Radeon 2900 Pro?

RV630 is ring based. They are 32-bit channels as well.

[Edit] - Fellix, it sounds like you are thinking that RV610/630 are the same - this isn't the case. RV610 is a crossbar based memory controller, although it does have two 32-bit channels

 

So does RV630 have 2x 32-bit channels in each of two ring stops, or 1x 32-bit channel in each of four ring stops?

LOL, I thought R5xx's 32-bit memory channels had gone for good.

Jawed

 

Well, I was about to like the idea of 16*32-bit config on R600, anyway.
Then tell me about the overkill thing.

So, Dave, looks like the same situation as it was for the previous R500 line, considering the ring-bus scaling.

 

2x32 for GDDR4 only, or GDDR3 as well? Going with a 64-bit memory interface would seem to be a cost-saving measure. Would it be worth it to include the additional tranceivers (~16) to allow 2x32 with GDDR3 DRAMs?

It might not dramatically impact the die size if one uses area-IOs (to avoid being pad-limited), but then you likely need a more expensive package to escape all the signals.

 

Some more fruitzilla rumor

 

I knew it!
Now, where is my candy.

 

With this specs the card can't compete against the 8800gts320mb than some user hope/suggest.

 

Does ATi (better said TMSC) produce so many R600s with two damaged/defective/broken SIMDs?

 

In my eyes r600 is a dead-end, impossible to make anything good in lower segments with this GPU because of the GPU performance and power consuption, but looks like rv670 won't coming this year so AMD not have other chance the release something in Q4.

As i said already this card won't be a hotseller, same happend with x800se/gt/gto(gto was a little better because on some AIB cards the disabled pipeline quad can be enabled with bios change) , x1800gto/x1900gt/gt rev2. (x800 gto2/x1800 gto2 i not count it was sapphire own brands with factory enabled quad).

AMD need to face almost the mission impossible with r600, they only hope if something can be enabled in some of the hd2900pro cards this make some hype and maybe help the cards sell better.

 

And yet if ATI's own site is to be believed the R600 Pro is 320 SPUs. Granted it's for the FireGL, but would they change the chip even further just to differentiate it from a consumer grade card when they don't do so either RV630 or R600XT(X)?

I'd be more inclined to think they'll follow suit as in the past with the FireGL being basically equivalent to the Consumer card other than bios and drivers.

An R600 with only a 256-bit memory controller whilst all the rest stayed the same would be an intriguing card in the mid-range.

Regards,
SB

 

With dual slot cooler and 180watt power consuption? this is not midrange, i'm sure not this card what midrange users with they 50-70$ PSU's dreaming on.

If this will be the hd2900pro what you write its won't cost 200$ for sure.

 

In my eyes we shouldn't read Fudo.
Ok, R600 isn't that, what we expected, but 2900Pro with same specs like FireGL 7600 would be great. Why? Does anyone really care about power consumption? No. Does anyone really care about temps? No. But may be about loudness.
Things like drivers or dirvers' stability are more important, and ATI's driver team does a really good job. R600's drivers are not perfect, but they get better and better.

But there is one thing, that makes me scared: http://www.3dnews.ru/video/512_bit_memory_bus_efficiency/ (Google translated: http://www.forum-3dcenter.org/vbulletin/showthread.php?p=5687891#post5687891)
What's wrong there?

 

The only strange thing I can see here is the odd drop in performance between 550~700MHz range, but that's way below stock speeds. Looks like a timing issue in synchronizing I/O sequencing.

 

Don't forget this card competion can be a 65nm GPU from NV.

 

All the other FireGL line cards have kept their specs to their respective consumer line products. Why would the 2900pro be an exception?

 

Where is the xtx with 700mhz core clock speed?

 

Probably the same place as the nVidia non-Quadro model that has 768MB of VRAM with a 384bit memory interface and only only 67.2 GB/s of bandwidth of the QFX4600.

You seem pretty hung up on clock speeds.

Workstation cards are generally clocked slower than their gaming counterparts, because they're built for quality and reliability, not just absolute top speed.

Don't forget, it already is in the form of the GF8600 series, whatever comes next is for when that's announced, and 65nm alone isn't the issue, it's architecture as the HD2400/2600 prove.

While power consumption matters to some, like myself, when building efficient rigs for things other than gaming. As the R9500Pro , X1900GT and GF8800GTS-320 have shown, if it offers great performance for the price, most people are willing to use they extra juice.

 

The FireGL line can differ in memory configuration, and clock speeds. I have never seen them differ in piplines or bus width from their consumer couter parts.

 

Except that it's been reported that the power cosumption of the R600 based FireGL cards with 1 gig of memory will be <225 watts (same as the R600 Consumer board) and that the R600 Pro based FireGL card with 512 megs of memory will be <150 watts.

A far cry from 180 watts there. Then again I wonder if people are still running around claiming the R600 boards consume over 240 watts of power?

And considering my R600 runs perfectly fine with a 400 watt power supply (albeit with only 1 HD), I'd say that yes, even the power profile of a R600 Pro could fit in the midrange. All depending on what they price it.

Question is, are they going to price it against the 8800 GTS 320 (probably lose in low resolution but win in high resolution) or the top 8600 cards?

Regards,
SB

 

You loose your pills?
Even the 100-110-120-130-140-150watt is crazy for midrange, check some ATi AIB forum x1950pro topics you will see how many midrange users have PSU problems with the x1950pro and this card consume less than 80watt in load.

BTW. don't belive anything what you see in AMD slides about power consuption, before rv630 released its hyped to the sky because of the much lower power consuption than g84, here is the reality, 65nm based hd2600xt vs 80nm 8600gt, and don't forget 8600gt avarage faster than the hd2600xt and already in the same price segment Link Link
ATi power consuption problems started with r520,now they have with rv630 the lower process lead and can't profit anything power consuption wise with this.
Lower process and high power consuption means that AMD cards core clock always already bruteforce when its released and still lack of performance against competion cards.

Of course NV 65nm GPU can leak too like the r600 on 80nm, maybe AMD hope this, but we see what NV can do on 90nm with g80, still hard to belive how NV can make 671million transistor GPU on 90nm with r580 power consuption what have 384million transistor.

When its r600 with 256bit and lower core speed it's won't cost 200$ for sure when its released.
Here is a easy question we not count anything else just performance, hd2900pro lose in low resolution against 8800gts320mb 52 vs 46 frame in the latest blockbuster game, and win in high resolution 14 vs 12 frame, what card you want to buy ?
8800gts320mb not a good product either, have slowdown problems what i'm sure NV never fix, so here is a bone again for ATi/AMD