The Official RV630/RV610 Rumours & Speculation Thread

[turtle]

There's the age-old problem of what the ROPs (and TMUs) can do. One R6xx ROP might be worth two R5xx ROPs, for example.

true enough.

Also, there is a distortion in RV530 associated, I believe, with it's superfluous bandwidth. Note RV560 has ~the same bandwidth as RV530...

That distortion clouds an analysis of scaling, I reckon. e.g. RV530 is essentially 1/4 of R580, yet it has 35-45% the bandwidth (depending on GDDR3 or 4). RV560 has ~45% of the bandwidth of R580+, with ~1/2 of the pipeline capability - this almost perfectly lines up.

The R6xx range looks like it's going to scale the other way, comparing the extreme high-end 512-bit bus scaled down to RV630. e.g. I'm thinking RV630 has 1/4 the pipeline capability, but 1/5 the bandwidth. Against that, R6xx should be, if the patent documents are to be believed, more bandwidth-efficient.

It's a shame we don't have any, rumoured, handy pipeline configuration nomenclature to argue over, e.g. 16-2-20-4 ...

I suppose if we're expecting R600 to be more than 2x faster than R580, then RV630 should be ~2x faster than RV530, i.e. in RV560 territory. Maybe with more bandwidth, say ~29GB/s.

Jawed

I agree with that whole-heartedly, and glad once again you can put it into a nice quote-bite.

128*2000/8 = 32GB/s

Which adds up. Fast GDDR3 or slow GDDR4, sounds right. 2000mhz is just a guess, it could be slightly higher or lower.

32GB/s is 20% of 160GB/s, which is a R600 at 2500mhz, which again, sounds like a reasonable assumption to make. Again, it could be slightly higher or lower, depending on if it uses 2200mhz, 2400mhz, or 2800mhz GDDR4. At any rate, ~20% bw sounds about right to me. I'll guess R600 uses 2400mhz GDDR4, and 153.6GB/s. 32GB/s is ~21% of that. Good call Jawed

I too wish we had the ol' 16-1-1-1, 4-1-1-1, 4-1-3-2, and 16-1-3-1 hints of yore...That was oh, so much fun.

 

[Ailuros]

320 this and that for R600, it sounds more to me like a 4*16 thingy ie 4 clusters.

I'd rather expect 2 ALU clusters for RV630 and hopefully more than 4 ROPs. While I don't disagree that RV6x0 ROPs shouldn't be comparable to RV5x0 ROPs, the real point of interest would be for me if hypothetical 2*RV530 performance is good enough to beat it's straight competitor.

A real reason to get excited would be if someone told me that RV630= 2*G73.

 

[vertex_shader]

Source from Taiwan manufacturer said AMD schedules the release of the next-generation entry-level graphic chip RV610 in May. As its release has already a month later than its rival, together with just 64-bit memory bandwidth, RV610 becomes a spot in the market recently.

Link

Sounds cheap, but the 64bit without fast memory can be a bottleneck in performance.
I not really understand what AMD doing, maybe the cost of the cards overall lower with 64bit-high clocked gddr3 ram, than 128bit-lower clocked gddr3 ram, and the performance will be the same?

 

[Techno+]

Link

Sounds cheap, but the 64bit without fast memory can be a bottleneck in performance.
I not really understand what AMD doing, maybe the cost of the cards overall lower with 64bit-high clocked gddr3 ram, than 128bit-lower clocked gddr3 ram, and the performance will be the same?

maybe

 

[turtle]

I think it's obvious the low-end has never been really used for gaming, more-so just an add-in card to allow for a display at a decent resolution beyond IGP is capable, Aero glass, and HD playback. It doesn't really need to be anything special, it just needs to get the job done. In that respect, it may be enough for 720p or so, and if it is, that <25w power usage is the key to the whole thing. I think they're shooting for a general purpose, non-gaming card with Rv610, perhaps good enough for HTPC's or laptops, built as small as humanly possible for low power while meeting their goal to maximize profit.

Any guesses on how big it will be? I'm guessing 40-50mm2.

How about Rv630 being <80mm2? Think there's any truth to that?

Doing straight 1/4 and 1/8 R600 (430mm2) + guesstimate on 65nm = <40mm2 (rv610) and <70mm2...but obviously nothing is linear. Still, I think <50mm2 and <80mm2 are very possible, and regardless of performance advantage or disadvantage to G84/G86, it's still possible the die sizes could be roughly half as large to G84/G86 if the 160mm2 on 80nm G84 rumor is to be believed.

