Anand has the details about r520,rv530,rv515

[Skrying]

This would imply that ATI made a load of early revision chips that are leaky and have design problems fixed in later revisions - enough to supply a whole couple of SKU's. I find it hard to believe given the delays and respins we've seen that ATI went ahead and made millions of known problematic early revision chips months back in the hope they could make them work as a low-performing XL product, while still chasing down the problem.

Not really that unbelievable. The issue with the chips is not that they are not functional, but rather they can thit the super high speeds of 600Mhz+. So instead of throwing them away or be unuseful, they can have a bit made and stock them for the launch while waiting for a respin to come out perfect.

Remember, the high end X1800 really do not sell that much at all, its more of a mind share product, its the X1600 and X1300 that are the big money makers, so it'd probably be in ATi's interest to save money by producing some of the better-bad spins into functioning chips for the slower end X1800s.

 

[Bouncing Zabaglione Bros.]

Well I don't know how it works but I thought the IHV's only started ramping production after all the kinks had been worked out. So I'm not sure how ATi could have enough "bad" chips to fill the early XL/LE demand.

Yes, that was my point exactly.

 

[Sinistar]

Not really that unbelievable. The issue with the chips is not that they are not functional, but rather they can thit the super high speeds of 600Mhz+. So instead of throwing them away or be unuseful, they can have a bit made and stock them for the launch while waiting for a respin to come out perfect.

Remember, the high end X1800 really do not sell that much at all, its more of a mind share product, its the X1600 and X1300 that are the big money makers, so it'd probably be in ATi's interest to save money by producing some of the better-bad spins into functioning chips for the slower end X1800s.

Sounds very logical to me.

 

[Dave Baumann]

Keep an eye open tomorrow, a tiny bit of the curtain may be lifted...

 

[Bouncing Zabaglione Bros.]

Not really that unbelievable. The issue with the chips is not that they are not functional, but rather they can thit the super high speeds of 600Mhz+. So instead of throwing them away or be unuseful, they can have a bit made and stock them for the launch while waiting for a respin to come out perfect.

Hmm, that does sound possible given that Orton has said that R520 was functional from day one, but not reaching required speeds. Sounds risky though. As a potential upgrader, I don't want one - I want the latest revision with all fixes and upgrades.

 

[Skrying]

Hmm, that does sound possible given that Orton has said that R520 was functional from day one, but not reaching required speeds. Sounds risky though. As a potential upgrader, I don't want one - I want the latest revision with all fixes and upgrades.

Yeah, I personally could see all of those older cores going out to the OEMs who will have no need but for the default valued cores. This could be how ATi is planning to meet full demand, not to bad an idea.

Keep an eye open tomorrow, a tiny bit of the curtain may be lifted...

!!!!!! I cant wait for that!

 

[wireframe]

I don't think the speeds and overclockability will be the key factors in the success of the X1800 line. If the rumored SKUs are correct then we have three models that all share the same hardware specification: the XL, the Pro, and the XT. This should give ATI ample room for binning. It also means that the consumer can be impressed by the top-of-the-line XT and still buy into that spec with the lower priced Pro or XL without feeling cheated or put in a difficult decision making situation.

In the case of the X800, I think this harmed sales quite a bit. You see this glorious X800 XT (PE) that is difficult to find and costs a fortune when you do. The model that was available and affordable was the Pro, but this had a quad disabled. Not quite the same thing and may make the potential customer second guess their decision to buy. This same 'problem' exists with the 7800 GTX and GT. However, it did not exist between the 6800 Ultra and GT and I think this was a big part of its success. The X800XL, coming late to the game was still a success and probably for the reasons I have outlined here. Imagine what that part would have brought to the table had it been introduced from the beginning.

I am not sure that overclockability is a very important factor. One may think so when looking at the success of the 6800 GT, but the X800XL is generally a poor overclocker and does very well. By ATI speed binning (we can think of this as factory overclocking) three hardware identical SKUs with the X1800, I think they have served up a smorgasboard (possible new PCB codename?) to the consumer and this will help sales tremendously. Even if people think "I'll buy the XL and overclock it to XT or Pro levels, saving me money" and failing in doing so, they will be offering parts at three price points that are not crippled in relation to the Big Brother (unless ATI decides to software limit some features, but we know how that turns out...).

PS. Could someone set me straight on the model progression, is it, slowest to fastest: XL, Pro, XT? For some reason I keep wanting to think it goes Pro -> XL -> XT. That's probably the X800 legacy "XL 16 pipeline effect" working on me though.

 

[wireframe]

Keep an eye open tomorrow, a tiny bit of the skirt may be lifted...

(edited to match the drool hanging off my face)

Alright! No more sleep for me! Ever! Bring it!

 

[Skrying]

Its XT > XL > Pro > LE.

 

[wireframe]

Its XT > XL > Pro > LE.

Thanks. The Digit-Life preview says as much, but I am sure I read something to the effect that "Pro will be the second from top model, followed by XL" somewhere and it just struck me as weird and then stuck in my head. So, everything is "just as it should be then"

 

[R300King!]

Not really that unbelievable. The issue with the chips is not that they are not functional, but rather they can thit the super high speeds of 600Mhz+. So instead of throwing them away or be unuseful, they can have a bit made and stock them for the launch while waiting for a respin to come out perfect.

