4870 X2 / GTX 280

[djnsmith7]

I've been looking for additional insight on the differences between these 2 cards. Good work.

 

[udneekgnim]

regarding the WoW Terokkar test, my frames always seemed to drop pretty badly when flying threw there, but I always assumed it was my CPU (4400+ X2)

guess it's good to know my video card (4870) might share in some of the responsibility (altough I run at 16*10)

my WoW performace with the 4870 has been inconsistent enough that I've been tempted to pop in my previous video card (7950 GT) and compare performance since WoW has an in-game benchmark command for flight paths

 

[sourthings]

Originally posted by: RussianSensation

Originally posted by: sourthings

I am constantly blowing cash on pc stuff, about $500 a week. Yes, it's true, I'd have to sneak a $2000 monitor past her

Wow that's $26,000 on computer hardware annually. You must have a Gallardo as your winter beater

Haha, I wish! I drive a honda, probably because I spend too much on computers

 

[Aikouka]

Nitro, I was wondering something about WoW and the 4870 X2. I tend to play two WoWs at once (separate monitors) and was wondering if the X2 would still keep using a single GPU or would it be smart enough to divvy up the tasks? With the lackluster amount of memory on the normal 4870, running two WoWs at once is a huge memory monster (I assume they're maxing out the vram and both use around 600+ MB of RAM a piece with my 4870).

So in other words, I guess would you notice a difference running two instances of WoW at the same time?

People may wonder why I run relatively high-end cards just to play WoW (as I don't play any other game), but playing both in windowed mode (lower framerate because of Aero still running) I can easily hover around 30-50 fps on each system with AA set to 1x (I tried it maxed out at 8x, but that's also when I had the huge slowdown that I discussed in another thread, so I set it back). The new expansion will be upgrading the shadows to real-time shadow instead of the current blob system, which will require even more graphic prowess and be even more taxing with two open at once.

If the X2 will intelligently separate the tasks... one WoW per GPU (although I doubt it'd be that clean), then it might be worth an upgrade.

 

[nitromullet]

Originally posted by: ShadowOfMyself
Gonna get flamed but that looks very Nvidia skewed to me, specially the choice of games... Its like you specifically picked games that dont scale or are nvidia favorites (like crysis)
If you had tested Call of duty 4, Race grid, etc youd see the difference

But it comes down to - if you wanna play THOSE games, sure buy an nvidia card, if you wanna play the rest, buy the X2

Yep, it's *not* a review of the two cards. If you actually read what I posted, you would know that I looked at both GRID and COD4. Both cards played both games well, and I only made FRAPS runs games where I noticed a significant difference between the two.

Also, before anyone goes too far with calling me biased, please remember that the video cards used for these graphs cost me $1000+ of my own money. I'm not a beta tester or sponsored by anyone. I give NV and ATI my hard earned, so I might be a bit more nit picky than AT or other reviews sites that get the hardware free in exchange for publicity.

I find it interesting actually that no one has commented on the lackluster performance of the GTX 280 in Quake 4 w/ 8xAA. This was absolutely abysmal IMO for the age of the game/engine and the level of video card.

Originally posted by: MarcVenice
Nitromullet, the similarities are, obviously, that the fps go down and up at the exact same time, by exactly the same amount. In the ice walk thru, both videocards perform EXACTLY the same. Could it be a cpu bottlebeck, because it's weird for them to be so close to each other.

Btw, the terrokar forrest implies that the GTX280 is twice as fast as a HD4870X2. It could be that only 1 gpu is working, and it could in fact be that the performance is actually beneath that of a single HD4870.

Sorry, I misunderstood what you meant. I thought you were implying that the WoW and ice walk through graphs were similar, as opposed to the two cards' performance in each graph.

In the ice walk through it's possible that there is a limiting factor between the two cards, although both cards were pretty respectable on this level. They both did pretty well on the harbor level as well. Crysis in general actually plays pretty well on the X2, but there are a few areas where it really suffers. Another point where the X2 showed some really low fps is in the final cut scene of the game. I didn't really think this was relevant to FRAPS because it's not actual game play, but the X2 showed fps between in the low teens while the GTX 280 kept above 20. It's a cut scene though, so it really isn't very important to me.

I think the similar areas of reduced frames in the outland flyer graph is simply because the entire run is a scripted sequence, so the events happening on screen are exactly the same for each run. Also, notice that this graph represents a longer bench than most of the others, so the graph isn't as granular in detail. You can also see that I have vsync enabled because they both tend to hit 60fps 'ceiling' pretty often.

