[Albuquerque]

Anand's results have me not believing though simply because I get better results on my system by a sizable amount over their 8800GTS 512. 4 FPS is nothing to forget about when we're down to the 30~ levels. That is a canned benchmark, so why the large difference?

I too get FAR better performance than they do, but I own a pair of 3870's in crossfire. So I also wonder what the story is with that review...

[Geo]

Dunno why [H] seems to get worse results than TR and Anand. Maybe something to do with the "driver tango" AMD was apparently engaging in during the review period, which is never fun for anyone.

I'll be curious to see how mobo compatibility holds up as this thing actually gets into the wild on a bunch of platforms. Will it work just as well on Intel and particulary NV chipsets? I seem to recall GX2 having some issues.

Man, another 10.5" board. I'm not putting another 10.5" board in this case, it was too traumatic last time. And this case is only 2 years old, and pretty nice. I'll have it for several more years in all probability. Even 10.25" would be a big help. Ah well.

[Albuquerque]

I would like to see some benchies of a pair of individual 3870's in crossfire put up against a single 3870X2. I want to know if the "new" package makes any differences, pro or con. I want to assume it's all down to the slower memory speeds on the X2, but I am starting to wonder if there are CF issues with the X2 that don't exist with the individual cards...

Some benchmarks of this card just fly in the face of my pair of 3870's; I'm not sure if it's an interface difference (I'm on an X38 board with the dual PCI-E 2.0 16x slots), a processor difference (4.3Ghz Wolfdale) or an OS difference (I'm on Vista64 Ultimate).

For example: the 3Dmark06 scores that are broken out to SM2 / SM3... My individual scores seem to be far beyond any of the X2's I've seen reviewed. The X2's are getting ~5700 / ~6200 respectively; mine are hitting ~8000 / ~9000. This difference can't JUST be a memory speed issue, so what else is holding these back? I know that's a terrible benchmark for comparison, but I'd assume my Sm2 / Sm3 scores should be at least comparable even if the total score won't be.

[wishiknew]

Bummer. That computerbase review used 5 different drivers for ati and 4 for nvidia cards.

[Shark Sandwich]

Originally Posted by Albuquerque

I would like to see some benchies of a pair of individual 3870's in crossfire put up against a single 3870X2. I want to know if the "new" package makes any differences, pro or con. I want to assume it's all down to the slower memory speeds on the X2, but I am starting to wonder if there are CF issues with the X2 that don't exist with the individual cards...

Check out Tech Report's review http://techreport.com/articles.x/13967

It's interesting, the 3870x2 comes out significantly ahead of the 3870 Crossfire in Call of Duty 4 and Half-Life 2, it's about equal in Unreal Tournament 3, and gets beat by Crossfire in Crysis.

I wonder if it's solely due to the differences in clock/memory speeds between the two, or if there is something else going on here.

[Albuquerque]

Yeah, I did see that. Interesting results, not sure what to make of them... It is true that the X2 has a higher default GPU speed than a standard 3870 card; but the X2 also has lower memory speed. So that might explain why some apps tie, others lose, and yet others win depending on what method you choose to side with.

[Skrying]

The cores in the X2 are clocked 50Mhz faster than those in the 3870 (825 to 775), the memory for the X2 is only at 900Mhz compared to 1125Mhz on the normal 3870. I would assume in games (and benchmark setups) where the core speed has a greater effect on the results that this would allow the X2 to perform somewhat noticeably faster than that of the 3870. This could also help explain away some of HardOCP's results which often use higher levels of AA when possible in their "scheme" of benchmarking. Still, there are some very large gaps in the differences that can not be fully explained by these relatively slight frequency differences.

The other important aspect here has to be the drivers. Are 3870 Crossfire pairs using current drivers? There is a great possibility that the latest driver AMD provided for testing is much better in certain cases than current available drivers. Though Tech Reports setup shows that they were using the same driver on all AMD cards.

As for my issue with Anandtech using a cut scene for their "testing": The cutscene likely doesn't (well I know it doesn't) represent true ingame FPS, and therefore the differences could easily appear to be much greater than they actually are. A "canned" benchmark becomes useless when it doesn't represent at all the experienced frame rate in game. I have no issues with using a games built in demo recording features and using that same demo in all tests but to use a cut scene is a bit shady and speaks volumes about the lack of care put into testing. I know there are time constraints but all other sites I've read so far went well beyond what it seems Anand did.

