Recent Tapeouts

[CMKRNL]

In the past few weeks the following have taped out:

NV31 in 0.13.
R350 in 0.15.
RV350. I'm fairly certain 0.13, but not 100% positive.

 

[Grall]

You sure about ATIs tapeouts? No announcements have been made, but maybe it is not in their interest to toot any horns right now?

*G*

 

[Evildeus]

In the past few weeks the following have taped out:

NV31 in 0.13.
R350 in 0.15.
RV350. I'm fairly certain 0.13, but not 100% positive.

Interesting, seems to confirm some trends.

 

[Bigus Dickus]

You've been pretty reliable in the past.

I expected to see NV31 in .13u and RV350 in .13u (which should replace R300 in the second half of next year as a similar performing lower cost chip/board).

What really surprises me is the R350 being taped out. I figured that if one existed, it would be .15u. However, I really expected to see a clock bumped DDR-II equipped R300 try and hold on against the NV30. I didn't expect a true core improvement until the R400 next fall.

This should really make things interesting. If the R300 and R350 are very similar (which one would guess they are), then the reference design for the PCB should be close. R350 could be brought to market fairly quickly. Will we see R350 boards then in March or April?

 

[Joe DeFuria]

Good ol' CKMNRL, er, CKLMNRN, er...damn whatever!

R350 in 0.15.
RV350. I'm fairly certain 0.13, but not 100% positive.

If RV350 is on 0.13, that would certainly explain the "contradictions" surrounding the R(V)350 and the process that it's on.

Oh, and Russ....stop reading now. More speculation on R(V)350.

Assuming R350 does now exist, and that it is a "superior" chip to the R300, I still do not subscribe to the "Add a second TMU and make pixel shaders match capability of NV30" theories.

I expect R350 to be little more than a "tweaked" R-300: mostly for faster clocks and possibly be better utilization of DDR-II. But who knows...ATI has been full of surprises lately!

I also suspect RV-350 to be a "real" 4 pipe version of the R-300. If that's true, and it's on 0.13 micron, it should be a Radeon 9000 replacement part to serve in the $100 market segment.

EDIT: And if the RV-350 turns out to be on 0.15 micron, then I would suspect it would be a replacement part for the Radeon 9500 non pro. Funnily, given the uncertainty of the 0.13/0.15 nature of RV-350...why not speculate that ATI has taped out two versions of the RV-350 chip?

 

[antlers]

And if the RV-350 turns out to be on 0.15 micron, then I would suspect it would be a replacement part for the Radeon 9500 non pro. Funnily, given the uncertainty of the 0.13/0.15 nature of RV-350...why not speculate that ATI has taped out two versions of the RV-350 chip?

Yes, a .15 micron 9500 replacement (or just a 9500? do we know when the plain old 9500 is due?) and a .13 micron to be the mobile DX9 chip, and perhaps the 9500 replacement when .13 micron process is cheap enough.

 

[MuFu]

From what I've heard about R400, the reason they are so excited about it is that they can move DX9 down into the "mainstream". So I'm assuming RV350 is a mainstream counterpart to R400 (not a Radeon 9500 replacement, as such) and we may see a staggered launch a la RV250/R300 next year - remember that they "sat on" RV250 for ages until the R300 launch (it taped out many months prior).

As for R350 - I must admit I have been apprehensive in the past that such a product even exists. They obviously feel it is necessary from a strategic point of view - to build on this apparent period of dominace in the performance sector. I'm pretty they'll remove some redundant funstionality and maybe remap the current die slightly to achieve ~400MHz operation.

MuFu.

P.S. I am also fairly certain RV350 is a 0.13u ASIC - from independent sources. Plus it is most likely the technological foundation of M10 - 0.13u and DDR-II looks likely for this "RV350M" part.

 

[WaltC]

'Kay, the only thing I've seen or heard about "R350" are rumors. People are picking up the FiringSquad nibblet which plainly states it is a rumor ("Word on the street is"....) Is this something real or merely rumors bouncing off each other and reflecting back?

