Leaked BENCHMARKS from the ATI 3870X2

[yogurt_21]

Lol, I was playin with Strick94U's comment above mine but dont bother quoting the GTX sli of 17574, I can more or less hit that with a single GTS 512MB! I can only imagine those GTX's were severly bottlenecked cause I beleive there are some 20,000+ scores in the 2006 thread, although I would love to see a couple of GTS 512MB's on say a quad at 4gig.

lol you can't compare stock to oced. sure your gts can hit 17k overclocked, but a stock one wont, and stock gts sli wont either. (as sli doesn't scal as well as crossfire) also at 3.3GHZ the cards are quite a bit more bottlenecked than your single. and to see a quad 4 gig plus dual 8800gts (g92) just got to futuremark lol.

http://service.futuremark.com/compare?3dm06=4608322
and I'd gather that these cards are far from stock as well.

 

[Tatty_Two]

New gen isn't that much higher if at all higher at 256-bit
In any case anyone else have any 2560x1600 scores?

Most have the free version I spose so probably not, and hmmmmmm G92 GTS versus G80GTS on same system equates to over 4000 3D Mark 2006 points if you overclock both cards to the max without hardmodding, thats pretty significant in my book Actually, not far off the differents between G80 in Sli!

Edit: Think I missed something there, do I understand you right ....you dont think two 8800GTS 512MB at stock will hit 17000 in 3D Mark 2006? The closest I could find in HWbot that was a quad doing 3.33gig was on two 8800GTS 512MB factory overclocked cards, there were a lot of GTS setups with MUCH higher scores than this but as you said earlier, both cards and CPU's were massively overclocked, here is a copy of the extract, agreed the score would be a bit lower with the CPU running alittle slower:

3Dmark 2006 - 17935 marks - cky2k6 (XtremeSystems) 6.1 points - [ ]
Processor: Core 2 Q6600 (2.4Gh... @ 3420mhz no image
Videocard: 2x GeForce 8800 GT 512... @ 683/1000mhz
Global Rank: 344th - 3.7 points
Hardware Rank: 15th 2x GeForce 8800 GT 512 Mb - 2.4 points
Description: Q6600 @ 3420MHz - 2x 8800 GT 683/1000/1729 - tuniq 120 & vf900
Verfication: link to forumpost verification link
Scan date: 28-12-2007 04:35
Compare url: http://www.hwbot.org/compare.do?resultId=681273

 

[Tatty_Two]

lol you can't compare stock to oced. sure your gts can hit 17k overclocked, but a stock one wont, and stock gts sli wont either. (as sli doesn't scal as well as crossfire) also at 3.3GHZ the cards are quite a bit more bottlenecked than your single. and to see a quad 4 gig plus dual 8800gts (g92) just got to futuremark lol.

http://service.futuremark.com/compare?3dm06=4608322
and I'd gather that these cards are far from stock as well.

Couldnt agree more but then again, thats some of the advantages of having a single card, it does not change the fact that in reality there is not a huge amount in it, on top of that, just look at the price comparison....but yes you are right but it's only fair to let the single card overclock if the competition has 2 cards!

 

[Ati Addictive]

I forgot to ask.....does speculation suggest these will be cheaper than two HD3870's??

count on a price between 300 and 350 euro so yes

 

[yogurt_21]

yeah, I've no problem with that, I'd much rather run single than mess with dual. the dualchip card war will come down to how well they oc really, as crossifre scales better which would give the dual 3870 and initial lead, but since the g92 base gts's overclock better, it might coem out as the better value. trouble is, neither side has launched a successful dual card previously soo.... I don't really think in either game support/drivers or cost that these will be a better buy than just getting two seperate cards, or better yet one single card and a good watercooler.