 

[nitromullet]

Originally posted by: Aikouka
Nitro, I was wondering something about WoW and the 4870 X2. I tend to play two WoWs at once (separate monitors) and was wondering if the X2 would still keep using a single GPU or would it be smart enough to divvy up the tasks? With the lackluster amount of memory on the normal 4870, running two WoWs at once is a huge memory monster (I assume they're maxing out the vram and both use around 600+ MB of RAM a piece with my 4870).

So in other words, I guess would you notice a difference running two instances of WoW at the same time?

People may wonder why I run relatively high-end cards just to play WoW (as I don't play any other game), but playing both in windowed mode (lower framerate because of Aero still running) I can easily hover around 30-50 fps on each system with AA set to 1x (I tried it maxed out at 8x, but that's also when I had the huge slowdown that I discussed in another thread, so I set it back). The new expansion will be upgrading the shadows to real-time shadow instead of the current blob system, which will require even more graphic prowess and be even more taxing with two open at once.

If the X2 will intelligently separate the tasks... one WoW per GPU (although I doubt it'd be that clean), then it might be worth an upgrade.

I don't know the answer to this... I have not tinkered with mutli-monitor with either card.

However, I would guess that if you really wanted to run WoW on two screens, your best bet would be to run two separate cards not on CF or SLI.

 

[chizow]

Nice job with the comparison, I especially enjoy the time vs. FPS graphs as it shows details most reviews do not. One thing I would've liked to have seen however is the frame times when the 4870X2 drops the lowest. It looks like the 4870X2 is showing microstutter in Crysis but that was taken at a random point, I'd be interested to see what the interval is at the lowest points in Crysis and WoW.

 

[nitromullet]

Originally posted by: chizow
Nice job with the comparison, I especially enjoy the time vs. FPS graphs as it shows details most reviews do not. One thing I would've liked to have seen however is the frame times when the 4870X2 drops the lowest. It looks like the 4870X2 is showing microstutter in Crysis but that was taken at a random point, I'd be interested to see what the interval is at the lowest points in Crysis and WoW.

I might be able to hook you up with when I get home this evening. The hardest part about showing the difference in intervals is that you really have to take a very small sample. If you don't, the graph has so many points it just looks like a straight line. To show that 'saw toothed' shaped graph you can really only graph 30-50 points, which in reality only represents about 1 or 2 seconds of gaming.

 

[Aikouka]

Originally posted by: nitromullet
I don't know the answer to this... I have not tinkered with mutli-monitor with either card.

However, I would guess that if you really wanted to run WoW on two screens, your best bet would be to run two separate cards not on CF or SLI.

Hmm I've thought about that... how does that affect other things such as just normal activities? Is it pretty much no different than what I have now (the "dual view"-esque setup)? I'd also have to upgrade my motherboard to go with two cards, so I'd probably wait for Nehalem (if reviews pan out well) to do that... no sense in wasting the money now .

 

[Dorkenstein]

Thanks for doing this research. At first I was confused and thought about getting another nvidia card but have decided against it.

 

[ShadowOfMyself]

Originally posted by: nitromullet

Yep, it's *not* a review of the two cards. If you actually read what I posted, you would know that I looked at both GRID and COD4. Both cards played both games well, and I only made FRAPS runs games where I noticed a significant difference between the two.

Yes I did read it, but my point stands, that way of doing an analysis is completely skewed, much like HardOCP

Obviously, the games in which you will notice the most difference are the ones in which the 4870 X2 doesnt scale, because when it does, the fps are so high that unless you have superhuman vision you wont notice it

Its like saying

"So guys, check this out to see when the 4870 X2 performs bad, if you wanna know when it actually performs how it should, go look at the reviews"

Which is fine if thats what you want, there are plenty of reviews out there having the 4870 X2 stomp the GTX, so we dont really need another one

 

[HOOfan 1]

Originally posted by: ShadowOfMyself

Obviously, the games in which you will notice the most difference are the ones in which the 4870 X2 doesnt scale, because when it does, the fps are so high that unless you have superhuman vision you wont notice it

That is kind of the point....the GTX 280 will run pretty much any game out there as high as you want to run it, except for Crysis, but the HD4870X2 is even worse in Crysis than the GTX 280.

