R580 is just a X1800 XT speed bump up!

[AthlonXP1800]

Rolo from Rage3D forum had an interview (http://www.rage3d.com/board/showthread.php?t=33832485) with ATI:

- ATi has some excess inventory on X800 products because of NV's high end SLI products. (could we see discounts for consumers here in the future maybe?)
- The refresh for X1800 will be Q1.
- No release date for X1000-based AIW products yet.
- Their top driver engineer (formerly from SGI) is working to improve OpenGL performance.
- They will release AGP versions of the lower end X1000 series, and that "probably" includes the X1600. However 3rd party board makers CAN make AGP versions of X1800 using an ATi bridge chip, but only if its a viable business in their eyes.
- ATi uses AMD chips in their servers
- ATI will have a new chipset for AMD platform in Q1. This new chipset (RD580) should be faster than NForce4. It'll have 2 x 16 PCI-E lanes for SLI plus extra features.
- Which leads into this tidbit: AMD is introducing a new socket platform(!) in Q2.
- Intel asked ATi to make up to 6 million chipsets for them in 2006, which Phil said would make ATi the largest 3rd party mfgr of chipsets.
- AIW products based on X1800 will most likely have the 550 Theater chip.
- The 480M and 482M mobile chipsets are released.
- Mobile X1000 products will use about 25% more power than 9xxx mobile products.
- Mobile X1000 product in Q1 '06
- Currently ATi has about 70% of the mobile discreet video market, and the outlook is to lose about 10% market share in '06. (The discreet video market is about 30% of the entire mobile market, with integrated video accounting for the other 70%)
- The R580 (no offcial name) will be a speed bumped X1800XT. Didn't know if he meant just a clock increase though. Don't look for it to be a real big change from the X1800XT.

X1800 XT probably will have the shortest lifespan until it replace by R580 in Q1 2006. R580 is not the next big thing with 32 pipelines as ATI fans been hoped for months but it turned out to be a speed bump up of X1800 XT, to me it look like another X800 XT PE to X850 XT PE, I think Nvidia dont have a problem with this compete with 7800 GTX while the performances boosted in future drivers just like it did with 6800 Ultra. If ATI will still have products delays next year with R580 then the situation hasnt helped ATI, products delays are killing ATI fast. In Q3 2005 ATI had $324m in cash for spending and fund running operations and $400m inventory, in Q4 2005 ATI had just $223m in cash and $350m inventory. In November, ATI could have around $100m in cash and yet another inventory write down, they still struggling to clear existing X800 products in inventory. It is surprisely X800 series are not selling very well as ATI hoped against Nvidia's very successful SLi products, ATI still not have Crossfire products as well as vaporware X1300, X1600 and X1800 XT.

[Demirug]

I am sure ATI will try to sell the R580 as 48 pipe chip in the same way the do this with X1600 as 12 pipe chip.

[AthlonXP1800]

I am sure ATI will try to sell the R580 as 48 pipe chip in the same way the do this with X1600 as 12 pipe chip.

:lol: I dont think so, not until R600 if it will ever see the light of day.

[Demirug]

:lol: I dont think so, not until R600 if it will ever see the light of day.

R600 will be at last a 72 pipe chip if I count like ATIs new preferred method. ;)

But they are not very consistent. They should have tried to sell R520 as 32 pipe chip and R580 as 64 pipe.

What I try to say is that counting pipes is not very usefull anymore.

[nutball]

R600 will be at last a 72 pipe chip if I count like ATIs new preferred method. ;)

To be fair the "R520 is 32 pipes" nonsense didn't come from ATI, it came from hard-of-thinking muppets who read nonsense on fanboy forums and believe it!

The slightly more sane and knowledgeable people from other sites consistently maintained that R520 was 16 pipes. Guess what? They were right! Funnily enough, those same people have also been suggesting that R580 *will* feature some significant architectural improvements over R520. #-#-#-# and all that. Fanboys won't listen of course because it doesn't suit their agenda.

But as you say, pipe count is rapidly losing its meaning. Some people might even have to start to think, rather than downloading programs to multiply two (incorrect) numbers together and mouthing off about fill-rates, etc. because their number is bigger!

[retsam]

To be fair the "R520 is 32 pipes" nonsense didn't come from ATI, it came from hard-of-thinking muppets who read nonsense on fanboy forums and believe it!

The slightly more sane and knowledgeable people from other sites consistently maintained that R520 was 16 pipes. Guess what? They were right! Funnily enough, those same people have also been suggesting that R580 *will* feature some significant architectural improvements over R520. #-#-#-# and all that. Fanboys won't listen of course because it doesn't suit their agenda.

