for so long the old way being tricked by fake geometry then seeing this almost makes you feel dumb for not noticing before. As I play current games now I've started to notice more and more the little tricks that game devs use to fool the eye and make you suspend belief. Tessellation truly brings the scene to life.

Image

#7
Posted 06/13/2010 05:01 AM   
Great review Ryme, I was looking for more info on scaling for when I return to the wonderful world of MGPU, and you definitely gave me enough to think about. Your singe [is that what happened to your hardware or a typo :D] card results are pretty similar to what I get with a X58 platform [which should give you a nice warm feeling inside :)].

I can confirm Metro 2033 is a graphics card killer, I average 30 FPS on 1 GTX 480 and a 280 for PhysX, and that's in 2D, guess I will be having to make some sacrifices when I put the goggles on. Have to say I'm not entirely sure why there are such low FPS though, but I think that's just ignorance on my part. I still think Crysis looks better, but I guess that's personal preference, I just like my games to be lighter and open rather than dark and dirty.

Would of liked to have seen a bit more on noise and heat especially with regards to SLi, but I guess you can't always get what you want.

I notice you did leave 1 area untouched, I wonder if we will see what these cards can do with a mild OC in the not too distant future :).

Thanks for the considerable time and effort you have put into this review, it's definitely helpful for me, as I'm sure it will be for many others, now go forth and plaaaaaaay !!!
Great review Ryme, I was looking for more info on scaling for when I return to the wonderful world of MGPU, and you definitely gave me enough to think about. Your singe [is that what happened to your hardware or a typo :D] card results are pretty similar to what I get with a X58 platform [which should give you a nice warm feeling inside :)].



I can confirm Metro 2033 is a graphics card killer, I average 30 FPS on 1 GTX 480 and a 280 for PhysX, and that's in 2D, guess I will be having to make some sacrifices when I put the goggles on. Have to say I'm not entirely sure why there are such low FPS though, but I think that's just ignorance on my part. I still think Crysis looks better, but I guess that's personal preference, I just like my games to be lighter and open rather than dark and dirty.



Would of liked to have seen a bit more on noise and heat especially with regards to SLi, but I guess you can't always get what you want.



I notice you did leave 1 area untouched, I wonder if we will see what these cards can do with a mild OC in the not too distant future :).



Thanks for the considerable time and effort you have put into this review, it's definitely helpful for me, as I'm sure it will be for many others, now go forth and plaaaaaaay !!!

#8
Posted 06/13/2010 09:32 AM   
Nice Ryme!!!

Now... OC those 480's to 900Mhz+ core and re-do the bench :)
If heat or stability becomes an issue... flash to ASUS or Galaxy's Bios (I can email u the bios files if u need them)
Nice Ryme!!!



Now... OC those 480's to 900Mhz+ core and re-do the bench :)

If heat or stability becomes an issue... flash to ASUS or Galaxy's Bios (I can email u the bios files if u need them)

MS-Physics (Thermal Dynamics)/BA-IT/BS-Electronic Engineering And still going for more ;)~ Maybe Chem. next?
Owner/Lead R&D: Socomcomputers.com
Ebay: http://www.ebay.com/usr/jake2gnt
Yes; I am Jerk.But I am a helpful jerk.I promise to help you to the best of my abilities.
My main rig:
OS- Win7 Pro x64 SP1 (core parking disabled)
CPU- FX8320 8 cores active @ 4.8Ghz 1.43vcore Liquid Cooled w/ chiller
-HT= 2750mhz / NB= 2450mhzRAM
-16Gb Viper 3 @ 2020Mhz (9-10-9-27)1.525v
Motherboard- Asus Crosshair V
Main Graphics- 3-way SLI EVGA GTX770 4Gb @ 1278Mhz core w/ 334.67 drivers Liquid Cooled
SSD1- Samsung 120GB (OS drive raid0)
SSD2- Samsung 120GB (OS drive raid0)
SSD3- Mushkin 240Gb (game drive)
HDD1- Samsung 1TB (media drive)
HDD2- Seagate Hybrid SSD/HDD 500Gb (back-up game drive)
Power Supply- TT Toughpower 1200w PSU
Samsung 27" 2560 x 1440p
STILL working Nvidia GPUs-FX5200,FX5700,6800GT,7900GT(x2),8800GTS(x2),9800GTX,GTX275(x2),GTX285(x2),GTX295(x2),GTX470(x3),GTX480,GTX660ti(x3)GTX770(x3)
HWbot Overclocking results
#192 of 2446 in United States Xtreme OC league