 

[nitromullet]

Originally posted by: ShadowOfMyself

Originally posted by: nitromullet

Yep, it's *not* a review of the two cards. If you actually read what I posted, you would know that I looked at both GRID and COD4. Both cards played both games well, and I only made FRAPS runs games where I noticed a significant difference between the two.

Yes I did read it, but my point stands, that way of doing an analysis is completely skewed, much like HardOCP

Obviously, the games in which you will notice the most difference are the ones in which the 4870 X2 doesnt scale, because when it does, the fps are so high that unless you have superhuman vision you wont notice it

Its like saying

"So guys, check this out to see when the 4870 X2 performs bad, if you wanna know when it actually performs how it should, go look at the reviews"

Which is fine if thats what you want, there are plenty of reviews out there having the 4870 X2 stomp the GTX, so we dont really need another one

That is not entirely correct. In the four days I've owned the card, I've found three games at pretty specific settings that show a difference between the two cards, so I looked into them a little bit more closely. Two of them favor the GTX, while the third one favors the X2.

Understand that I didn't go looking for low frames on either card.

- Since Crysis has come out, it has pretty much been a standard part of checking out performance for every new card I've purchased.

- I've been playing WoW for a while now, so I know the parts of the game that push video cards. Surprisingly, no one has made the point that the fps for both cards with 8xAA are actually pretty sad considering the price of these cards and the game we're talking about.

- I played Quake 4 on an 8800GTX back when it came out, and I was pretty disappointed by its lack of playability with 8xAA. When I got my GTX 280 I gave it a shot, and was still disappointed. So, I figured I'd give the X2 a try, and it did a much better job.

 

[Tempered81]

Nice review Nitromullett. You know 4870x2 > gtx280 right?

 

[Janooo]

Which one are you going to keep?

 

[Wreckage]

Originally posted by: ShadowOfMyself
Gonna get flamed but that looks very Nvidia skewed to me, specially the choice of games... Its like you specifically picked games that dont scale or are nvidia favorites (like crysis)
If you had tested Call of duty 4, Race grid, etc youd see the difference

But it comes down to - if you wanna play THOSE games, sure buy an nvidia card, if you wanna play the rest, buy the X2

Crossfire has scaling issues in many games. It's not skewed, it's reality. This is why it's generally better to go with a single CPU solution if a fast enough one is available. Clearly the GTX280 holds up well enough and as a bonus costs less.

 

[NitroTurtle]

First off, thanks for the review nitromullet (my nitro brother!). I personally found this info very useful, as I spend about 95% of my gaming time playing WoW (GM of Monolith). I'm currently running an 8800GTS 512MB, and while it works reasonably well, I do get occasional slowdowns during raids, even with no AA. The rest of my machine is very nice, but I feel my video card is the weakest link, and I've been looking for a suitable replacement. After reading the recent reviews, I was sold on the 4870x2 and planned to get one after the price dropped just a bit.

I play at 1920x1200 in a maximized window (dual 2707WFP), and I would really like to be able to crank up the AA and still cap my framerate during raids. I'm also currently playing the WotLK beta, and the performance hit with the new shadows is definitely noticeable. It's not uncommon to see it sit around 30-40 fps in Northrend, and that's only using the 2nd of 4 shadow settings with no AA. So, I feel it's time to upgrade and would like to get the best card money can buy. I'd prefer to stay away from multi-card solutions, since with SLI I lose the ability to run dual monitors while gaming and even with CF I would need another MB as I'm currently using a P35 board.

So anyways, which card would you recommend solely for WoW? Like I said, I was sold on the 4870x2 and just waiting for the price to drop a bit after the initial release. After reading this though, I'm having some doubts. The money isn't that big of an issue, I just want the best performance possible right now with a single card.

Thanks again for your review. It's been impossible for me to get any performance data on either card for WoW, as it's never used in any benchmarks. I realize it's not the newest tech, but with the sheer number of people playing the game, you'd think someone would test it. But it seems people think you can max it with a 6800GT, and that's simply not the case once you get into the higher resolutions and settings in a raid environment.

 

[nitromullet]

Originally posted by: jaredpace
Nice review Nitromullett. You know 4870x2 > gtx280 right?

That is what I keep hearing...

Originally posted by: Janooo
Which one are you going to keep?

To be honest, I'm not sure yet. I know my sig says GTX 280, but the 4870 X2 is actually installed as I type this.