But as you say, pipe count is rapidly losing its meaning. Some people might even have to start to think, rather than downloading programs to multiply two (incorrect) numbers together and mouthing off about fill-rates, etc. because their number is bigger!
__________________QFT .... at the end of the day ... its all about the benchmarks...

[Demirug]

nutball, the 72 pipes for R600 are not based on the "R520 is 32 pipes" rumor. It is based on the way ATI counts pipes for the X1600 cards.

The R580 will have the same architectural improvements that the X1600 cards already have.

[msxyz]

It would be better to think in terms of shader cores as the old analogy 1 pipeline = 1 pixel processing unit is not valid anymore.

[nutball]

nutball, the 72 pipes for R600 are not based on the "R520 is 32 pipes" rumor. It is based on the way ATI counts pipes for the X1600 cards.

I know. Although I quoted your post I was kind of aiming my comments at the OP. Apologies for the bad netiquette!

[Muppet]

[QUOTE=nutball]To be fair the "R520 is 32 pipes" nonsense didn't come from ATI, it came from hard-of-thinking muppets who read nonsense on fanboy forums and believe it!

QUOTE]
Hey man don't hastle the Muppets, or else :)

[Graphicmaniac]

it would be a great marketing spot pruduce a

R580 with
800mhz gpu (16 pipe) and
1800mhz ram

anyway i wait the true graphic revolution that comes with the gpus after W Vista release :P

[AlphaWolf_HK]

- ATi has some excess inventory on X800 products because of NV's high end SLI products. (could we see discounts for consumers here in the future maybe?)

I should hope so for ATI's sake. Not even the x850 XT can match the 7800GT, yet the 7800GT can be found for a cheaper price.

[Redeemed]

So, who do you think will have the first ENTIRELY DX10 compliant vcard- ATI or nVidia? And who's card will perform best under Vista? I'm THINKING nVidia, but I'm also NOT counting ATI out.

[Dazz]

I don't know i think the R580 will be 24 & 32 pipelines however no really feature changes, kinda like the X800 is not that much diffrent from the 9700

[rohit]

No 32 piped cards this round.
Inquirer guyz should be Burned/Burried alive.
A$$HOLES got me into Hi-Hopes of a 32piped AGP card. :rant:

:p

[AthlonXP1800]

So, who do you think will have the first ENTIRELY DX10 compliant vcard- ATI or nVidia? And who's card will perform best under Vista? I'm THINKING nVidia, but I'm also NOT counting ATI out.

According to Microsoft on DirectX 10, they said both Nvidia and ATI will peform the same. Nvidia will launch DirectX 10 card first and ATI launch it 6 months late. But in reality, ATI will be on the brink of liquidation or bankruptcy in 4 months if they continue to lose money, they dont have enough money to fund $300m R&D for next generation R6xx in 2006 and still have difficulties to dump obsoleted R9x00, X300, X700 and X800 products from inventory. The long delayed Crossfire motherboards and X1800 XL are not selling very well and Crossfire X800, Crossfire X850, Crossfire X1800, X1300, X1600, X1800 XT and X1800 AIW cards are delayed too nowwhere to be seen.

[retsam]

well i think your forgetting to factor in the money they will make for the xb360 licenses...

[Sazar]

Or the fact that they are still have excellent market share in many markets and are not completely reliant on mid-high end game cards.

Currently most mid-range desktops still have an x300 as an option rather than an nvidia component.

I don't think athlonxp1800 is seeing the market in the way it really operates.

Ati can't afford to screw up repeatedly and needs to ensure when it has the parts available that they are selling but its hardly do or die right now.

[AthlonXP1800]

well i think your forgetting to factor in the money they will make for the xb360 licenses...

No, ATI wont make lots of money from Xbox 360 as much as Nvidia did with Xbox, Nvidia made $100m every quarterly and had fall out with Microsoft, complainted that the fee was 10 times higher. This time with XBox 360, ATI will make around 10 times less with $10m from it.

[Treason]

But in reality, ATI will be on the brink of liquidation or bankruptcy in 4 months if they continue to lose money, they dont have enough money to fund $300m R&D for next generation R6xx in 2006 and still have difficulties to dump obsoleted R9x00, X300, X700 and X800 products from inventory.

Four months eh? Funny I don't see any massive downsizing at ATi at the present... Are they diehard ATi employees or something? :rolleyes:

[Shyru]

So how can they do a speed bump with the X1800?
The R520 has roughly 300M transistors. And they need to have it run at 600+ MHz to compete with the 7800 GTX. The G70 has also roughly 300M transistors, but runs with 470 MHz to achieve comparable speed.
So in my opinion the G70 is cleary the efficience winner: One transistor does more per Mhz than a transistor of the R520.
So what does ATI want to do? Increase the Mhz even more? I doubt this to be possible. Increase the pipes? How should they do that? If they increase to 24 Pipes the r580 would perhaps have 400M transistors. I think they will not be able to run this chip monster at 600+ Mhz! So they need to decrease frequency and will end up with a chip with 24 Pipes which runs at 500 MHz and is as fast as the R520. So, in summary i see no possibility to speed bump the r520, because it so inefficient (transistor-wise).
What are your ideas? Whats that i've overlooked?