I've seen some speculation here and elsewhere about an R300 w/ DDRII---for what? It's already got more bandwidth than the DDRII nv30 product announcement. What would be the point? I kind of thought the whole point of going to a 256-bit bus was to avoid DDRII for the time being. Personally, I don't see any need for it until, possibly, an R400 on .13 microns clocked @ maybe 450MHz +.

I'm fairly skeptical of ATI quickly jumping to .13 microns as a knee-jerk reaction to nv30, especially prior to nv30 actually shipping @ announced specifications. I wouldn't expect to see a .13 micron Rxxx before summer '03. But I do think a Q1 '03 release maybe of an "Ultra" R300 with faster DDR1, maybe some tweaked pipelines--maybe even some speed-binning and better cooling at .15 microns--for a speed of ~400Mhz would be pretty cool... (While beating the nv30's price by $100-$150US.)

 

[jvd]

walt i doubt they will need to beat the geforce fx pricing . If they bring out a 400+ chip i'm pretty sure it will beat nvidia's offering.

 

[DadUM]

From what I've heard about R400, the reason they are so excited about it is that they can move DX9 down into the "mainstream". So I'm assuming RV350 is a mainstream counterpart to R400 (not a Radeon 9500 replacement, as such) and we may see a staggered launch a la RV250/R300 next year - remember that they "sat on" RV250 for ages until the R300 launch (it taped out many months prior).

Your statements don't seem to make sense to me. Given past ATI product history, we can see a few things. The Rx00 products were top of the line. That means, R400 would be top of the line and Rx50 parts would be the parts to "move DX9 down into the 'mainstream'". How is it that R400 will be the part to bring DX9 down to mainstream? Isn't that what Radeon 9500 supposed to be?

Second, R250 was the "mainstream" design of R200, so the guess that RV350 would be the "mainstream" counterpart to R400 doesn't seem to fit. Given past numbering, R(V)350 should be a lower end R300 design cleaned for "mainstream" users.

 

[µße®LørÃ￾]

I think I heard that the Parhelia 2 (Pitou) has taped out as well......

Not that I'm buying another one that is

 

[Joe DeFuria]

I'm fairly skeptical of ATI quickly jumping to .13 microns as a knee-jerk reaction to nv30, especially prior to nv30 actually shipping @ announced specifications. ...

Well, ATI moving to 0.13 wouldn't be a knee-jerk reation to NV30, it would be primarily in the interests of lowering costs...additional speed boost is just a nice side benefit.

The thing is, up to this point, there was no RX50 part...just RVX50 ones. The difference this time, is that there is a more advanced process (0.13) just waiting to be exploited. Back in the previous gen, 0.15 was it. So for ATI do do something different this time around, shouldn't be a "complete" shock.

 

[Dave Baumann]

Something doesn't stack up here.

RV200 = R100 on lower process.
RV250 = Cut back R200 on same process.

If there is any sense at all in ATI's internal naming that would imply RV350 is a cut back R300 on .15. If it was a R300/R350 core in .13 then going by their their history you'd expect it to be called RV400.

 

[MuFu]

From what I've heard about R400, the reason they are so excited about it is that they can move DX9 down into the "mainstream". So I'm assuming RV350 is a mainstream counterpart to R400 (not a Radeon 9500 replacement, as such) and we may see a staggered launch a la RV250/R300 next year - remember that they "sat on" RV250 for ages until the R300 launch (it taped out many months prior).

Your statements don't seem to make sense to me. Given past ATI product history, we can see a few things. The Rx00 products were top of the line. That means, R400 would be top of the line and Rx50 parts would be the parts to "move DX9 down into the 'mainstream'". How is it that R400 will be the part to bring DX9 down to mainstream? Isn't that what Radeon 9500 supposed to be?

I meant that when they introduce R400, it's counterpart will be a DX9 part for the mainstream. R400 is hi-end, of course. The "V" is the important part of the nomenclature - you shouldn't read to much into the Rx00/Rx50 thing.