 

[Tatty_Two]

count on a price between 300 and 350 euro so yes

Thats a nice price, interesting to see what good old rip off Britain will charge tho

 

[EastCoasthandle]

Source
If this is any indication of what we can expect I understand why some are excited (return of competition) and why others are so worried (return of competition) . Not because of it's sheer power but how competitive this card is using less power which is 40 watts less then GT using SLI and it will be even less using GTS in SLI.

But lets look at the scaling:

Side note: in order to see this side by side you may need a resolution of 1680x1050 and higher and have your web browser maximized.
As you can see the 3870 scales much better then it's competitor. Part of the success of the 3870 X2 will hinge on how well crossfireX drivers are in the end. Time will tell. However, by and far the whole design for X2 is much more innovative then GX2 IMO.

 

[TonyStark]

The people of Hothardware have gotten there hands on the new ATI 3870X2 card.
The card as show on the first foto contains 2 PCIe supplyconnectors, a 6 pin and a 8 pin.
The lenght of the card doesn´t look at all of the same size as the 8800 Ultra card or GS as i posted yesterday in a another thread, it is maby longer than your motherboard.

Also a Chineze site has leaked specifications of a Benchmark in 3Dmark 2006. The copy they have used contained 2 cores that ran on 775MHZ, while in the meantime the 1 GB did his job at 2250 MHZ. The only benchmark they posted has been done with 3Dmark 2006 with a resolution of 2560X1600, they established a score of 9573 points. (For the people who are wondering what a score would be on a resolution in 3Dmark 2006 of 1280x1024 will be the same score as 3870 in crossfire mode. Further of all the system contained a Intel C2D E6600 with 2 GB of DDR 2 memory. Further of all i have a foto without the cooling on it so that you can clearly see the PLX chip that exchanges data between the two GPU´s.

For the original article go to http://my.ocworkbench.com/bbs/showthread.php?p=425415#post425415

Also again many thanks to Hardware.info for getting use nice previews of the card.

Thread title is misleading. "BENCHMARKS" should be changed to "single 3DMark score" ....

 

[imperialreign]

@EastCoast - thanks for that post, man! It's nice being able to actually see the comparitive numbers side by side for once

I find it really interesting, based on those charts, that the 3870 setups appear to completely oust the 8800 GTs at a "gaming" resolution, while the 8800's seem to do better at a higher res. Anyhow, again based on those charts, it looks like any actual difference in frame rates is negligable for the most part. A 5-10 FPS difference between the cards is neck and neck, and hell, at those high FPS to begin with we wouldn't actually see the difference.

If ATI bring the new cards in around the $500 mark (as they've also stated before they no longer plan to develop anything over that price), I think they'll have a winner on their hands.

 

[strick94u]

i dont´t think you understand but the resolutions where much higher with that card than your puni little benchmark! don´t really understand why u post here!
So green actually suxs compaired to this

PS: Ati Beat the hell out of nvidia with there 3850 GPU

Let’s just say I hope the next generation of video cards would be able to outperform this generations. 2 GPU’s twice the memory buffer it is a huge card with nice numbers. I post on here because its an open forum, I don't understand your anger at such a tiny comment about color at least I am not so much a fanboy it would ever keep me from buying an ATI card, can you say the same? I usually make that decision based on reports like this one however I wait till the actual silicone is out and tested by trusted sites with drivers released with the card not built for running one or two benchmarks. That way I know for sure I know the card I am about to drop 10 percent of a paycheck on is worth buying.Now when you make a blanket statement about Ati beating the hell out of nvidia wither 3850 gpu what does that mean out perform,out sell,or did they get a 3850 and beat the hell out of Nvidia with it? Please elaborate

 

[Tatty_Two]

So bottom line, if you want 2+ independant card setups, the HD3870 is the better choice, it scales better in multi cards setups, is cheaper and the single card performance gaps betwen the 2 cards are pretty much eradicated, if you prefer just single card setups then the 8800GT is the faster card and the rest is down to how much price difference there are between the 2.

It will be interesting to see in that case, how much the GX2 beats (if it beats, but it should as scalability will be less of an issue than it is in Sli) this dual card by and what the price comparisons are.