Yeah, this is a shot in the darn, but I think it's possible, if not likely, RV660 may be a similar die size to G84, have a larger bus, and completely take that part to town. Yeah, noone wants to think about how "late" that would be compared to G84, but the truth is it seems to be the reality; like it or not. All we can do is hope it (and they) is/are competitive in the same market upon their release, and arn't delayed beyond the competition window.

 

[Topman]

http://forums.vr-zone.com/showthread.php?t=133860

oh no!

 

[Geo]

It's a little bit contradictory right now, isn't it? Why would AIBs have them and be commenting on performance if AMD is going back for another spin? And who was expecting to see them in March anyway?

 

[vertex_shader]

And who was expecting to see them in March anyway?

No rumors before scheduled the card coming out in march, its a vrzone AMD "news", i not need to say more :smile:

 

[Arty]

If that is RV630 indeed then its way too big than anyone was expecting. Its more than "half a R600" on 65nm big. Its almost ~60%, but then considering shrinks dont always work out exactly it could be half R600.

That could also explain the higher TDP.

 

[vertex_shader]

If that is RV630 indeed then its way too big than anyone was expecting. Its more than "half a R600" on 65nm big. Its almost ~60%, but then considering shrinks dont always work out exactly it could be half R600.

That could also explain the higher TDP.

The RV630 rumors are the following now: 65nm-bad performance-huge power consuption and now the core is big like alaska, i can't belive this 4 things in teh same time come true.

 

[Arty]

The RV630 rumors are the following now: 65nm-bad performance-huge power consuption and now the core is big like alaska, i can't belive this 4 things in teh same time come true.

Way to early to comment on performance, its supposed to launch in May. I'm not sure how far along the drivers are at this stage.

 

[vertex_shader]

Way to early to comment on performance, its supposed to launch in May. I'm not sure how far along the drivers are at this stage.

CJ suggest here that AMD rename the rv630 from x2600 to x2400, this say more than 1000 words about performance

 

[Blazkowicz]

Sounds cheap, but the 64bit without fast memory can be a bottleneck in performance.
I not really understand what AMD doing, maybe the cost of the cards overall lower with 64bit-high clocked gddr3 ram, than 128bit-lower clocked gddr3 ram, and the performance will be the same?

probably, just like geforce 7300GS has 400MHz ram which was quite fast in my mind. if the higher end RV610 has say 800Mhz GDDR3 it should be quite usable for some gaming (while not being the greatest thing on earth, sure)

 

[Arty]

CJ suggest here that AMD rename the rv630 from x2600 to x2400, this say more than 1000 words about performance

We'll see. Because thats a way big 65nm chip that'll end up being an X2400.

 

[vertex_shader]

We'll see. Because thats a way big 65nm chip that'll end up being an X2400.

Maybe its not a rv630 core in the picture, picture too small.

 

[Arun]

Assuming R600's die size is ~420mm2 on 80nm, and RV630's die size is ~160mm2 on 65nm... Then, if the architecture was very scalable (which you would expect it to be), it would be nearly exactly half a R600. It would certainly make sense if RV670, RV630 and RV610 were 75%, 50% and 25% of R600 respectively.

 

[trinibwoy]

Guess you guys haven't seen this tidbit yet - http://www.vr-zone.com/?i=4757

Also, we have gotten hold of an early performance test of the RV610LE running on 965 board with Core 2 Extreme X6800 processor. At 1024x768 resolution, 3DMark05 scores stood at 20xx and at 1280x1024 resolution, 3DMark06 scores stood at 12xx. The performance suffers largely due to its 64-bit memory interface and also drivers aren't optimized yet. We will provide you a glimpse of R600, G84 and G86 performance in our upcoming articles.

 

[vertex_shader]

Guess you guys haven't seen this tidbit yet - http://www.vr-zone.com/?i=4757

Funny, we don't know any performance number for AMD new generation cards except the lowest performance and cheapest one, vrzone news writers need a big brain wash, they are totally gone crazy..

Who want to know a 50$ card 3dmark numbers? its not me for sure
In this price segment the 3d performance not counting.

 

[Arty]

Guess you guys haven't seen this tidbit yet - http://www.vr-zone.com/?i=4757

Thats an LE, so the PRO & XT could be in the 8500GT range despite being 64bit.

 

[trinibwoy]

Who want to know a 50$ card 3dmark numbers? its not me for sure
.

Well considering there is a whole other thread for "high-end" stuff.....