Radeon9500 the only mainstream DX9 card available this year?
- Thread starter Sabastian
- Start date
Evildeus said:
What ever will happen with regards to the NV30 it will more then likely an announcement or paper launch of some sort. I seriously doubt that it will be more then just that.
Derek Smart [3000AD]
Regular
alexsok said:
It will probably come out early next year. From what I know, this year is out of the question. We [devs] are still at Beta 2 and MS is probably going to release Beta3 sometime next month.
As for ATI's mojo day, it is highly unlikely that MS will release DX9 around the schedules of the #2 (as distant #2 that is) video card maker. My bet is that they work around the #1 card (in case you were wondering, that would be nVidia) maker.
Derek Smart [3000AD said:
For whatever it's worth: Philip Taylor (MS) stated this two days ago on DIRECTXDEV@DISCUSS.MICROSOFT.COM:
But it'll be tight with Beta3 barely out...
Prometheus
Newcomer
And why do we need DX9 so bad?No DX9 games around!!!
I think microsoft should try to work with developers for developers and not just release a different version of DX so that graphics companys can better market their new products!!
I couldn't care less if DX9 comes this year or the next a long as devs can adopt it and really use it, not like DX8 which doen't seem to be much supported(pixel&vertex shaders which DX8 was do much advertised)
I think microsoft should try to work with developers for developers and not just release a different version of DX so that graphics companys can better market their new products!!
I couldn't care less if DX9 comes this year or the next a long as devs can adopt it and really use it, not like DX8 which doen't seem to be much supported(pixel&vertex shaders which DX8 was do much advertised)
nooneyouknow
Newcomer
Derek, with all due respect, I would urge you to review the latest NPD market number. They certainly paint a different story. For the US Retail numbers, ATI has 42% (they even had a press release on it). Hardly a distant #2.
You may be confused by the Nvidia marketing propaganda machine telling you that they are #1 by a huge margin, but that simply would not be true.
Also, there are more markets than just retail. Realistically, retail accounts for at most 20% total market. Distribution is another critical channel. I don't have the share numbers there as I think there is no third party tracking those. Then you have OEM, AIB, and mobile on top of that. I won't even go into world wide either. Bottom line, that may be the information that you have to the best of your knowledge, but it is certainly not fact.
Just my two cents worth and likely to be flamed.
You may be confused by the Nvidia marketing propaganda machine telling you that they are #1 by a huge margin, but that simply would not be true.
Also, there are more markets than just retail. Realistically, retail accounts for at most 20% total market. Distribution is another critical channel. I don't have the share numbers there as I think there is no third party tracking those. Then you have OEM, AIB, and mobile on top of that. I won't even go into world wide either. Bottom line, that may be the information that you have to the best of your knowledge, but it is certainly not fact.
Just my two cents worth and likely to be flamed.
Oompa Loompa
Newcomer
It is difficult to imagine how we could have it both ways, though.
For the R9500pro to be a $150 "mainstream" card that still yields good margins for ATI, it is difficult to understand how it could mount the same huge (presumably moderate yield) 0.15 um die, with only the memory bandwidth reduced.
I predict that a card with this configuration would need to cost at least $225 to be cost effective.
For the R9500pro to be a $150 "mainstream" card that still yields good margins for ATI, it is difficult to understand how it could mount the same huge (presumably moderate yield) 0.15 um die, with only the memory bandwidth reduced.
I predict that a card with this configuration would need to cost at least $225 to be cost effective.
Why wouldn't all 9500's be 4 pipe, and all 9700 be 8 pipe, with suffixes corresponding to memory speed? Having both an 8- and 4-pipe 9500 seems ill-thought out. It'd be cheaper if ATi only had to make the 8-pipe R300, but if they're making a 4-pipe variant, I'd expect them to use it across the whole 9500 line.
I'm done talking out of my derriere, now.
I'm done talking out of my derriere, now.
<speculation>
What if they all use the same die, but some won't run at 325Mhz and other's have issues in some of the pipelines? Now what if they could downclock some chips and disable pipes in other chips?
With being able to do this, they first try to make 9700 chips, turn some of the rejects into 9500-Pro chips, and some of those rejects into 9500 chips. This could maximize the return per wafer, thus making it financially feasible.
</speculation>
--|BRiT|
What if they all use the same die, but some won't run at 325Mhz and other's have issues in some of the pipelines? Now what if they could downclock some chips and disable pipes in other chips?
With being able to do this, they first try to make 9700 chips, turn some of the rejects into 9500-Pro chips, and some of those rejects into 9500 chips. This could maximize the return per wafer, thus making it financially feasible.
</speculation>
--|BRiT|
This goes back to the redundancy thing - which I now have reason to believe is what is going to happen (well, actually, ever since I read the 9700 instruction manual I suspected it, but I have heard more).
They are not actually going to fab a separate '4 pipe' card, merely use the ones that didn't quite yeild with 8 pipes and turn some off to make a 9500. Of course, one thing that could snag this plan is yeild - to plan along these lines indicates that they are betting on a low yeild for 9700 class R300's, however if the yeild is better than expected then they may not have enough parts that fall below that level.
This can be quite a dangerous path to take - with an expensive core you want to be pushing as much to the high end as you can; we see that with the introduction of Ti4200 this fast became NV's most popular Ti board, but obviously gave the lowest margins.
