Printable View
Sen I can guarantee you that those numbers are wrong. There is not enough markup given those prices for profit, it has to be 2.5 to 3 times under retail to work. A $500 dollar retail card would need to wholesale @ $200 tops which means ATI's production costs for a card in total would need to be another 2.5 to 3 times under wholesale. So a complete card would cost ATI around $80 or less.
ATi only manufactures chips last time I checked, sapphire makes the Built By ATi cards.
wonder how much 512mb of GDDR 3 costs...
and what about a pcb?
a 7800gtx costs around $300 to make, this is the complete card ram pcb et all. this came from gibbo who is the buyer at overclockers in the uk, he has dealing directly with nvida and their partners so its pretty accurate. i would expect the r520 to be fairly similar maybe a bit more for the faster ram.
perhaps $300 is thier wholesale cost and not the cost of manufacture?
ATI top-end chipsets used to sell for about 40-60 bucks. but that was before sapphire got exclusive build rights for the x800 line. about $200 for a high-end card to about $250, figuring costs and labour, but that's only becasue of large-number runs. low-number runs become more expensive, of course.
Those are underwhelming numbers....sheesh.
its also unreleased and un final hardware.Quote:
Originally Posted by Vapor
and how many chips does one wafer make? :rolleyes:Quote:
Originally Posted by Der_KHAN
Unreleased by a couple of days....they had better scores when Macci was goofing around, IIRC (still at stock speeds).Quote:
Originally Posted by Revv23
Anyway, they've had working prototypes for HOW long now? It had acceptable performance a couple of months ago, what happened to that?
Anandtech released a pre NDA roadmap:
http://anandtech.com/video/showdoc.aspx?i=2532
I wonder what they mean when they say,Quote:
Originally Posted by vapb400
"First of all, ATI's traditional core design can do 'more' per clock cycle (at least on the R420 design) than NVIDIA."
Ever since the GF6 series the R300 design (which only til the R520 series has been used) NVIDIA's architecture has been doing much more per clock than ATI's.
I once read an arguement in favor of the higher efficiency of the R300+, but I forget it. I swear it made sense though. IIRC, it *MIGHT* have been more efficient in terms of performance/watt.
well I'm mostly refering to the shader architecture. I don't know what they were refering to exactly.
well it seems that ati only has 16 pipes max for retail
Quote:
The bottom line is that R420 has the potential to execute more PS 2.0 instructions per clock than NVIDIA in the pixel pipeline because of the way it handles texturing. Even though NVIDIA's scheduler can help to allow more math to be done in parallel with texturing, NV40's texture and math parallelism only approaches that of ATI. Combine that with the fact that R420 runs at a higher clock speed than NV40, and even more pixel shader work can get done in the same amount of time on R420 (which translates into the possibility for frames being rendered faster under the right conditions).
Geez, you guys never read the links they give, eh?Quote:
As an extension of the fp32 vs. fp24 vs. fp16 debate, NV40's support of Shader Model 3.0 puts it at a slight performance disadvantage. By supporting fp32 all the way through the shader pipeline, flow control, fp16 to the framebuffer and all the other bells and whistles that have come along for the ride, NV40 adds complexity to the hardware, and size to the die. The downside for R420 is that it now lags behind on the feature set front. As we pointed out earlier, the only really new features of the R420 pixel shaders are: higher instruction count shader programs, 32 temporary registers, and a polygon facing register (which can help enable two sided lighting).
That's all theoretical performance--I want to see an example (bench or game) where clock for clock it outperforms the NV40....I know they're out there, I'm just wicked lazy right now.
I did that, and they gave the NV40's second math unit almost no credit at all, like it never existed. There isn't always going to be texturing needed, and when there isn't, ATI's previous architecture can't compete with the NV40's shader pushing capabilities. This changes of course on a case-by-case basis.Quote:
Originally Posted by cadaveca
You don't need a example. The 6800GT and X800XL perform almost exactly the same. The GT is clocked at 350MHz, the XL is clocked at 400MHz. That's good enough for me, anyway.Quote:
Originally Posted by Vapor
Heh X850XT CrossFire performs like single X1800Pro in 3D2k5. X1800XT ought to be 7800GTX killer :D
You mean how many working chips does they make on single wafer ? ;)Quote:
and how many chips does one wafer make?
Rev.A R520 costed ~$350 for single working chip - when reasonable price has to be ~$50 per chip or less.
Silicon has to be clean, but several thousands for a wafer - that`s too much. several thousands for big piece of it to cut in slices - that sounds right to me.
With 600Mhz and 16 pipes? I don't think so. :(Quote:
Heh X850XT CrossFire performs like single X1800Pro in 3D2k5. X1800XT ought to be 7800GTX killer
my new word of the day :D (not intended to insult spelling error, just a funny word like suckle :P)Quote:
Originally Posted by Cooper
Is it me or does it seem like ATI is pushing crossfire with X8's series in mind? Who would want to buy a X8 anything at this point? I mean as soon as you buy it your out dated. Anyway as a ATI fanboy I'm very disapointed with EVERYTHING concerning ATI lately. For the past 6 months or so I havent even been able to recommend a ATI card, friends, clan members, family members. I have to refer everyone to Nvidia, geez I hate being a ATI fan.
No you're not an ATI fanboy. If you were you'd be ignorant of ATI's vulnerable position right now. Trust me, there are real ATI fanboys out there, even on XS, that would say with a straight face that some X800 series cards perform better and have better image quality than a 7800GTX, even with all the bells and whistles being utilized.Quote:
Originally Posted by Exa.bit
You're the good kind of "fanboy". The kind that likes a company, but still observes who offers the best solutions at the time and won't mislead those who want advice. Props to you. :toast:
Yes each wafer can on average make over 10,000 chips..and at the cost of $7,000 per wafer... it is less than $1 to make a chip.. and lets just say only 20% of the chips are used that is still less than $5... so what makes you think they cost $50 apiece?Quote:
Originally Posted by Cooper