Remember, the high end X1800 really do not sell that much at all, its more of a mind share product, its the X1600 and X1300 that are the big money makers, so it'd probably be in ATi's interest to save money by producing some of the better-bad spins into functioning chips for the slower end X1800s.

Of course, and when Sander asked for some R520 benchmarks the sly board partner gave him exactly what he asked

for...of course it was an early pre-production R520 with leaks and/or possibly other problems/lower clocks,

etc. Oh, this is classic. You're probably right. They may be using the problematic early R520s and using

them for X1600s and X1300s, etc. I just hope the lastest R520s are the beauts and are problem free and then

after speed binning these, this is where you'll get your X1800XT, X1800XT PE, etc. I bet we'll see SOME X1800

XTPEs well over 700Mhz. If this is true, IF, then the benchmarks Sander posted were most likely what ATi would

be coming out as an X1600 or X1300.



Wow, if those numbers were basically from an X1600 and they're right up

there with nVidia's lastest, then ...well, wow. Can't wait to see what an X1800XT can do.

 

[Arty]

Keep an eye open tomorrow, a tiny bit of the curtain may be lifted...

*goosebumps*

And I'm not kidding.

 

[Ailuros]

I bet we'll see SOME X1800 XTPEs well over 700Mhz...

 

[Skrying]

Yeah, I'm really hoping ATi doesnt do an XTPE. I think its just a bad idea have a super super high end card that is stupidly rare to have. If you promise a card, please have enough that you can meet orders.

 

[trinibwoy]

If this is true, IF, then the benchmarks Sander posted were most likely what ATi would be coming out as an X1600 or X1300. Wow, if those numbers were basically from an X1600 and they're right up there with nVidia's lastest, then ...well, wow. Can't wait to see what an X1800XT can do.

Man if you really think there is even the remotest possibility of ATi's ~ $250 12-pipe part running with Nvidia's $600 24-pipe flagship then you've have reached a whole new level of IHV worship my friend

 

[Ailuros]

How sure are you guys that that won't be the case with the XT? As soon as the 6/8 quad nonsense found an end, we now are going to enter the ultra-high clockspeed era?

It belongs into the same trashcan IMO as the rumoured 7800 "Ultra" stuff. While I'd expect in the foreseeable future higher clocked Gxx-whatever GPUs at 90nm, I have severe doubts that we'll see another 110nm version of the G70. Vendors of course will continue to sell higher clocked 7800's by their own initiative.

To put that mess above into perspective IMHO it will be X1800XT vs. 7800GTX for this round, until the next batch of "refreshes" from either/or arrives and yes that's just me.

 

[Skrying]

Man if you really think there is even the remotest possibility of ATi's ~ $250 12-pipe part running with Nvidia's $600 24-pipe flagship then you've have reached a whole new level of IHV worship my friend

But what if it did!!

Nah, I expect X1600 to hang around 6800GT performance personally, depending on how much those pipelines have been changed. Probably slightly slower, still pretty damn powerful. I expect the 7600 to be around 5%, either faster or slower.

What I'm curious about now is that X1300 and the 7200. Hmm, if either of those parts reached close to 6600GT performance then they'd be mouth watering low end parts!

 

[Skrying]

Ail, I believe if Nvidia countered with a 90nm G70 part with ultra high clock speeds then I'd expect ATi to wait a bit and drop the R580 on it. I really dont know how ATi is going to play the R520 and R580 together, it seems so close to the R600 and G80, even though they are also so far away.

 

[trinibwoy]

What I'm curious about now is that X1300 and the 7200. Hmm, if either of those parts reached close to 6600GT performance then they'd be mouth watering low end parts!

Yeah I think with the last three generations - R300 -> NV40/R420 -> R520/G70, hardware has steadily outpaced software demands. Hopefully UE3 and F.E.A.R will put the X1300/7200 back in their place as useless 3D solutions

 

[Pete]

But what I was getting at was that another G70 variant may not necessarily be a wise economic decision upon the launch of the R520 even if it is faster.

If the top end R520s are all 512MB, we're 99% likely to see a response in kind from nVidia, and I think this would coincide nicely with an Ultra moniker (and pricetag).

What strikes me as very strange is that the XL will be available before the XT - have we ever seen a case where the slower SKU is released first even though it's based on the exact same chip except with lower clocks?

Not that I can recall (Xmas was being facetious when he said X700XT, I think), but the new importance of availability at launch may have contributed. IIRC, both the X800P and 6800GT were much more available than the XTPE and U, though both tiers were launched together. Maybe ATI wants to avoid some of the backlash that arises from low availability (and thus super-high initial prices).

nV might not have been under as much clock pressure with G70 simply because of their pretty sizable time advantage (and the fact that they seem to have traded clocks for die size, as ATI did with R300), so maybe they were able to shoot for more conservative clocks than ATI can with R520.

It just sounds to me like the XT is going to be hand-picked XL's that are capable of higher clocks.

That's always been the case, but it may be somewhat exaggerated here because of the apparent trouble ATI's had with R520 hitting their anticipated speeds.

I expect to see Crossfire aplenty tomorrow. I'd be surprised if this shed light on R520. I'd be very surprised if we saw RV515 launched tomorrow, or just Avivo or Kaleidescope revealed.