3870x2 reviews are here and where to buy

[Munky]

Originally posted by: Syntax Error
I guess Nvidia's G80 and the 8800GTX/Ultra were just that good that AMD barely beats it down this time. A year later.

*Sigh* Sticking with the 8800GT, price-to-performance FTW.

The g80 was so good that even Nvidia will need 2 cards sandwiched together to beat it.

 

[apoppin]

Originally posted by: Rusin

Originally posted by: Sylvanas

Originally posted by: Rusin
dennifloss:
Well.. you don't need double the frame rate for cut scenes. It would have been nice if AnandTech would have tested gameplay-performance :/

Where did Anand say they benched cut scenes? And as if the results taken don't imply the real world ramifications. If you are referring to kyles' defence of the Hard OCP article chanting 'canned benches' everywhere and that it would take too long to bench more than 4 games then you really have to look at their reviews....most of the time they use different IQ settings to compare totally different cards, they use a different driver revision than the one AMD held the launch back for aswell (the one Anand used), this gives no real indication of what we want to know.

Well for example COD4 test is pretty far from actual game play (smoking scene). Thing is that these scenes run differently than actual game play.. If noticed that in UT3 these cut scenes work better with my rig than actual game play.
----

Yes they are testing with different IQ settings. They are looking for best playable settings which gives better picture of what that performance difference is in real life situation.

----
Drivers:

Anandtech:
8-451-2-080123a

HardOCP:
8-451-2 + latest crysis fixes

HardOCP has the most useless benchmarks - anywhere

 

[Munky]

Originally posted by: Rusin
Yes they are testing with different IQ settings. They are looking for best playable settings which gives better picture of what that performance difference is in real life situation.

And yet, I find their benches useless, because:
1. I don't necessarily agree with them on what constitutes "playable" fps in each game
2. IF I were to turn down settings to get playable fps, I might choose different settings from the ones they chose

 

[dennilfloss]

First price quote I've seen in Canada.

http://www.extreme-pc.ca/showp...&menu2id=99&menu3id=39

 

[happy medium]

Originally posted by: dennilfloss
First price quote I've seen in Canada.

http://www.extreme-pc.ca/showp...&menu2id=99&menu3id=39

They have two versions 875 core, 960mem and 825 core, 900 mem at 539$ and 500$.
Seems they are making overclocked versions.

Added to list thanks, Dennifloss

 

[dennilfloss]

Yup, they sell overclocked versions. I am currently using their OC version of the Diamond 3870 (see rig).

 

[nanaki333]

anybody know if you can run a 3870X2 card with a regular 3870 and it'll act as 3 cards in crossfire, or do you have to run 2x 3870x2?

 

[happy medium]

I read somewhere you can crossfire with single gpu's (3870's) or another x2 with a driver update sometime in March. I can't find it now.



Pole added.

 

[olmer]

Yes we ?need? a good product from AMD ? i still think the correct pricing can make it a winner over 8800GTX. Unfortunately it is very unlikely - £320 in the UK, where GTX can be bought for £235...

 

[DerekWilson]

Originally posted by: Rusin

Originally posted by: DerekWilson

Originally posted by: Rusin
dennifloss:
Well.. you don't need double the frame rate for cut scenes. It would have been nice if AnandTech would have tested gameplay-performance :/

AnandTech does not benchmark cut scenes in any of the games tested here.

Quote from your review:
"A surprisingly successful FPS on the PC, Call of Duty 4 also lacks any sort of in-game benchmark so we benchmark the cut scene at the beginning of the first mission. We start the frame rate counter as soon as we're in the helicopter and stop once the man in the chair gets shot."

hmm ... Anand must have changed the test method on that one since I was running them.

 

[apoppin]

Originally posted by: olmer
Yes we ?need? a good product from AMD ? i still think the correct pricing can make it a winner over 8800GTX. Unfortunately it is very unlikely - £320 in the UK, where GTX can be bought for £235...

it looks like it may be a really good deal
-in a couple of months when the new GPUs come out

 

[WaTaGuMp]

Not impressed enough to buy one. :brokenheart:

 

[Mem]

Originally posted by: olmer
Yes we ?need? a good product from AMD ? i still think the correct pricing can make it a winner over 8800GTX. Unfortunately it is very unlikely - £320 in the UK, where GTX can be bought for £235...

OCuk have the £299 HIS model in stocklink.

 

[*kjm]

I?m just darn sick of all these retailers marking up the cards $50+ MSRP for both ATI and Nvidia! I won?t buy till they drop to MSRP or below.... I'll have to sit on my 1950pro for some time yet.

 

[WaTaGuMp]

Frys is also showing the Diamond 3870X2 at 449.00.

 

[happy medium]

Originally posted by: WaTaGuMp
Frys is also showing the Diamond 3870X2 at 449.00.

Thanks I'll add it to the list.

 

[happy medium]

Originally posted by: *kjm
I?m just darn sick of all these retailers marking up the cards $50+ MSRP for both ATI and Nvidia! I won?t buy till they drop to MSRP or below.... I'll have to sit on my 1950pro for some time yet.

MSRP is 449.00$ for 3870x2. I seen a few places selling them for that.

 

[Demoth]

Keep in mind if buying his card that indications are you need PCIe 2.0 to get proper performance. If your not planning on a MB upgrade in the near futue, might want to consider that. Plus, as has been pointed out, speed increase is heavily dependant on individual games having correct driver support. This could be a future issue, especially for less popular titles down the road, where you could be limited to below single card speeds. You will also need a good quality PSU with a lot of juice coming off the 12V rail under load and noise levels will be relatively high.