#9
Posted 06/13/2010 10:53 AM   
Awesome Review! /woot.gif' class='bbc_emoticon' alt=':woot:' />

Ryme, I know this review was "long time coming" and a lot of effort /yes.gif' class='bbc_emoticon' alt=':yes:' /> Your are a Champion here at NVIDIA Forums, Thank You for a honest and straight to the point Review :D TJ
Awesome Review! /woot.gif' class='bbc_emoticon' alt=':woot:' />



Ryme, I know this review was "long time coming" and a lot of effort /yes.gif' class='bbc_emoticon' alt=':yes:' /> Your are a Champion here at NVIDIA Forums, Thank You for a honest and straight to the point Review :D TJ

CM HAF 932 - ASUS Rampage IV Extreme- Intel i7 3930K - Mushkin Redline 16GB 2133 - Antec HCP 1200kw - Galaxy GTX 670 3 way SLI - 2 WD 1TB Black,WD 640GB Black, WD SSD 64GB - Thermaltake CLW0217 Water 2.0 Extreme - ASUS 24x - Logitech G9 - Logitech G105 - I INC 28" Monitor at 1920 X 1200 (Native) Wow Gaming is great! Speakers - Logitech 530's - Win 7 Ultimate 64bit

Posted by Rsabatino, techjesse you are the people's champ of overclocking

Vantage P70249ORB
Image

#10
Posted 06/13/2010 06:36 PM   
"The fact of the matter is, if you are playing with high end equipment that needs high end cooling, how can you complain about some fan noise? That makes no sense to me. Turn up your 7.1 Dolby souround sound or crank up your headset and quit your whining. "


LOL Bravo! That's what I've been thinking.

Good review. Now get those cards on a x58 and watch those frame rates go even higher. ;-)
"The fact of the matter is, if you are playing with high end equipment that needs high end cooling, how can you complain about some fan noise? That makes no sense to me. Turn up your 7.1 Dolby souround sound or crank up your headset and quit your whining. "





LOL Bravo! That's what I've been thinking.



Good review. Now get those cards on a x58 and watch those frame rates go even higher. ;-)

#11
Posted 06/13/2010 06:56 PM   
Impressive review sir. Now to really show the difference from generation to generation you need a pair of 9800's to show people how slow they are now adays :D
Impressive review sir. Now to really show the difference from generation to generation you need a pair of 9800's to show people how slow they are now adays :D

Using Left over CPU coolers to keep phones cool since 2012.

Image

#12
Posted 06/13/2010 10:25 PM   
3D Mark Vantage run - stock GPU clocks

[img]http://farm5.static.flickr.com/4019/4697388365_2282c73a8b_b.jpg[/img]

Thanks for the kind words guys.

@Bone - LOL I totally missed that and they are all wrong because I made a template and just kept changing the numbers! Ha but yes I did singe off a few hairs by touching the heat pipes on the 480's after not letting them cool enough before trying to take them out.

Seems like I've run into an issue with running a GTX260 Core 216 as a PPU. I ran the batman Physx benchmark and my average frame rate dropped down to 66FPS. As far as I know everything is configured correctly and from watching GPU usage it looks like it's working. Wonder why the loss? I also had to disable the 260 for the vantage run because it lowered my score by about 1,000 points. Any advice from those running dedicated Physx cards would be awesome. I'm about to take it out if it's not helping.
3D Mark Vantage run - stock GPU clocks



Image



Thanks for the kind words guys.



@Bone - LOL I totally missed that and they are all wrong because I made a template and just kept changing the numbers! Ha but yes I did singe off a few hairs by touching the heat pipes on the 480's after not letting them cool enough before trying to take them out.