Originally posted by: NitroTurtle
First off, thanks for the review nitromullet (my nitro brother!). I personally found this info very useful, as I spend about 95% of my gaming time playing WoW (GM of Monolith). I'm currently running an 8800GTS 512MB, and while it works reasonably well, I do get occasional slowdowns during raids, even with no AA. The rest of my machine is very nice, but I feel my video card is the weakest link, and I've been looking for a suitable replacement. After reading the recent reviews, I was sold on the 4870x2 and planned to get one after the price dropped just a bit.

I play at 1920x1200 in a maximized window (dual 2707WFP), and I would really like to be able to crank up the AA and still cap my framerate during raids. I'm also currently playing the WotLK beta, and the performance hit with the new shadows is definitely noticeable. It's not uncommon to see it sit around 30-40 fps in Northrend, and that's only using the 2nd of 4 shadow settings with no AA. So, I feel it's time to upgrade and would like to get the best card money can buy. I'd prefer to stay away from multi-card solutions, since with SLI I lose the ability to run dual monitors while gaming and even with CF I would need another MB as I'm currently using a P35 board.

So anyways, which card would you recommend solely for WoW? Like I said, I was sold on the 4870x2 and just waiting for the price to drop a bit after the initial release. After reading this though, I'm having some doubts. The money isn't that big of an issue, I just want the best performance possible right now with a single card.

Thanks again for your review. It's been impossible for me to get any performance data on either card for WoW, as it's never used in any benchmarks. I realize it's not the newest tech, but with the sheer number of people playing the game, you'd think someone would test it. But it seems people think you can max it with a 6800GT, and that's simply not the case once you get into the higher resolutions and settings in a raid environment.

Glad you enjoyed it. I understand where you are coming from with WoW benchmarks too. Because it's an MMO on an older engine, it doesn't get the attention it deserves in video card reviews.

As far as a recommendation... I didn't really want to get into that, but honestly I prefer the GTX 280 in WoW right now because I like my transparency AA. I will say this though, the 4870 (even non-X2) does play WoW really well at 4xAA.

You probably should also be aware that the AA mode I run in WoW is only available with nHancer, as NVIDIA has removed support for it in the NV control planel. From nHancer, just select enchance application AA, Combined mode 8xS, Super Sampling, Transparency AA, and then select 8x AA in the game. I use this mode because it offers the best image quality with the least performance hit. I did try different AA modes with the Radeon, but I did not see any differences in performance.

I also don't have beta access to WotLK though, so I have no idea how that will be. I do know that WotLK is no longer a TWIMTBP NVIDIA title and that Blizzard has gotten in bed with ATI, so it's possible that WotLK is going to be optimized for ATI.

 

[nitromullet]

Originally posted by: nitromullet

Originally posted by: chizow
Nice job with the comparison, I especially enjoy the time vs. FPS graphs as it shows details most reviews do not. One thing I would've liked to have seen however is the frame times when the 4870X2 drops the lowest. It looks like the 4870X2 is showing microstutter in Crysis but that was taken at a random point, I'd be interested to see what the interval is at the lowest points in Crysis and WoW.

I might be able to hook you up with when I get home this evening. The hardest part about showing the difference in intervals is that you really have to take a very small sample. If you don't, the graph has so many points it just looks like a straight line. To show that 'saw toothed' shaped graph you can really only graph 30-50 points, which in reality only represents about 1 or 2 seconds of gaming.

Well, I was wrong (very wrong) about that the graphs for the intervals would look like for the entire run. I added complete interval graphs for all three of the Crysis runs.

 

[BFG10K]

Originally posted by: ShadowOfMyself

Gonna get flamed but that looks very Nvidia skewed to me, specially the choice of games... Its like you specifically picked games that dont scale or are nvidia favorites (like crysis)

No, he picked games where he noticed a difference between the cards and graphed the results. That the 4870X2 is generally less reliable and consistent is the nature of multi-GPU, not him being ?biased?.

Originally posted by: ShadowOfMyself

But it comes down to - if you wanna play THOSE games, sure buy an nvidia card, if you wanna play the rest, buy the X2

Actually what it comes down to is if you want a more robust solution then go single card whether the color is red or green.

 

[BFG10K]

Originally posted by: nitromullet

I find it interesting actually that no one has commented on the lackluster performance of the GTX 280 in Quake 4 w/ 8xAA. This was absolutely abysmal IMO for the age of the game/engine and the level of video card.

