Did ATI hold back on R520 to make room for R580?

geo said:
As well, did you notice Komb's smiley re "R580 won't be released any time soon"? There begin to be Signs and Portents showing up in public that Anand's "CeBit" prediction may have been too conservative.

Those public portents haven't reached critical mass yet, but they do seem to be growing.

Yep I sure noticed - seems those boys in red have got an ace or two up their sleeves.
 
trinibwoy said:
Yep I sure noticed - seems those boys in red have got an ace or two up their sleeves.

The owie portion, based on the hints on NV 512mb pricing conversation, might be another move up of $100 on top-end sku MSRP for both IHV to support that accelerated timing and king-of-the-hill duel.

Edit: And, btw, while certainly the performance crown competition is part of that, I can't help but think that SLI is partly to blame for the blowup in high-end sku pricing we've seen the last year. SLI proved there is a viable, if small, market that is willing to fork over quite a bit of money above the prices at release of X800/G6800.
 
Last edited by a moderator:
geo said:
The owie portion, based on the hints on NV 512mb pricing conversation, might be another move up of $100 on top-end sku MSRP for both IHV to support that accelerated timing and king-of-the-hill duel.

I wouldn't be suprised if in a couple product cycles the FX's and EE's of the GPU world occupy the $800-$1000 price segment that their CPU brethrin enjoy today.
 
trinibwoy said:
Yep I sure noticed - seems those boys in red have got an ace or two up their sleeves.

Wavey's comments on GDDR3/4 with R580 could be read that way too, for example.
 
geo said:
Wavey's comments on GDDR3/4 with R580 could be read that way too, for example.

So the expectation is that G71 isn't going to be able to handle GDDR4 once it becomes available in volume?
 
R580 won't be around for awhile. Most likely this 512 MB "GTX" will go for around $600 or more and ATI will simply drop the price of the X1800 XT 512 to around $500. This is great as it means lower prices all around, $450 for 256 X1800 XT, $400 for X1800 XL, etc.
 
trinibwoy said:
So the expectation is that G71 isn't going to be able to handle GDDR4 once it becomes available in volume?

Huh? I haven't seen anything like that. I just meant that as ATI has started hinting about GDDR4 for R580 that Wavey has been saying he doesn't expect to see it on the first R580, but on a refresh. Both ideas --that it won't be on the first one, but that there will time for a refresh of R580 before Vista hits in earnest, suggest he's leaning "sooner rather than later" on R580. Or at least that's what these chicken bones I just cast say. :p So I was just offering that example as another "portent" on R580 timing.

Don't know nuttin about 90nm G7x high-end other than "not this year" --unless they snuck one in on us with those eye-popping AIB clocks on the 512mb "GTX". :LOL: But I'd think we would have heard by now with a chip shot to prove it.
 
geo said:
Huh? I haven't seen anything like that. I just meant that as ATI has started hinting about GDDR4 for R580 that Wavey has been saying he doesn't expect to see it on the first R580, but on a refresh. Both ideas --that it won't be on the first one, but that there will time for a refresh of R580 before Vista hits in earnest, suggest he's leaning "sooner rather than later" on R580. Or at least that's what these chicken bones I just cast say. :p So I was just offering that example as another "portent" on R580 timing.

Ah ok, I didnt have Dave's comments for context so couldnt really follow :)
 
trinibwoy said:
sireric has voted "Ati will release its r580 monster and sweep the floor with nvidia" on a recent poll at Rage. Funny stuff :LOL:

http://www.rage3d.com/board/showthread.php?t=33833964
Hey!

Actually, if NVidia has access to 1.1ns GDDR3, why shouldn't ATI? (Although it's fun to remember that NVidia was the first IHV to get GDDR3 actually on a card, wasn't it?)

Damn the run-up to Christmas may turn out to be more fun than I was expecting...

Jawed
 
trinibwoy said:
Yep I sure noticed - seems those boys in red have got an ace or two up their sleeves.
If R580 is on-schedule, then wouldn't its schedule be for a December release?

R480 was supposed to be a December release - I know, it was a month late. But shouldn't R580 be launched before Christmas, for the same reasons that R480 was?

http://www.beyond3d.com/forum/showthread.php?t=15682

Though that press release has more references to X800 than it should :p


Jawed
 
Last edited by a moderator:
Jawed said:
If R580 is on-schedule, then wouldn't it's schedule be for a December release?

