NV45?

lyme said:
Did ATI praise the yeild rate? or just haven't said anything about it?

They've praised the 130nm yeild rate. Bit early to tell what things are like with a larger ASIC such as R420 though.
 
lyme said:
4) As far as I know both ATI and nVidia do not design their chips by hand, they use automated tools. (unlike intel and amd who do).
AMD uses automated tools to a large extent, the chips will off cause undergo some hand tweaking during their lifetime. Intel has also used automated tools for Prescott.
 
rainz said:
OCZ will release 650mhz memory sticks .. but dunno if we can apply the same speed for the video cards

RainZ

OCZ sucks as a company. Just search the reseller ratings page. A lot of people report on the items purchased from them either not working at all or merely being cherry-picked memory that even then don't work all that great at the advertised specs. And as been stated before, their 650Mhz memory is nothing but DDR at 325Mhz.
 
DaveBaumann said:
http://biz.yahoo.com/rc/040512/tech_ibm_1.html

"We seem to have turned the corner on those (design issues). We believe we understand them and we are now making rapid progress on them. But we still are not quite to our target -- to our objective -- but we are getting very close," Kelly said.

Seems like it's going in the right direction at least.
 
I still don't understand why people want the nv45 to come out so soon . Personaly when i bought ym geforce sdr and 4 months later the ddr version came out with much much better performance for the same cost i was pissed.

Same with my ti500 and the geforce 4s .

I can understand if the card sucks .

But don't u want to be safe in the knowldege that for 6 months your card is going to be the best from that company ?

Thats what i liked with my 9700pro. Thats what i liked from my 8500.
 
jvd said:
I still don't understand why people want the nv45 to come out so soon . Personaly when i bought ym geforce sdr and 4 months later the ddr version came out with much much better performance for the same cost i was pissed.

Well I have no intention buying any card from this generation at all. I hope technology makes another huge leap late this year or early next. :)
 
Well I'llmost probably buy this gen card(ATI most probably). My GF3 is starting to hurt now with all the new games coming out. I really can't see SM3.0 coming out in force(not at least for the next 6 months) and the X800 Pro seems a good preformer at a low cost.

Prices from South Africa.

ATI X800 XT = $499

SA exchange rate = R6,83/$1 so should = R3408

The price I've just seen by someone here is R4500

X800 Pro IceQ R3650
X800 XT Platinum IceQ R4500
Radeon 9800PRO R2500

We tend to get ripped off by suppliers here :( but I guess they have to make money too.

I intend to import my card/get someone to buy it for me and send it as a present.

US
 
AlphaWolf said:
Well I have no intention buying any card from this generation at all. I hope technology makes another huge leap late this year or early next. :)

I have my doubts about another technology leap like this one. Mostly because of the "problems" with manufacturing processes. I have no idea on how many transistors it'll take to double the performance again but i suspect that it'll take a 300+ million transistor chip to double the performance of the NV40. And i doubt that such a large chip will become easy/cheap to manufacture anytime soon.

And Ati needs to switch to FP32, SM3.0 (or better :)) for it's next gen so they'll probably not be in a better position either.

Though perhaps there are huge things to be done when it comes to the efficiency of the shading pipeline and maybe we'll see some large advancements there instead of just adding more shading units.
 
Good omens for R420 -

http://www.digitimes.com/NewsShow/Article.asp?datePublish=2004/05/14&pages=A2&seq=3

Taiwan Semiconductor Manufacturing Company (TSMC) and United Microelectronics Corporation (UMC) have both improved yield rates for processes using low-k dielectrics and now expect output to increase in the next few months.

TSMC employs black diamond (BD) low-k material at its 0.13-micron copper and 90nm processes. TSMC has shipped 30,000 8-inch equivalent wafers using low-k processes since its adoption in the second half of last year, according to company sources. TSMC shipped about 10,000 low-k wafers in 2003, according to a February 3 EE Times report.