Second, R250 was the "mainstream" design of R200, so the guess that RV350 would be the "mainstream" counterpart to R400 doesn't seem to fit. Given past numbering, R(V)350 should be a lower end R300 design cleaned for "mainstream" users.

There is no R250. RV250 is the mainstream counterpart to R300 - hence RV350 to R400. There's my reasoning.

Dave - you mention process changes - I don't think that is that significant in terms of the naming conventions, although I see your point. They refreshed Rage6c onto 0.15u without changing anything (just prefixed stepping info). Since fab advances don't really correlate exactly with the annual/biannual product cycle I don't see why they should reflect them in the nomenclature. I'll see if I can dig up some more info regarding R350/RV350/R400.

MuFu.

 

[Mariner]

How about RV350 = 0.13 micron 4 pipeline 128-bit DDR at around 450MHz. This assumes that DDR2 will be much cheaper at the time than the larger transistor count required for 8 pipelines.

With improved efficiency, this might be able to hover around the performance of R300 in most situations similar to the equivalence of R200 & RV250.

 

[Bigus Dickus]

How about RV350 = 0.13 micron 4 pipeline 128-bit DDR at around 450MHz. This assumes that DDR2 will be much cheaper at the time than the larger transistor count required for 8 pipelines.

With improved efficiency, this might be able to hover around the performance of R300 in most situations similar to the equivalence of R200 & RV250.

That would make sense. The 9500 chips can't compete with the current R300 because of the speed limitations of a .15u die, and because the chips are full 9700 cores. On .13u, an optimized 4-piped variation could probably surpass 500MHz, keeping pace with or slightly behind the current R300 with half the pipes, and a 128-bit memory interface utilizing DDR-II for nearly the same bandwidth as the current R300. The part would be much cheaper, perhaps partly due to the memory architecture, but certainly due to the vastly reduced die size and packaging requirements.

I expect the R350 to be another monster of a chip on a 256 bit bus though. I can feel my feet getting warm already.

 

[Nebuchadnezzar]

R350 = Core enhached R300 w/ possibly PS & VS 3.0 [Source; diverse ATi employees said it would be 'DX 9.1'], 0.15µ, it should have 375+ MHz, DDR2 at 400+ MHz, should be the new Super High-End card (Or at least be even w/ the NV30)

RV350 = Cut down R300, i'm not sure of this, but it should have 8 Pipes on 128 bit bus or 4 pipes on 256 bit, 0.13µ (that's absolutley sure), 425+ MHz, DDR2 ~350 MHz on 256 or 450+ on 128, this thing should be as fast or a little bit faster than R300 and replace it (Just like the R9000 did to the R8500) ...and a lower clocked version to replace R9700 & R9500Pro (Pure speculation)

EDIT: (Damn, 2nd one , just like Mariner said, i could be olny 4 Pipe at 128 bit (But then they must clock it at ~500 MHz !!!)

And

MuFu wrote:

From what I've heard about R400, the reason they are so excited about it is that they can move DX9 down into the "mainstream". So I'm assuming RV350 is a mainstream counterpart to R400 (not a Radeon 9500 replacement, as such) and we may see a staggered launch a la RV250/R300 next year - remember that they "sat on" RV250 for ages until the R300 launch (it taped out many months prior).

That would mean that R400 is more than DX9 (DX9.5 ) and it places all 'old' R3X0 cores to the mainstream because the R400 will be new High-End. (That would be great if there would (at least ATi) be olny DX9+ cards on the shelves! 8) )

P.s. LOL i used a bit to much the word 'should'

 

[Dave Baumann]

IMO, R350/RV350 probably wont have any feature enhancments. I find it probable that if R400 is scheduled for later this year it will be their PS/VS 3.0 / DX9.1 part.

 

[Nebuchadnezzar]

IMO, R350/RV350 probably wont have any feature enhancments. I find it probable that if R400 is scheduled for later this year it will be their PS/VS 3.0 / DX9.1 part.

But then why should ATi have said that R350 would be DX9.1 ???

 

[Geeforcer]

But then why should ATi have said that R350 would be DX9.1 ???

What the hell is DX9.1, anyway?