R480 was supposed to be a December release - I know, it was a month late. But shouldn't R580 be launched before Christmas, for the same reasons that R480 was?

http://www.beyond3d.com/forum/showthread.php?t=15682

Though that press release has more references to X800 than it should :p

Jawed

That was before (and arguably part of the reason that drove in this direction) both IHV's started tightening up the announce/availability timeline tho, right? If you read Orton's answer at the last conference call about timing, he on the one hand defends announcing X1800XT with a month gap to availability because of X1800XL and X1300, to give a full view of the new family. . .and on the other suggests it doesn't mean they don't want to continue tightening that gap more towards the model NV has been going with of late when that particular consideration isn't in play (i.e. a whole family at once).
 
geo said:
That was before (and arguably part of the reason that drove in this direction) both IHV's started tightening up the announce/availability timeline tho, right?
Let me be honest and say I'm far too new at this game to know what happened in previous years (i.e. 2003, 2002) when Christmas refresh-SKUs were announced.

What's the point of announcing a winter refresh before Christmas if it's not ready to buy before Christmas?

But I dunno - too new at this game, as I say.

Also the X850XT refresh was supposed to be about availability, and those cards were supposed to be available before Christmas - but slipped well into the new year. So ATI really mucked-up there.

One would hope that if R580 has been planned as a winter refresh for the last 18 months or so that, notwithstanding the R520-balls-up, the damn thing would appear on time.

Maybe I'm just putting too much faith in the idea of a winter refresh SKU :rolleyes:

Certainly ATI should have been expecting NVidia to release a winter refresh. That's what R580 would have been targetted-at all along, surely?

If you read Orton's answer at the last conference call about timing, he on the one hand defends announcing X1800XT with a month gap to availability because of X1800XL and X1300, to give a full view of the new family. . .and on the other suggests it doesn't mean they don't want to continue tightening that gap more towards the model NV has been going with of late when that particular consideration isn't in play (i.e. a whole family at once).
There's no reason why ATI can't announce later to close the "paper" gap between announcement and shelf.

Anyway, this is gonna be fun. If 7800U releases 14/11 and X1900XT releases 28/11 :drool: <--B3D needs a drool smiley.

Jawed
 
wireframe said:
Why would ATI create a successor like the R580 if they didn't think the important thing was to advance shader performance?

Just to be clear, I myself don't believe the premise of this thread anymore. The point I'm trying to make is that shader performance (or lack thereof) is only one possible performance bottleneck. It will depend on the application which one gets hit first. And with the games that will be available at R580's launch, will it often be the shaders limiting performance? Specifically at high resolutions,AA and AF settings?
 
Let me get this straight. There are people who think that a publicly traded company with recent financial difficulties and legal issues is holding back technology so that it would be sold in 4-6 months from now?

My answer to that is, no @#$%ing way.
 
Last edited by a moderator:
mrcorbo said:
The point I'm trying to make is that shader performance (or lack thereof) is only one possible performance bottleneck. It will depend on the application which one gets hit first.
Obviously, but the trend is to make more use of shaders. This can be thought of as a more general transition from data storage bandwidth and capacity to processing capacity, offsetting that "wall." (this is something like file compression on CPUs: You reduce size but increase processing need to work with the file. In this case, data can be computationally generated instead of stored (for example: procedural textures))

And with the games that will be available at R580's launch, will it often be the shaders limiting performance? Specifically at high resolutions,AA and AF settings?
This is a very valid point. It is unlikely that the titles available at launch will demonstrate the true difference in processing power between the two. The R580 should still enjoy the necessary ~30% advantage that a new part needs and this gap will grow over time. Again, this will obviously vary between titles, but I think we can agree that it is the latest titles, and therefore probably the most shader intensive titles, that need it the most and what will drive people to upgrade. (Note that I do not believe the R580 will ever be 3x as powerful shading-wise as the R520 in real usage, but I do believe the R580 can be twice as powerful.)
 
ondaedg said:
Let me get this straight. There are people who think that a publicly traded company with recent financial difficulties and legal issues is holding back technology so that it would be sold in 4-6 months from now?

My answer to that is, no @#$%ing way.