[Silent_Buddha]

Not that I'm interested "yet" in going multi-GPU, but a couple interesting bits I've taken fromt his that make it far more attractive than SLI...

1. Extended desktop via multiple monitors on Crossfire systems (not just the 3870x2). And in addition...extended desktop without having to match resolutions on all monitors. IE - It would be possible to do my setup of a 30" monitor at 2560x1600 flanked by 2 x 20" monitors in Portrait view at 1200x1600, and be able to game at 4960x1600 assuming the game supported custom resolutions. Now, having the horsepower to drive that resolution probably doesn't exist for newer games, but it would be a fun experience in older games.

2. If you already have 1 or 2 3870's...you can pair them up with a 3870x2 for Tri or Quad crossfire. IE - You're investment in 3870's doesn't go to waste when you buy a 3870. That's pretty cool. The downside? Drivers for this apparently have been pushed back to March.

Now if only they could come up with something better than AFR. /sigh.

Regards,
SB

[John Reynolds]

The "3dfx way" of evaluating video cards is DEAD. It did have its time and place, but we are beyond that now. Any person using those methods to influence your video card purchase is likely irresponsible in doing so. You might even consider them liable. And I think that is going to come bubbling to the surface more and more as the industry matures.

Defending your test methodology is one thing, but this rhetoric is just ridiculously overdramatic.

Glancing over some of the reviews, the performance is pretty much what I think most of us expected. Hopefully we'll see some nice gains over the coming months as the scaling issues are worked out/improved on.

[Albuquerque]

Generally agreed with performance being "approximately" what was expected -- the grand total of all reviews shows the X2 to perform pretty much like a standard CF'd pair of individual 3870's. And of course, it comes with the tradeoffs of heavy driver dependance and games that don't always "work".

Nice thing was, it seemed like ALL the games "worked" in some fashion or another, albeit some got less gains than others. The people who would buy a card like this are also the people who likely run high resolutions with high AA and AF; this card seems at home in those situations in nearly all cases.

I'd like to get ahold of this beta driver to see what it does with my standard pair of 3870's now

[Geo]

Btw, and tangentially, has anyone seen a power adapter to turn two 6pin into one 8pin pcie power yet?

[Albuquerque]

Would Single-6-to-Single-8 work for you Geo?

Thanks go to Brolloks over on OCForums.com for that link btw...

[Arnold Beckenbauer]

Originally Posted by wishiknew

Bummer. That computerbase review used 5 different drivers for ati and 4 for nvidia cards.

It's normal.

http://www.google.com/translate?u=ht...&hl=en&ie=UTF8

Anmerken we need at this point still, that all today's dual-GPU technology (both Crossfire and SLI) in a few games with an annoying problem, as soon as the frame near or below 30 frames per second. In diesem Fall kann es passieren, dass der Spielablauf sehr ungleichmäßig und stockend wirkt, obwohl der 3D-Titel eigentlich noch spielbar sein müsste. In this case, it can happen that the flow of play very uneven and stagnant effect, although the 3D titles still have to be playable. Worin genau die Ursache für dieses Phänomen liegt, bleibt weiterhin unbekannt. What exactly is the cause for this phenomenon is still unknown.

I love Google.
But this is the problem, which makes all that SLI and Crossfire stuff useless.
http://www.forum-3dcenter.org/vbulle...d.php?t=371844
http://translate.google.com/translat...hl=de&ie=UTF-8
(I know, it's Google translation, but I hope you can read it and understand.)

[Albuquerque]

I'm not sure I get why that link insinuates that "SLI and Crossfire stuff is useless"

The translation is pretty rough, but the scaling (in games that aren't CPU bottlenecked) looks to be very good for a 3-card system. You never expect a 100% increase per-GPU-added, but obviously quite a few REAL games continue to see admirable gains going from two -> three cards.

So help me out -- which piece did I miss? I saw some framerate timing stuff that didn't make much sense; are they talking about lag or something?

[Squilliam]

Originally Posted by Geo

Btw, and tangentially, has anyone seen a power adapter to turn two 6pin into one 8pin pcie power yet?

