The Official RV630/RV610 Rumours & Speculation Thread

Quite probable!

 

I smell paper launch, just checked some webshop and not much have HD cards listed some hours before NDA expire
Geizhals have 4 hd2600 card listed 3 is in the same online shop.

Maybe its not a sign of paper launch, its a sign of no one care about this cards.

 

Active shaders 80?

 

You would be correct. 5200 was a worthless POS. A Radeon 8500 or 9000 would obliterate them, let alone a GF4 Ti. But they were marketing genius. People still buy them in droves today, but they are called FX 5500 now. Still NV34.

I think a lot of people here underestimate how little the majority of computer users care about their video card. Before FX 5200, we had GF4 MX. Before that there was GF2 MX 100. And then TNT2 M64, NV Vanta, and finally the various 2D cards with worthless 3D (S3 Virge, anything #9). I'd go so far as to say that the majority of PC users are as happy with an IGP as most gamers apparently are with integrated audio.

 

QX6600 2600XT GDDR3 2k6 sm2.0/3.0 1532/1966
QX6700 2600XT GDDR4 2k6 sm2.0/3.0 1788/2268

 

http://radeon.ru/?select=more&f=2007_06&new=95

something strange there is HD2600XT doing better than 8600GT in simple pixel/vertex shaders and worse in complex shaders

 

I'm not too good at understanding 3D Mark. Are those numbers looking good or bad right now for the 2600XT?

 

HD 2400XT Review:
http://www.hartware.de/review_720.html

translation by google

 

nothing wrong with that as it just means that 8600 is probably triangle setup limited (or something like that..) where the 2600 is not

 

It's time for The Official HD2400/HD2600(RV610/RV630) reviews thread!

http://www.ocworkbench.com/2007/powercolor/hd2600xt/g1.htm

Waiting...

 

Radeon HD 2600XT GDDR4 @ PConline translation

 

+1

http://www.hexus.net/content/item.php?item=9187&page=1

 

ATI has a 390 million transistor part now in HD2600XT, that would get absolutely destroyed in any game benchmark by 140m X850XT.

Nice.

 

I'll reserve my final judgement until at least one additional driver revision is released. Sure doesn't look that good right now though.

But the 2400 should be really interesting for HTPC's, when they've fixed the drivers problems that is.

 

Hothardware.com have a review up. I was actually surprised how low some of the figures are, even at 1280x1024 (which I consider the best resolution for the size of LCD these cards are likely to be partnered with). 10-15fps was not uncommon.

Neither nvidia or ATi midrange offerings are anything to shout about and both go to show how good a bargain the 8800GTS 320MB is.

Shame the 7900GS and 1950Pro cannot do DX10, but apart from that they really still supply the goods

 

It's hard to believe the dichotomy presented by these initial reviews. The 2600 basically fails in Hexus' review (to be fair, they said they had problems--but they still published their #s), but it's pretty decent in PCOnline's. Looking at HotHardware's piece, maybe that's b/c--IIRC--PCO didn't use AA that often.

I don't know how helpful it is to compare transistor counts of SM4 and SM2 cards. That road only leads to 256 SLI'ed Voodoos.

A thread at Ars asked to recommend a 7900GS or 8600GT as they were priced the same, and my first instinct was the GS, duhhh. Looking over some reviews, though, I was really surprised to see that the GT held its own against the GS overall, though maybe at framerates too low to call "wins." It was crushed in some titles (40% slower in CoD2), but it was decently faster in newer ones and pretty even overall. Most surprisingly, it didn't tank with AA. I say this looking at Computerbase's overall scores in the 2900XT review, and Firingsquad's and TechReport's 8600 reviews (though TR used Vista, which apparently is still a shaky playing field).

But the X1950P pretty much stomps the 8600 and 2600 in HH's test, and the 8800GTS does come out looking like a kick-ass card, but 320MB is still too little for its speed. So many trade-offs...

 

But it supports DX10.
IEEE floating point for the win! Wasn't that what consumers wanted?

It would have been very interesting to have seen what those 390 million transistors and reasonable power draw could have achieved at the DX9 level. Until Vista has taken over, and that is a long way off, DX9 is its own market just as AGP has been for a while. IMO it should have its own dedicated hardware at the new process nodes. Would possibly be quite profitable for AMD as well to attack and own the Windows XP market.

 

Not much, given ATI's recent products.

Oh, I guess the R580 line, was close to that many transistors, though half as big G71 performed just as well.

And, most of ATI's mid-range was just partly disabled R580's last gen, so you'd be looking at ~350m transistor parts all throughout the lineup last gen.

BTW, these reviews are terrible. It's amazing that the "big sites" like Anandtech mostly buy you graphs that are actually not stupendously confusing and nigh unreadable, especially when plastered with a huge watermark like OCworkbench's.

If you want to know why Anandtech is more popular than most review sites, 75% of it alone probably lies in the clarity of their graphics. Apparantly an amazingly hard thing to get right.

 

Features cost transistors. You pay in performance or in size/cost/power draw. It is a question of balance, what constitutes a good compromise at a given time. The whole point of having a dedicated ASIC doing a job in the first place is that you try to make an optimal use of resources for the job you need to get done rather than going the general route.

I'd say that the ever increasing power draws is the most obvious consequence of pushing the feature envelope since DX8, but the price is paid in performance/transistor as well.

Be that as it may, for the first time we have a situation now that these new transistors are mostly useless for those that run Windows XP. XP users would definitely be better served by another design. And they are the vast majority right now, and are likely to be a large market for years to come.

 

These cards are a disaster. You should see the posts on Hardocp. Performance is just awful. Half 8600GTS in some cases. Which itself is a terrible card.

I hope for ATI's sake that these cards are broken as per the rumors, because if they knowingly planned for this level of performance, I just dont know what to say.

4 ROPs? Really?

Oh well, supposedly the ultra low power one will sell them something to OEM's.

Umm, dont the features add performance/transistor (when used by games)? Otherwise why have them? It would be more efficient then to run the graphics on super bulked up SM2 hardware, or whatever.