Ati moving to 0.13 faster than I thought?
- Thread starter Joe DeFuria
- Start date
megadrive0088
Regular
yeah, whatever process the new chip is, on 0.13 or 0.15, its still a new chip. The R350. why would ATI have to tape out another R300, unless it was a 0.13 R300. R350 implies a refresh, an enhanced core, even tho the architecture is the same or very similar.
Joe DeFuria said:but we're talking about new chips being taped out.
Joe, more than one source confirms that the another Canadian company has also taped out their next generation chip... and I doubt it being to 0.15µm...
based on time line on these confirmations I have received, it has been happened already some time ago or then they have had already more than one tape-out for chip. So, either they (at least) have first samples some time ago or then they are already resolving possible bugs out of it.oh well, only time will show the critical questions: what and when...
Joe DeFuria
Legend
All right...
Since I consider myself a middle of the road kinda guy, I say the next ATI product will be based on 0.14!
Since I consider myself a middle of the road kinda guy, I say the next ATI product will be based on 0.14!
Joe DeFuria said:
LOL!
that was a good one
Thanks for that Joe
I can go to bed smiling. and you are right...
who cares about the tech, if it does the trick for my purposes and it is answer to my needs... that's the main thing everyone at least should be up for. Not some fancy tech and marketing names that are just made up to get you to buy the product though you wouldn't necessarily need it.megadrive0088
Regular
an 8 pipe Parhelia?
Prometheus
Newcomer
I can't understand why everyone says Ati can't have a 0.13 card this spring!!Is ATI that useless that it can't follow nvidia?I'm waiting for an official announcement and then get my hopes up but just think about it,Ati's product cycle is 9 months IIRC.
megadrive0088
Regular
With ATI's concurrent development teams now firing on all cylinders, there is no reason they couldn't have new chips every 6 months. I fully expect R350 in Q1 or Q2.
what R350 consists of is another matter though.
what R350 consists of is another matter though.
When ATI does move to .13u I would be really interested to know what road block they encounter. Where the simular to the ones NV hit with the NV30? Different? We all know that its not a simple process to do this with dies around 110 million transistors so something will come up. Just what and the work arounds is what I would like to know....
Ichneumon & DaveBaumann
suspicion or sources?
Ichneumon & DaveBaumann
suspicion or sources?
Bigus Dickus
Regular
Check out CMKRNL's new thread on recent tapeouts.
I though we would see RV350 next year, but I figured R350 was a phantom part... nothing more than a clock bumped DDR-II version of R300.
It looks like there is a true R350 chip on .15u that has been taped out recently. I wonder what improvements they could possibly have made to the core to boost performance, given the apparent miracle they pulled with clock speed the first time around, and the rather large transistor count of the chip for a .15u die.
More flexible shaders to compete with nV's PR? That's wouldn't really affect performance. Retooled for even higher clock speed? I highly doubt a second TMU... but what other reasons could there be for a new tape-out?
Perhaps this needs its own thread.
I though we would see RV350 next year, but I figured R350 was a phantom part... nothing more than a clock bumped DDR-II version of R300.
It looks like there is a true R350 chip on .15u that has been taped out recently. I wonder what improvements they could possibly have made to the core to boost performance, given the apparent miracle they pulled with clock speed the first time around, and the rather large transistor count of the chip for a .15u die.
More flexible shaders to compete with nV's PR? That's wouldn't really affect performance. Retooled for even higher clock speed?
I highly doubt a second TMU... but what other reasons could there be for a new tape-out?Perhaps this needs its own thread.
Nappe1's post appears to have ache!
Bigus Dickus said:
By far the most straightforward assumption would be a simple die shrink, but it seems that isn't the case.
When I typed up a response to DB's post on the apparent naming scheme contradictions (the main gist was to point out the number part was mainly relative to "when it was designed", and the V part only had to do with whether it was cheap for ATI to make or not), I did some thinking on what each could offer (before I discarded the post as it took a much longer and convoluted path to get that point across).
I stand by my guess from 2 months ago (I won't requote it again as I did EDIT: before) that the R350 makes the most sense as a professional level card. This would be from greater vertex processing power (to compete with the nv30 which is not so cost bound in the professional segment and seems like a pretty viable product there as a result), and perhaps greater quality (for example, though the practical gains from 96-bit to 128-bit might not matter, professional users may insist on it). Also, with the implications about displacement mapping (still not clarified, but circumstantial evidence points at less functionality than expected for the R300), this "quality" issue might be addressed as well.
Now, the first could be done by simply clocking higher, but really that isn't the only way. I think if they've had this design on the drawing boards for a while, all of the above are possibilities.
The one thing I don't really see a need for is significantly more bandwidth.
I even think this fits in with my general impression of specific statements from ATI concerning their lineup this coming year (I really need to do a through search of Rage3D to find these statements...perhaps I'll add a link later).
More flexible shaders to compete with nV's PR? That's wouldn't really affect performance. Retooled for even higher clock speed?
Perhaps this needs its own thread.
I don't see more flexible shaders, seems a waste of effort unless this was far easier for them to add than I'd think. The R400 is plenty of time to address that, IMO. I think the 128-bit "thing" might be a perception issue in the "professional" target market they are trying to create ("cinema shading"). I think the clockspeed thing is possible, but I just can't see how they can go further on 0.15 (but we couldn't see before, so...
) . Finally, it seems to be a clear trend that a 2nd TMU is pretty much a concept that keeps coming up because of a fixation on how things were done in the past. The earliest I see a 2nd TMU is R400, but with R300's (and nv30's) performance with just one, I don't even see why there would be a point. Anyone?