-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
ATI Tape Out Three Successful!!!
Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Hey, what about Mountain Dew pop?! 😉
It's in Hot Deals:
http://forums.anandtech.com/messageview...atid=40&threadid=1636358&enterthread=y
I'm a bit surprised at everyones shock over a $600 launch MSRP on this card?
A. It has no competition other than 6800GT/U SLI, which cost more.
B. Prices of high enjd cards at launch have been going this direction for a while:
May 2003 high end pricing
October 2000
So high end cards have been launching at $500 for five years. You didn't think that would ever change?
Keysplayr
Elite Member
X, not to rain on the parade just yet, but it was not specified if the poor yields were just for a particular piped card. 32 or 24. Just said yields were not that great but did not specify. So, theoretically, if the R520 was intended to be a 24 pipe card all along, then that would mean poor yields, although high clockspeed, on a 24 pipe R520. We need at least a glimpse of what the actual specs for R520 really are and were always intended to be by ATI. So, I am hoping they are only referring to a 32 pipe poor yield, but I have doubts of course.
Only pure speculation on my part, but With a total of (3) tape-outs, I tend to think the first two were poor designs with non-working features or something, not just poor yeilds. I think R520 would have been shown more behind closed doors at E3 and Comdex(?, or whatever the other show was) but was'nt because it was not totally working.
imported_X
Senior member
As I read the article, it was poor yields that drove the retaping processes. Since the headline says the third retaping was successful, it seems logical that yields are now up to acceptable levels.
biostud
Lifer
Not knowing anything about production of chips, but I think I read something about that once the chips is designed they get a first sample that runs MUCH slower than final specs, probably a few Khz. They use this to find flaws and whatever you need to improve. They then continue with new revisions until they have a chip that works as it should without errors. Then they have a sample for the final production line and they then need to improve the producoction without too much wafer loss.
bunnyfubbles
Lifer
If it is indeed more than 600MHz it does make it hard to believe its more than 24 pipes - although 24+8 or some odd combo - but 32 and 600+MHz doesn't seem likely, you'd think it would flat out destroy the 7800GTX, even an Ultra would most likley be a futile counter attack.
600+MHz and 24 pipes with working and available crossfire solution upon launch seems the most likely/logical bet. A 24pipe part @ 600+MHz should give them the single card advantage, of which a 7800Ultra would feature a clock increase to close that gap. Then there'd be Crossfire to claim the multi GPU advantage nVidia has had with SLI. The only need for 32 pipes would be to attack SLI pretty much without Crossfire, especially at 600+MHz.
Although then there's the problem of such low yeilds, it doesn't seem likely they'd have so much trouble if the card wasn't going to be a beast. However more than 24 pipes and more than 600MHz sounds too good to be true, although if it is the price will definately come to match, I would expect over $600, $700-800 even... yikes.
600+MHz and 24 pipes with working and available crossfire solution upon launch seems the most likely/logical bet. A 24pipe part @ 600+MHz should give them the single card advantage, of which a 7800Ultra would feature a clock increase to close that gap. Then there'd be Crossfire to claim the multi GPU advantage nVidia has had with SLI. The only need for 32 pipes would be to attack SLI pretty much without Crossfire, especially at 600+MHz.
Although then there's the problem of such low yeilds, it doesn't seem likely they'd have so much trouble if the card wasn't going to be a beast. However more than 24 pipes and more than 600MHz sounds too good to be true, although if it is the price will definately come to match, I would expect over $600, $700-800 even... yikes.
24 or 32 it wont matter as long as there is something to challenge the 7800 then we can hope to see some price drops. The INQ also goes on about the 7800 going to 500 shortly too and a lower version of the 7800 street pricing at 400 meaning we can see some 6800 price drops coming also. Im even seeing the 6600gt cards going to 150 yesterday after rebate.
This is great I am looking forward to getting a new video card soon.
Keysplayr
Elite Member
Ah, I see what your thinking. But a successful tape out does not automatically mean acceptable yield as the article says.
As it states: "IT TOOK ATI three tape-outs of R520 silicon to make it right but it turned out to be third time lucky. The first R520 silicon worked at high speeds but only few chips worked, we hear."
So, you can have a 100% successful tape out, but that does not mean good yields are guaranteed.
No, it doesn't necessarily guarantee good yields. But from the statement about the first tapeout "but only few chips worked" to "third time lucky" it could be inferred that ATI is now satisfied.
I'm just crossing my fingers and toes that there isn't a FOURTH. I'd like to see what this card is all about sometime in the near future and how it stacks up against the 7800GTX.
thecoolnessrune
Diamond Member
I thought 100% yields were impossible. I thought because things were produces so fast, there would always be some that didn't work. Just think, there may be hundreds of AMD X2 4800+ that could be sold for $250 because they ran 100 Mhz slower 😀 Or Pentium D 840 EE for $300. Or FX-57 for $400. Or 7800GTX for $150. Drools into supplied cup. Empties cup. repeats.
Concillian
Diamond Member
I believe the 'tape out' process is the process of making all the masks necessary for the photoresist developing. These determine the topography of the chip.
They couldn't possibly go from taping out to finished product in a week given the complexity of the chip. It most certainly takes more than 7 days from start to finish of a wafer of chips of this caliber, probably more on the order of 2-3 weeks. However, who knows how far along in the process they were when the Inq published their last article.
The 'tape out' process nowadays is simply uploading a computer file to a machine that makes the masks. I would guess that the masks aren't incredibly expensive either, but that the primary cost comes in terms of the rental of TSMC fab time, labor and resources... and the wafers themselves.
I would presume the cost of a 'tape out' estimated by the Inq in the previous article includes:
- design costs for making the new revision
- the real 'tape out' process (making the masks, etc...)
- cost of TSMC resources to make the 'test batch'
- the cost of all involved reliability and yield testing necessary for the 'test batch'
I don't know for sure, but those are my guesses.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