Yes you are correct, but from what I hear TSMC has lowered its price on the .15 micron process thus making the scheme cheaper. AFAIK the yields on the .15 um process are good and ATI is going into mass production of the R300 core. This whole process of using one core is beneficial in that ATi does not have to create a whole new chip design for another product and we all know how costly that can be. The rumors are suggesting the Radeon 9500 is not a 256 bus then ATI will have no problem selling the chips. With the case of the Radeon 9700 it is the 10 layer PCB (making it more expensive) needed to accommodate the 256 bus that card manufacturers don't like. If anything the margins on a Radeon 9500 chip should be similar to that of the Radeon 9700 chip because of the extra expense of the PCB on the Radeon 9700 on the card manufacturer but only slightly lower as it saves the card manufacturer to have the mid range 128 bus and needs only an 8 layer PCB.(This is speculation on my behalf and please correct me if I am wrong. Thanks in advance.)
I do think however that the yields on the R300 are acceptable and whatever the matter is ATI will produce the product to stay in the game but I doubt very much that ATi will be having margin issues such as they did with the Radeon 8500....... If anything ATi will experience a boost in revenues and the company who will suffer in their margins will be nvidia for the time being. Why do you suppose that nvidia is nearly trading at 10% of its all time high?
EDIT: Further the demand for the Radeon 9700 is high this helps ATi manage better margins.
That's a much more sensible take, Dave. The only reason I doubted it was because I thought it would be hard to know exactly how a chip may fail (i.e., are the pipes structured in a way that makes disabling half easy? are the pipes the most failure-prone component? will it have cut-down vertex shaders as well?) Also, I figured that if yields were good, ATi could always just sell a slower 8-pipe part as a 9700 (non-Pro). At this point, are we even going to see 9700's, or is ATi trying to make as much as they can with a premium part? Even nVidia released both the 4600 and 4400 simultaneously, IIRC, while they postponed the 4200.
If the 9500 Pro is indeed an R300 with slower memory, it should sell extremely well, as I'd wager people can settle for 10x7 for half the price. I'm not sure why they wouldn't pay the extra $40 for double the pipes, though. I guess we'll see street prices for the 4-pipe 9500 go much lower than those for the 9500 Pro.
Anyway, it's nice to see ATi pick up nV's slack and continue driving the market forward. I might just put off getting a cheap 8500/4200 and wait for a reasonable 9500/Pro.
If the 9500 Pro is indeed an R300 with slower memory, it should sell extremely well, as I'd wager people can settle for 10x7 for half the price. I'm not sure why they wouldn't pay the extra $40 for double the pipes, though. I guess we'll see street prices for the 4-pipe 9500 go much lower than those for the 9500 Pro.
Anyway, it's nice to see ATi pick up nV's slack and continue driving the market forward. I might just put off getting a cheap 8500/4200 and wait for a reasonable 9500/Pro.
It appears that both ATi and Nvidia are looking for DDR-II memory before it is due on the market. Also DDR-II is not expected to be on the market this year. This doesn't bode well for the NV30 to be matched with it ..... at least not by november. BTW if indeed the NV30 is matched with current DDR what is the likelyhood of it severly outperforming the Radeon 9700?
Graphics standard for DDR-II expected
By Jack Robertson
EBN
(09/27/02, 12:24 PM EST)
The JEDEC Solid State Technology Association, as expected, is formally working to adopt a DDR-II "look alike" standard for frame buffer memory of graphics cards.
Although JEDEC deliberations are confidential, sources from the standards body said the quarterly meeting this month heard presentations for a proposed graphics version of its basic DDR-II standard for PC main memory.
Graphics card makers were pushing for a fast track adoption by the end of the year of the DDR-II specification tailored to their specific needs. However the JEDEC sources believed this was overly ambitious and any DDR-II graphics version won't be finalized until early next year.
ATI Technology and Nvidia Corp. are expected to introduce graphics cards with DDR-II memory chips even before a new JEDEC standard is approved. Since graphics card memory is point-to-point mounted directly on the board, vendors can validate chips and work out interfaces themselves. The JEDEC graphics standard would be an added benefit by eliminating any of the small proprietary twists memory suppliers now try to add to their DDR-II chips to lock in customers.
http://www.ebnews.com/story/OEG20020927S0035
Graphics standard for DDR-II expected
By Jack Robertson
EBN
(09/27/02, 12:24 PM EST)
The JEDEC Solid State Technology Association, as expected, is formally working to adopt a DDR-II "look alike" standard for frame buffer memory of graphics cards.
Although JEDEC deliberations are confidential, sources from the standards body said the quarterly meeting this month heard presentations for a proposed graphics version of its basic DDR-II standard for PC main memory.
Graphics card makers were pushing for a fast track adoption by the end of the year of the DDR-II specification tailored to their specific needs. However the JEDEC sources believed this was overly ambitious and any DDR-II graphics version won't be finalized until early next year.
ATI Technology and Nvidia Corp. are expected to introduce graphics cards with DDR-II memory chips even before a new JEDEC standard is approved. Since graphics card memory is point-to-point mounted directly on the board, vendors can validate chips and work out interfaces themselves. The JEDEC graphics standard would be an added benefit by eliminating any of the small proprietary twists memory suppliers now try to add to their DDR-II chips to lock in customers.
http://www.ebnews.com/story/OEG20020927S0035