You can use the 6 pin in the 8 pin socket. The 8 pin is only needed for overclocking etc As I recall.

[AlexV]

Originally Posted by Squilliam

You can use the 6 pin in the 8 pin socket. The 8 pin is only needed for overclocking etc As I recall.

It's only used in order to get the OverDrive tab in the CCC. You can OC even with 6+6, but through other apps. Or you can do some putzing around and trick the card into believing you have 6+8.

[AnarchX]

Originally Posted by Albuquerque

So help me out -- which piece did I miss? I saw some framerate timing stuff that didn't make much sense; are they talking about lag or something?

Inhomogeneous frame-output-times, so 40 measured FPS feel in some games like 20 real.

Here for example:

Frame, Time (ms)
1, 0.000
2, 11.231
3, 52.277
4, 62.364
5, 110.195
6, 120.274
7, 167.977
8, 178.608
9, 225.865

30 FPS Avg, but every second FPS, which you can feel, is only 20FPS.

Real(Single-GPU) 30 FPS would be ~33ms -> ~66ms -> ~99ms -> ...

[Arnold Beckenbauer]

Originally Posted by Albuquerque

I'm not sure I get why that link insinuates that "SLI and Crossfire stuff is useless"

... .

So help me out -- which piece did I miss? I saw some framerate timing stuff that didn't make much sense; are they talking about lag or something?

The problem are these different frametimes between single frames. Your framecounter says you play with 60 fps, but it feels like 30 fps.

[Albuquerque]

Originally Posted by Arnold Beckenbauer

The problem are these different frametimes between single frames. Your framecounter says you play with 60 fps, but it feels like 30 fps.

Ahhhhh, I saw the numbers but the translation left me scratching my head. I shoulda just looked AT the numbers, as that's what I ended up just doing and it now makes sense.

This looks like a rendering artifact from AFR -- a supertile or scissor mode wouldn't be capable of such odd timing. At least, I can't make it work in my head

[Kanyamagufa]

On one hand, I'm particularly impressed with how well Crossfire seems to be improving when compared to SLI - many benchmarks on a number of sites show that Crossfire, whether it's utilized through a X2 or two cards, is scaling excellently. Very impressive in my opinion, especially considering the state Crossfire was in just six months ago.

On the other hand, reading how many driver releases some of these sites had to deal with was a little unnerving to say the least...

Oh well, regardless of all that I still think it's great AMD is working this hard on their drivers.

[Albuquerque]

Originally Posted by Slyne

Which brings me to the main point of my post, WHERE IS BEYOND3D's REVIEW?
3870/3850 are still MIA, as are 8800GT/GTS512. Is the site going to be renamed BeyondReviews now? I was always looking forward to B3D's different take on GPU reviewing. I totally appreciate articles such as the one on PerfHUD, but is it incompatible with reviewing hardware?

I know the staff are under some pretty serious time and money constraints, so I wouldn't hold this against them.

If I had some different video hardware to test with (8800 of any sort, maybe a 2900 or two) then I'd love to build a review with my 3870's. I could do Vista64, Vista32, even Xp32. I've got more than a few apps to test with; albeit I'm sure there's more to be had.

[Slyne]

Originally Posted by Albuquerque

I know the staff are under some pretty serious time and money constraints, so I wouldn't hold this against them.

Time, nothing I can do about it.
Money, I'm not opposed to helping, given the option. I can't be the only one.

[swaaye]

Don't forget about the Digit-Life guys. They always put together pretty in-depth reviews over there. I especially like how thorough they are with showing you photos of the cards and their coolers.

3870X2 is sure a geometry processing monster, btw. Too bad it falters on pixel processing as usual.
http://www.digit-life.com/articles3/...0-3-part1.html

[Albuquerque]

Originally Posted by swaaye

Don't forget about the Digit-Life guys. They always put together pretty in-depth reviews over there. I especially like how thorough they are with showing you photos of the cards and their coolers.

3870X2 is sure a geometry processing monster, btw. Too bad it falters hugely on pixel shader processing as usual.
http://www.digit-life.com/articles3/...0-3-part1.html

Duly noted, adding to the pile right now!

[ECH]

...