Something is up with nVidia and Quake 4. Even on my 8800 Ultra I always noticed severe slowdowns when shadows were enabled and even 1-2 enemies were onscreen, even with regular 4xAA.

Even dropping to 1600x1200 with 4xAA I?d be getting 70-80 FPS averages in SP benchmark runs but as soon as an enemy appeared on the screen (and hence generated shadows) the framerate would plummet to 25-30 FPS at times. I?m guessing the dips in your graph are when enemies where onscreen.

Disabling shadows ?fixed? the issue and I always thought it was a game problem, but based on your results I'm starting to think it's an nVidia driver problem.

When most reviewers benchmark Quake 4 they use MP benchmarks which don't have shadows and I believe that?s why the problem was never detected. Prey is far more demanding than Quake 4 yet I never experienced such slowdowns at the same settings on the same card.

 

[Aikouka]

Originally posted by: NitroTurtle
Thanks again for your review. It's been impossible for me to get any performance data on either card for WoW, as it's never used in any benchmarks. I realize it's not the newest tech, but with the sheer number of people playing the game, you'd think someone would test it. But it seems people think you can max it with a 6800GT, and that's simply not the case once you get into the higher resolutions and settings in a raid environment.

No one tests WoW because it is simply too hard to get a test that is static. The typical lull in framerates that I see in WoW are attributed to the massive amounts of differing data used in Shattrath (tons of different models and skins combined with a massive amount of NPCs and players). Also spell effects, from what I recall on my 8800 GTX, tend to be really bad on my framerate. It was bad enough that I could literally turn away from a brazier and gain 10 FPS even with more people on my screen. It was pretty bad in Hellfire with how the zone has the random fire spurts out of the ground. I remember sometimes when they spit up, my framerate would drop to below 10 (note, this is with two clients open, so it wouldn't be so drastic with only one and that one being full screen, not windowed-maximized).

WoW is pretty taxing on many fronts though... the sheer amount of memory used... the amount of data needing to be transferred off the source (HDD) when switching areas and such.

 

[videogames101]

Originally posted by: tboo
I just upgraded from a GTX 280 to a 4870x2. I have a 30 inch monitor & run at 2560x1600 & I do notice a difference. My main gripe on this new card is its LOUD as hell & puts out way more heat the the GTX 280. I do have everything crammed in an Antec 900 gaming case. Maybe I need to upgrade to a bigger full size case?

I love my 1200.

 

[praesto]

Just take nitromullets results for what they are. He isn't showing, nor trying to show us a definite face-off between the hd 4870x2 and gtx 280. As he mentions from the start, he has chosen to benchmark the games where he noticed a difference. And that is really the key here. With a monitor only displaying 60 fps at max, what difference does it make that a 4870x2 renders 160 fps where a gtx 280 renders 90? He won't notice the difference, aka he doesn't benchmark the game. My point is, the hd 4870x2 for sure is better in many games fps wise, as you can see in tons of reviews, but nitromullet doesn't have the same aim as the reviewers.

 

[chizow]

Originally posted by: nitromullet

Originally posted by: nitromullet

Originally posted by: chizow
Nice job with the comparison, I especially enjoy the time vs. FPS graphs as it shows details most reviews do not. One thing I would've liked to have seen however is the frame times when the 4870X2 drops the lowest. It looks like the 4870X2 is showing microstutter in Crysis but that was taken at a random point, I'd be interested to see what the interval is at the lowest points in Crysis and WoW.

I might be able to hook you up with when I get home this evening. The hardest part about showing the difference in intervals is that you really have to take a very small sample. If you don't, the graph has so many points it just looks like a straight line. To show that 'saw toothed' shaped graph you can really only graph 30-50 points, which in reality only represents about 1 or 2 seconds of gaming.

Well, I was wrong (very wrong) about that the graphs for the intervals would look like for the entire run. I added complete interval graphs for all three of the Crysis runs.

Wow, they're really all over the place throughout. I saw some in there very close to 100ms! Also I noticed that Vsync looks to be enabled in a few of the tests. Is that Vsync + Triple Buffering or just Vsync, and how is it enabled? I think CPS570, myself and others would be interested if enabling both is still inconsistent/broken with AFR CF/SLI solutions.

Anyway tough decision for you I imagine, deciding which to keep. Nice thing about the X2 this time around is it doesn't suffer from lower frame buffer and single GPU performance is still quite good in games that do not scale, but its clear all of the problems with multi-GPU still exist.