[AthlonXP1800]

Four months eh? Funny I don't see any massive downsizing at ATi at the present... Are they diehard ATi employees or something? :rolleyes:

Things can happen suddenly and completed unexpected, it did happened to Commodore and 3Dfx, they both thought they finally had left the mess behind and look postively to the future but they never looked at their quarterly cash flow statements, ATI probably did exactly the same.

[MUYA]

So how can they do a speed bump with the X1800?
The R520 has roughly 300M transistors. And they need to have it run at 600+ MHz to compete with the 7800 GTX. The G70 has also roughly 300M transistors, but runs with 470 MHz to achieve comparable speed.
So in my opinion the G70 is cleary the efficience winner: One transistor does more per Mhz than a transistor of the R520.

You are forgetting the quad blocks or "pixel pipelines" each GPU has. To compensate for the lack of 2 quad blocks, the R520 is clocked higher to achive the sameish fillrate as the G70.

[Shyru]

You are forgetting the quad blocks or "pixel pipelines" each GPU has. To compensate for the lack of 2 quad blocks, the R520 is clocked higher to achive the sameish fillrate as the G70.
Thats clear, i never said something different. I'm not talking about efficiency per Pipeline (ATI is clear winner there), I'm talking about efficiency per Transistor.

The R520 has only 4 quad blocks, but yet has roughly the same amount of transistors as the G70. So with the same amount of transistors, nvidia achieved to build 6 quad pipes. And with this 6 quad pipes and 470 MHz they achieve the same speed as the r520 with 4 quad pipes and 600+ MHz. Note: They have nearly the same amount of transistors! So looking at the transistor count nvidia's G70 is more efficient, because they have the same speed with the same transistor count but less MHz.
So still: what is ATI gonna do?
Increase MHz even more? I doubt it!
Increase Pipelines? I don't know how! (See my original post)

[AthlonXP1800]

Or the fact that they are still have excellent market share in many markets and are not completely reliant on mid-high end game cards.

Currently most mid-range desktops still have an x300 as an option rather than an nvidia component.

I don't think athlonxp1800 is seeing the market in the way it really operates.

Ati can't afford to screw up repeatedly and needs to ensure when it has the parts available that they are selling but its hardly do or die right now.

Well you need to looked at Commodore's past with Amiga and 3dfx's past with Voodoo.

Amiga was very successful product and had excellent market share 11 years ago. Commodore made the same mistakes just like ATI at present in 1993 after posted $356.5m losses in Q2 as result of discounts on excess inventory of obsoleted A500, A500+ and A600s that nobody wants to buy that got old ECS graphics chipset, slower CPU, less memory and running older Amiga OS 2.0, everybody want to buy new and latest Amigas with advanced AGA chipset, faster CPU, more memory and running modern Amiga OS 3.0 as well as Amiga CD32 consoles. But later in the year, Commodore faced production difficulties and products delayed as they had less than $107m in cash, not enough to made 3 million Amigas to met very high demands.

A year on in 1994 Amiga continued to smashed all sales records and most customers didnt buy it but rather waited for next generation 64 bit RISC based Amigas that were 5 times faster than the fastest Pentium and have world first full programmable 2D and 3D graphics on the core of HP-PA RISC chipset long before Nvidia did. That situtations did not helped Commodore in March 1994 when they posted the smallest $8.2m loss and panicked to discovered that they have no money left. Commodore started prepared to scaled down operations and employees layoff, after a month of ongoing neogitations with the creditiors and shareholders about their rescue plan to save the company but how can they get rid of excess stock of A500, A500+ and A600 that nobody want? The creditors and shareholders see no other alternatives for Commodore to get rid of it in return for a large loan and they then refused to give Commodore loan to continue the business and see the launch of next generation RISC based Amigas in 6 months. In the end, Commodore International Limited had filed for liquidation and shutdown the operations and closed door for business on 29 April 1994.

The fall of Commodore was took people completed by shocked and unexpected, Commodore US management seriously screwed up repeated in 2 years after made many mistakes and had 4 bad quarterly reports in a year. 3Dfx management did the same in 2000 and then in the end of year they had only $9.39m in cash, not enough to continue the company operations and no money to fund R&D for 3dfx Rampage and the Ramage launch.

ATI like Commodore and 3dfx never learn from so many mistakes.