Seems like I've run into an issue with running a GTX260 Core 216 as a PPU. I ran the batman Physx benchmark and my average frame rate dropped down to 66FPS. As far as I know everything is configured correctly and from watching GPU usage it looks like it's working. Wonder why the loss? I also had to disable the 260 for the vantage run because it lowered my score by about 1,000 points. Any advice from those running dedicated Physx cards would be awesome. I'm about to take it out if it's not helping.

Image

#13
Posted 06/13/2010 11:10 PM   
Great review of what & how it should be done, also it's about time people see things that a laymans eyes can understand.........CHEERS

BTW: I think the scores that where low on the one game was due to console limited codes for the pc ver, of course I could be wrong, not that I haven't been in the past ;).
Great review of what & how it should be done, also it's about time people see things that a laymans eyes can understand.........CHEERS



BTW: I think the scores that where low on the one game was due to console limited codes for the pc ver, of course I could be wrong, not that I haven't been in the past ;).

#14
Posted 06/13/2010 11:34 PM   
[quote name='420Ryme' post='1072971' date='Jun 14 2010, 12:10 AM']3D Mark Vantage run - stock GPU clocks

[img]http://farm5.static.flickr.com/4019/4697388365_2282c73a8b_b.jpg[/img]

Thanks for the kind words guys.

@Bone - LOL I totally missed that and they are all wrong because I made a template and just kept changing the numbers! Ha but yes I did singe off a few hairs by touching the heat pipes on the 480's after not letting them cool enough before trying to take them out.

Seems like I've run into an issue with running a GTX260 Core 216 as a PPU. I ran the batman Physx benchmark and my average frame rate dropped down to 66FPS. As far as I know everything is configured correctly and from watching GPU usage it looks like it's working. Wonder why the loss? I also had to disable the 260 for the vantage run because it lowered my score by about 1,000 points. Any advice from those running dedicated Physx cards would be awesome. I'm about to take it out if it's not helping.[/quote]

You broke over 30K Ryme, Ah the sweet 30K mark /woot.gif' class='bbc_emoticon' alt=':woot:' /> YEA Ryme, Now Overclock those beasts and hit it again /woot.gif' class='bbc_emoticon' alt=':woot:' />
[quote name='420Ryme' post='1072971' date='Jun 14 2010, 12:10 AM']3D Mark Vantage run - stock GPU clocks



Image



Thanks for the kind words guys.



@Bone - LOL I totally missed that and they are all wrong because I made a template and just kept changing the numbers! Ha but yes I did singe off a few hairs by touching the heat pipes on the 480's after not letting them cool enough before trying to take them out.



Seems like I've run into an issue with running a GTX260 Core 216 as a PPU. I ran the batman Physx benchmark and my average frame rate dropped down to 66FPS. As far as I know everything is configured correctly and from watching GPU usage it looks like it's working. Wonder why the loss? I also had to disable the 260 for the vantage run because it lowered my score by about 1,000 points. Any advice from those running dedicated Physx cards would be awesome. I'm about to take it out if it's not helping.



You broke over 30K Ryme, Ah the sweet 30K mark /woot.gif' class='bbc_emoticon' alt=':woot:' /> YEA Ryme, Now Overclock those beasts and hit it again /woot.gif' class='bbc_emoticon' alt=':woot:' />

CM HAF 932 - ASUS Rampage IV Extreme- Intel i7 3930K - Mushkin Redline 16GB 2133 - Antec HCP 1200kw - Galaxy GTX 670 3 way SLI - 2 WD 1TB Black,WD 640GB Black, WD SSD 64GB - Thermaltake CLW0217 Water 2.0 Extreme - ASUS 24x - Logitech G9 - Logitech G105 - I INC 28" Monitor at 1920 X 1200 (Native) Wow Gaming is great! Speakers - Logitech 530's - Win 7 Ultimate 64bit

Posted by Rsabatino, techjesse you are the people's champ of overclocking

Vantage P70249ORB
Image

#15
Posted 06/14/2010 02:00 AM   
  1 / 3    
Scroll To Top