 

[EastCoasthandle]

@EastCoast - thanks for that post, man! It's nice being able to actually see the comparitive numbers side by side for once

I find it really interesting, based on those charts, that the 3870 setups appear to completely oust the 8800 GTs at a "gaming" resolution, while the 8800's seem to do better at a higher res. Anyhow, again based on those charts, it looks like any actual difference in frame rates is negligible for the most part. A 5-10 FPS difference between the cards is neck and neck, and hell, at those high FPS to begin with we wouldn't actually see the difference.

If ATI bring the new cards in around the $500 mark (as they've also stated before they no longer plan to develop anything over that price), I think they'll have a winner on their hands.

Nice observation! A lot of people (specially those who are hooked on faster frame rates) simple don't know that most games show no difference between 33FPS and 38 FPS when that game requires a high frame rate in order to play it as intended. In this case the game gets no faster, performs no better and is no smoother then at the lower frame rate (in my experience). Another thing people forget is that each game requires a certain frame rate in order to be played properly. If you don't achieve that frame rate consistently it's pointless which card is faster.

If both achieve the frame rate needed then the faster card doesn't change the immersion of the game. For example lets say WiC needs 50 FPS to play. If both competiting cards play the game at 53 FPS and the other 58 FPS it provides no greater advantage then at 53 FPS. The immersion of how the game plays and how it reacts to you playing it doesn't change. All you did was pay more to justify why you run fraps!

IMO it's the gamings benchmark greatest illusion if you don't properly understand what the numbers mean. In most cases, it means nothing if you don't know what's generally required for each game benchmarked. Is it safe to generalize (take a guess) that all games benchmarked require somewhere between 45-50 FPS in order to run properly? It's a tough call...

 

[Ati Addictive]

Yeayy Duhh ofcourse but it is still cool to have a great card in your computer, videocards like these are just simply for benchmarking and smooth game rendering like crysis.

 

[Wile E]

Hmmm, I hope Palit will send me a couple of these for testing. I'd like to see what they can do, and if CrossfireX can be enable with 2 of them on an X38 board.

 

[imperialreign]

Nice observation! A lot of people (specially those who are hooked on faster frame rates) simple don't know that most games show no difference between 33FPS and 38 FPS when that game requires a high frame rate in order to play it as intended. In this case the game gets no faster, performs no better and is no smoother then at the lower frame rate (in my experience). Another thing people forget is that each game requires a certain frame rate in order to be played properly. If you don't achieve that frame rate consistently it's pointless which card is faster.

If both achieve the frame rate needed then the faster card doesn't change the immersion of the game. For example lets say WiC needs 50 FPS to play. If both competiting cards play the game at 53 FPS and the other 58 FPS it provides no greater advantage then at 53 FPS. The immersion of how the game plays and how it reacts to you playing it doesn't change. All you did was pay more to justify why you run fraps!

IMO it's the gamings benchmark greatest illusion if you don't properly understand what the numbers mean. In most cases, it means nothing if you don't know what's generally required for each game benchmarked. Is it safe to generalize (take a guess) that all games benchmarked require somewhere between 45-50 FPS in order to run properly? It's a tough call...

sadly, though, the whole market at this point has been built around a numbers war. There are too many consumers out there that will buy a product because it "averages" 5 FPS more than it's competitor - based solely off 3rd party testing of the products on differing 'high-end' systems . . . speaking of which, how often do you see everyone's 3rd party testing producing the same numbers as the next guys? For the most part, you don't. While card A fares exceptionally well on review site X, card B instead beats out card A in the same tests on review site Y.

poorly informed consumers, IMO, are what keep the green vx red rivalry going at this point - granted, there's a lot of loyalists to one camp or the other, but we don't go around boggarting threads until someone else flames it up . . .

 

[Makaveli]

Couldn't have said it any better myself!