It's not that simple though, is it? Can a company in such difficulties afford to completely write off a product? Nvidia had some problems with NV3x, but they stuck to their guns, riding it out instead of completely panicking. However, I think the main reason the R520 exists now is for the X1800XL and not the XT. But they need something to show.

On the bigger scale, I don't see what the problem is. I think it's looking good for ATI, not excellent, but good. The XT is performance king (or looks like it) and the new hardware may have some room for driver/software improvements. Under no circumstances do I think you could say the XT is inferior to the GTX. Therefore, there is no reason for ATI to skip on using the R520 for the XT, earning some revenue on what has become a very expensive project while they wait to release the R580.

PS. I think the announcement of R580 being "ready and working" is not meant to be taken as a "soon to be launched," but as a "can be launched soon". In the 3D industry you ive and die by the products in your pipeline. Nvidia effectively killed off everyone in the early market through execution, execution, execution. Companies like 3dfx were powerless against it. It didn't matter how current products stacked up, Nvidia always threatened to release a newer, more powerful model if it didn't have the performance crown. Analysts want to see that companies are prepared for this kind of attack.
 
mrcorbo said:
Just to be clear, I myself don't believe the premise of this thread anymore. The point I'm trying to make is that shader performance (or lack thereof) is only one possible performance bottleneck. It will depend on the application which one gets hit first. And with the games that will be available at R580's launch, will it often be the shaders limiting performance?
This is very much up for debate. I've been pondering whether it's worth starting a thread on this: "R580 performance - R580:R520 as RV530:RV515?"

I prefer to err on the side of caution because RV515 is hindered by its old-skool memory controller and RV530 has double-rate Z.

But I've had thoughts that the 3:1 shader:texture ratio may increase texturing speed as well as increasing shader speed - so...

Specifically at high resolutions,AA and AF settings?
High-AA on R580 will benefit from 1.1ns memory. Indeed R580 will gain more from 1.1ns memory than G70, simply because ATI's new memory controller is using memory bandwidth more efficiently.

As I said earlier, if NVidia can utilise 1.1ns RAM, why can't ATI?

Jawed
 
wireframe said:
It's not that simple though, is it? Can a company in such difficulties afford to completely write off a product?
There's no reason why ATI should write-off R520, as I said earlier in the thread:

http://www.beyond3d.com/forum/showpost.php?p=613556&postcount=9

I also think that the price chasm twixt X1600XT and X1800XL - $200 - indicates that X1800XL/XT are going to drop very heavily in price.

And to me the best reason why is R580. With X1800XL at $250 and X1800XT at $300/350 (256/512MB) there's room for $450/$500/$550 X1900XL/XT-256/XT-512 parts. It's a pisser that it'll be after Christmas before R580 hits the streets.

But now we're (well some of us :p ) shifting towards expecting R580 before Christmas :cool:

Blimey that would be so cool if it was true. With R590 releasing in May with 2GHz+ GDDR4 :drool: <--B3D smiley needed

Jawed
 
Jawed said:
But now we're (well some of us :p ) shifting towards expecting R580 before Christmas :cool:

Blimey that would be so cool if it was true. With R590 releasing in May with 2GHz+ GDDR4 :drool: <--B3D smiley needed
That just seems so very highly unlikely. Without knowing the numbers, just a crystal ball here, I think that any gains made on early R580 would be offset by too rapidly dropped margins on R520.

I'm going to stick to my guns and say ATI will stick to theirs. They will ride out this storm because the situation is not so bad. In no way do they immediately need an R580 part. Let's see what the 512MB version of Nvidia's 7 Series brings to this mix.

PS. I am thinking that R580 is the focus of ATI at this point, but it is one of those tempting levers I think they will not pull in panic and that is probably a very good thing.
 
Well, ATI management has been talking to investment analysts again. Couple of excerpts from the TD report:

Management indicates it has achieved extremely strong yields on the 90nm process. ATI intends to follow-up with another high-end desktop discrete chip (code-named R580) in the Feb/06 timeframe.
And
ATI already has in-house chips designed on the 80nm node, and expects to be first to market with 65nm.
(And just to defuse speculation before it starts, the R580 reference did not appear anywhere near the 80nm reference.)

Out of respect for the actual "analysis" part of the analyst report, I haven't included all the financial details, which I imagine would bore most people here anyways.
 
Back
Top