In a technology forum this month, company president Rick Tsai said TSMC would start shipping 10,000 low-k wafers per month “in a few months.â€￾
 
So that means they've done 20,000 wafers in 2004. I'm too tired, anyone wanna do some back of the envelope calculations assuming all of these were for R420, and 100% yields, how many cores this could be?
 
Bjorn said:
AlphaWolf said:
Well I have no intention buying any card from this generation at all. I hope technology makes another huge leap late this year or early next. :)

I have my doubts about another technology leap like this one. Mostly because of the "problems" with manufacturing processes. I have no idea on how many transistors it'll take to double the performance again but i suspect that it'll take a 300+ million transistor chip to double the performance of the NV40. And i doubt that such a large chip will become easy/cheap to manufacture anytime soon.

And Ati needs to switch to FP32, SM3.0 (or better :)) for it's next gen so they'll probably not be in a better position either.

Yes I've been wondering the same thing. I wonder if we're going to see the re-emergence of multi-chip solutions in consumer space within the next 1 or 2 generations from NV/ATi.
 
DemoCoder said:
So that means they've done 20,000 wafers in 2004. I'm too tired, anyone wanna do some back of the envelope calculations assuming all of these were for R420, and 100% yields, how many cores this could be?

Around 2.4 milion.
 
DemoCoder said:
So that means they've done 20,000 wafers in 2004. I'm too tired, anyone wanna do some back of the envelope calculations assuming all of these were for R420, and 100% yields, how many cores this could be?

Assuming a 16x16 mm die, they should get ~102 die per wafer for a maximum total of 2 million.

Realistically, assume 50% of the wafers are R420 and 50% yeilds for a total of 500K.

Aaron Spink
speaking for myself inc
 
I'd imagine more than a few wafers went to RV360s, no? More than to R420, I'm thinking, if the midrange sells that much more than the high-end. Would it be fair to say that your 50/50 figure is still somewhat optimistic, Aaron?
 
More than to R420, I'm thinking, if the midrange sells that much more than the high-end.

Although ATi would be just as keen to get as many R420s into the marketplace as possible, it's their flagship product. Besides the RV360 dies is a lot smaller than the R420, so they can get that many more dies from a wafer. 50/50 wafer production split actually seems a little down for the R420 being honest.
 
Pete said:
I'd imagine more than a few wafers went to RV360s, no? More than to R420, I'm thinking, if the midrange sells that much more than the high-end. Would it be fair to say that your 50/50 figure is still somewhat optimistic, Aaron?

Depends what the sale rate of the rv360 is. Assuming it is a ~10x10 mm die, you'll get ~280 die per wafer. I would assume the yield is fairly high for the RV360 (~80%), so 10K 200mm equivlent wafers would work out to ~2.2 Million parts.

Anyone know if they are still making RV350 based parts as well? I would assume that ATI is doing much more than 2M 9600 XXX parts in ~4 months. Besides, you already have the mask set for the non low-k part and costs should be lower for that process.

Anyone have any idea how many 9600 XT's ATI has sold in the last couple of months?

Aaron Spink
speaking for myself inc.
 
aaronspink said:
Anyone know if they are still making RV350 based parts as well? I would assume that ATI is doing much more than 2M 9600 XXX parts in ~4 months. Besides, you already have the mask set for the non low-k part and costs should be lower for that process.

They are trying to push this further and further down the range, so, yes they will still be making RV350 as the wafer costs will be lower and there is only a clock speed advantage for the low-k variants.
 
Ante P said:
NV45
Bus interface: PCI-E 16x
Memory width: 256 bit
PCB: P211
Performance: 3x 5950U (same roadmap says 2x 5950 for NV40U)
Sample Availability: April/May
Retail Availability: Early Q3

Interesting indeed. I wonder how they will give the NV45 50 percent more performance than the NV40U but the retail availability in early Q3 is even more interesting: Maybe the NV40 was mainly about getting a cost-no-object chip out to regain the lead before the R420 while waiting for the real profitable NV45 to hit the market?
 
Back
Top