Agree with you 100%

 

[EastCoasthandle]

sadly, though, the whole market at this point has been built around a numbers war. There are too many consumers out there that will buy a product because it "averages" 5 FPS more than it's competitor - based solely off 3rd party testing of the products on differing 'high-end' systems . . . speaking of which, how often do you see everyone's 3rd party testing producing the same numbers as the next guys? For the most part, you don't. While card A fares exceptionally well on review site X, card B instead beats out card A in the same tests on review site Y.

poorly informed consumers, IMO, are what keep the green vx red rivalry going at this point - granted, there's a lot of loyalists to one camp or the other, but we don't go around boggarting threads until someone else flames it up . . .

Sad but true in the end some pay more and get the same gaming experience. Some know it and are just loyalists while others just don't know better. That's why I consider some of these benchmark results an illusion because some just don't know that a higher number doesn't always mean better performance. You have to know what's required for that game first in order for that benchmark to make sense.

 

[Ravenas]

I forgot to ask.....does speculation suggest these will be cheaper than two HD3870's??

Eh, that would be great, but kinda hard to believe without seeing it.

 

[TonyStark]

8800GTX SLI 3DMark06 score (@2560x1600) = ~9.9k

CPU used: Core 2 Extreme X6800 2.93GHz

 

[TonyStark]

8800GTX vs HD3870 X2 (3DMark06 @ 2560x1600)


8800GTX + Core2 Quad (Q6700 @ 3.5GHz):
SM2.0: 3244
SM3.0: 2828

3870 X2 + Core2 Duo (E6600 @ 2.4GHz):
SM2.0: 4494 +38%
SM3.0: 4476 +58%


source: http://www.xtremesystems.org/forums/showpost.php?p=2690580&postcount=281

 

[EastCoasthandle]

8800GTX vs HD3870 X2 (3DMark06 @ 2560x1600)


8800GTX + Core2 Quad (Q6700 @ 3.5GHz):
SM2.0: 3244
SM3.0: 2828

3870 X2 + Core2 Duo (E6600 @ 2.4GHz):
SM2.0: 4494 +38%
SM3.0: 4476 +58%


source: http://www.xtremesystems.org/forums/showpost.php?p=2690580&postcount=281

Oh yeah, nearly forgot about that:

 

[erocker]

It's starting to be leaked that the card is doing 17000+ with 3d06. I'm assuming it's with a q6600.

 

[Tatty_Two]

Not quite sure of the message at post 44?.....2 different generations of cards, one three times the price of the other etc etc, if the point is to say that 2 HD3870 are faster than two 8800GTX's in Sli/Xfire then fine, go find a similar bench with the 8800GTS G92, that will be closer.....has anyone said they are not, is that relative to the thread topic? my point earlier was that 97% of the population dont buy 2 cards, they buy one, if the only way you can stack up some positives for a card is to emphasis what 3% of the population can get out of 2 cards then IMO there is little point. My other point was if I remember correctly, I pointed out that one 8800GTS G92 was close to 2 HD3870's score, which again if I remember correctly....was true, so again whats the point? unless you are saying it is better to go out and by 2 cards costing more to get one cards worth of performance.

I agree with everything you have said about FPS....gaming experience etc etc, along with the benefits of better dual card scalability but as I said, so few people in the world buy 2 cards at a time (including most people who run Sli/Xfire as the whole point of that really is to buy one and when you can afford another, buy another) that the majority of people do not judge a cards performance by how it performs in XFire.....that was all.

 

[Mussels]

hell even on this forum, a large bunch of techie nutcases... its maybe 1 in a 100 who have dual cards, and almsot every one of them has two high end cards on very expensive systems (better than mine, some by a huge amount)

Also a very big reason why you see such big benchmarks with ATI? Because the best quad core platforms only support crossfire and not SLI...

 

[xfire]

^^ you have nvidia to thank for that with the whole you can't support sli without our chips thing.