A good card overall, but a niche card for enthusiasts mainly.

 

[keflex]

Originally posted by: apoppin
HardOCP has the most useless benchmarks - anywhere

Agreed! Benchmarking a game based on how it actually performs in-game is totally useless!

Now, let's back to running through empty levels or watching someone smoke a cigar for 30 minutes; that's where the real truth lies!

 

[schneiderguy]

Originally posted by: Rusin

Originally posted by: Sylvanas

Originally posted by: Rusin
dennifloss:
Well.. you don't need double the frame rate for cut scenes. It would have been nice if AnandTech would have tested gameplay-performance :/

Where did Anand say they benched cut scenes? And as if the results taken don't imply the real world ramifications. If you are referring to kyles' defence of the Hard OCP article chanting 'canned benches' everywhere and that it would take too long to bench more than 4 games then you really have to look at their reviews....most of the time they use different IQ settings to compare totally different cards, they use a different driver revision than the one AMD held the launch back for aswell (the one Anand used), this gives no real indication of what we want to know.

Well for example COD4 test is pretty far from actual game play (smoking scene). Thing is that these scenes run differently than actual game play.. If noticed that in UT3 these cut scenes work better with my rig than actual game play.

Isn't the point of benchmarking the cards measuring how they compare to each other? Results taken from a canned benchmark OR a gameplay benchmark are valid to compare cards to each other. If Card A gets 40 fps in a cutscene and Card B gets 20 fps in the same cutscene running a gameplay bench isn't going to magically switch the cards' positions. Card A will still be around twice as fast as Card B. When I look at a review I'm not looking for how well a card runs a game, I'm looking for which card is faster in that particular game.

Kyle and his different settings for each card method are stupid. Who made him the "expert" that decides what settings to turn up and down? Doing gameplay benchmarks on one section of a game is just as limiting as doing a benchmark on a cutscene. If Kyle really wanted to get a true number of how well a card performed in a game he should play through the whole game with FRAPS running and take the average when he finishes. At least then he'll have less time to say stupid things.

 

[hooflung]

Originally posted by: Syntax Error
I guess Nvidia's G80 and the 8800GTX/Ultra were just that good that AMD barely beats it down this time. A year later.

*Sigh* Sticking with the 8800GT, price-to-performance FTW.

I guess you didn't see that the card is really only second to the SLI GT which is 2 cards and only available in Nvidia chipsets.

X2 looks like the first 2-GPU card I've ever seen that I'd actually buy if I had the money. Considering that the drivers are getting better and better AND a boom head shot feature I think AMD is on to something.

When you get Crossfire X into the mix... sorry to say but Nvidia is in trouble.

 

[CSHawkeye]

Originally posted by: keflex

Originally posted by: apoppin
HardOCP has the most useless benchmarks - anywhere

Agreed! Benchmarking a game based on how it actually performs in-game is totally useless!

Now, let's back to running through empty levels or watching someone smoke a cigar for 30 minutes; that's where the real truth lies!

That doesn't make any sense, of course everyone wants to see how it benchmarks with the benchmarking tools, but in game performance shows us what you expect when you are using this in your computer in gaming.

 

[Nurn]

Originally posted by: Demoth
Keep in mind if buying his card that indications are you need PCIe 2.0 to get proper performance. If your not planning on a MB upgrade in the near futue, might want to consider that. Plus, as has been pointed out, speed increase is heavily dependant on individual games having correct driver support. This could be a future issue, especially for less popular titles down the road, where you could be limited to below single card speeds. You will also need a good quality PSU with a lot of juice coming off the 12V rail under load and noise levels will be relatively high.

A good card overall, but a niche card for enthusiasts mainly.

Anand says this GPU doesn't support PCIe 2.0.....

"On the board we've got two 3870 GPUs, separated by a 48-lane PCIe 1.1 bridge (no 2.0 support here guys). Each GPU has 16 lanes going to it, and then the final 16 lanes head directly to the PCIe connector and out to the motherboard's chipset."

 

[hooflung]

Originally posted by: Nurn

Originally posted by: Demoth
Keep in mind if buying his card that indications are you need PCIe 2.0 to get proper performance. If your not planning on a MB upgrade in the near futue, might want to consider that. Plus, as has been pointed out, speed increase is heavily dependant on individual games having correct driver support. This could be a future issue, especially for less popular titles down the road, where you could be limited to below single card speeds. You will also need a good quality PSU with a lot of juice coming off the 12V rail under load and noise levels will be relatively high.

A good card overall, but a niche card for enthusiasts mainly.

Anand says this GPU doesn't support PCIe 2.0.....

"On the board we've got two 3870 GPUs, separated by a 48-lane PCIe 1.1 bridge (no 2.0 support here guys). Each GPU has 16 lanes going to it, and then the final 16 lanes head directly to the PCIe connector and out to the motherboard's chipset."

I didn't read that as the card not being PCIe 2.0 but the interconnect between the two integrated gpu subsystems as not being PCIe 2.0.

 

[keflex]

Originally posted by: schneiderguy
When I look at a review I'm not looking for how well a card runs a game, I'm looking for which card is faster in that particular game.

You do understand that if synthetics shows that A is faster than B, but in real-world applications B > A, then the synthetic is pretty much useless, right? Kind of like how in vivo studies are alot more accurate than in vitro studies.

Of course, if all you do is buy bleeding edge hardware just to run benchmarks...