ATI X600 and X300 officially announced?

[DSC]

http://www.digit-life.com/news.html?105369#105369

http://www.techreport.com/etc/2004q2/ati-pcie/index.x?pg=1

http://www.anandtech.com/video/showdoc.html?i=2066

http://www.firingsquad.com/hardware/ati_pci_express/

X600(RV380) and X300(RV370) doesn't support 3Dc, PS2.0b or any features from X800. Both are 4x1 just like the 9600XT.

 

[Bjorn]

X600(RV380) and X300(RV370) doesn't support 3Dc, PS2.0b or any features from X800. Both are 4x1 just like the 9600XT.

3Dc can be "emulated" by the DX5 alpha channel trick. And i guess it's questionable if no PS2.0b support is a problem since these cards won't be able to run shaders of that length in realtime anyway.

What i want to know is, where's the X700 ?

 

[Lezmaka]

I figured X600/X300 would be based off of X800, but I guess I should have gone with my gut (eg, code names). It looks like ATI wants something in mid-low range for PCI Express now and new mid-low range cards are at least several months out. But what do I know, I'm just guessing.

 

[Dave Baumann]

Interview with Dave Orton[/url]]Well, I think four is needed now.

What’s interesting is that you could argue there are more offerings down below, but that’s not the case. We used to design to a 10x10 die, then a 12x12 and then the R300 was 14x14 and now we’re at 16x16, so you see what’s really happened is we’ve added more SKU’s in the high end as time has gone by, so if you’re at 16x16, what’s the next one down?

We think that there still is a market for a $10 part. Integrated isn’t taking over the world, so there’s a huge opportunity for discrete parts out there for about a $10-$12 selling price. So, with that at the bottom end and you’re high end part you ask if one will fill the gap, and the answer is probably not, you might need two. So, how you implement those total 4 SKU’s is the question – is it 3 design centres and some migrant technology or is it 4 design centres?

However, as you’ll see this fall from us, and our RV lines, a total of 4 parts too.

 

[Joe DeFuria]

Well, everything makes perfect sense so far.

There's no earthly "reason" to have the R4xx core down in the low-to-mid end right now. The few extra features of the R4x tech just don't make the low end any more marketable, and would only serve to increase chip costs, AFAIC.

What the low end "needs" is acceptable, (4 pipe) PS 2.0 performance, reasonable AA (Multisample, not Supersample) options. Anything else is just fluff.

However, as you’ll see this fall from us, and our RV lines, a total of 4 parts too.

I take this to mean that this fall, we'll see ONE more new part. At this time, we have:

RV370, RV380 and R420.

For a total of three new parts. What's "missing" is a theoretical RV420, to fill the gap between the X600 and the X800 Pro. This is currently filled quite nicely with the Radeon 9800 / 9800 Pro actually, but I'm sure ATI wants to replace that chip with a cheaper one to reflect current market pricing. An 8 pipe R420 based card on 0.11 should fit the bill nicely, running with either 128 bit or 256 bit memory bus interface.

 

[Dave Baumann]

OK, now this is a development since TSMC's messaging last month:

TSMC's 0.11-micron process technology is fundamentally a direct shrink of its industry-leading 0.13-micron process. The process will be available in both high performance and general-purpose versions. Though actual results are design-dependent, TSMC's 0.11-micron high performance process also includes transistor enhancements that improve speed and reduce power consumption relative to its 0.13-micron technology. It also delivers devices that are 15 to 20 percent smaller in size. When produced in TSMC 300mm Fabs, this combination results in a significant cost savings on a per-die basis.

I don't think that means they are going to be offering Low-k variants, although it might hint that they are getting more and better benefits from the shrink than initially expected.

 

[Joe DeFuria]

OK, now this is a development since TSMC's messaging last month:

Most interesting. I was just thinking about the notebook part implications of the 0.11 process. With only the "low cost" 0.11 process, an 8 pipe part might not be particularly attractive as a notebook offering due to power consumption.

This news now makes it more likely IMO that the RV420 will be built on the "high performance" 0.11 process. This would make the RV420 a nice candidate to replace the R350 chip in the mid-range discrete segment, and also be the new high-end notebook part. (RV360/RV380 moving down a notch.) Finally...8 pipes for notebooks.....but will there be a 256 bit bus for notebooks as well?

 

[Thunderbird]

So what happened to offering disabled quads further down the line besides just the x800 pro?

 

[Bjorn]

Finally...8 pipes for notebooks.....but will there be a 256 bit bus for notebooks as well?

I would think that having lower clocked RAM and a 256 bit bus would be a benefit for laptops when it comes to heat and power consumption. And i wouldn't bother with a 8 pipe card without a 256 bit bus to go along with it.

Interesting news when it came to the 0.11 micron process though. I wonder if this is a new discovery ? And if it is, maybe the IVH's are going to have to change their plans a bit.

 

[Joe DeFuria]

So what happened to offering disabled quads further down the line besides just the x800 pro?

I'm thinking this is very "unprofitable" for anything beyond disabling one quad. I see something like a "2 quads disabled" R420 based board as more of a "back-up plan" for ATI, which they would only release under one of two (or both) following conditions.

1) Working 3 and 4 quad Yields are so poor, that it would make more sense for ATI to sell 2 quad disabled chips, than to just throw them out.

2) nVidia puts out a killer $250-$300 card, and ATI needs something to compete with it, so they put out a dual quad disabled R420 just to prevent handing over that market space to nVidia by default.

I'm guessing that the "so far theoretical" RV420 is ATI's part for the $250-$300 market, coming this "fall." If at all possible, ATI would prefer to just let the current stock of Radeon 9800 parts ride it out until then.

However, lots will probably depend on what the 6800 non-ultra shapes up like. (Actual performance, price, and availability.)

 

[Joe DeFuria]

I would think that having lower clocked RAM and a 256 bit bus would be a benefit for laptops when it comes to heat and power consumption.

So would I, actually. But lower clocked ram AND a 128 bit bus is even better.

And i wouldn't bother with a 8 pipe card without a 256 bit bus to go along with it.

Here, I disagree. I'm sure lots of people would welcome a Radeon 9500 Pro type card in a lap-top. This will be particularly true if shader based games actually hit the market soon (Half-Life2). To be clear, I would certainly offer an 8 pipe card with the option for 256 bit bus, but I would definitely support 128 bits.

 

[Bjorn]

Here, I disagree. I'm sure lots of people would welcome a Radeon 9500 Pro type card in a lap-top. This will be particularly true if shader based games actually hit the market soon (Half-Life2). To be clear, I would certainly offer an 8 pipe card with the option for 256 bit bus, but I would definitely support 128 bits.

A 128 bit bus would be ok if they could use reasonably high clocked RAM which i doubt.

Though i agree that it will be interesting to see how shader limited f.e Half Life 2 turns out to be. Perhaps a 128 bit bus won't cause any troubles with a 8 pipe card. At least not in new games.

 

[anaqer]

Somewhat related, but thinking in the desktop space... what would be cheaper overall (eg. PCB cost included) : good old DDR1 (slower ones even) on 256bit, or the relatively fresh GDDR3 on 128bit?

 

[CapsLock]

Trying to connect the dots from limited data... I'm wondering since we haven't seen or heard anything about the X800SE 8 pipe product, if this could actually be the 4th product line (RV420) on .11 instead of an R420 with two quads disabled.

The 2 quads disabled theory doesn't make sense since this is going to be a high quantity production product which couldn't be supplied by defective R420's (unless yields are horrible).

That could be quite cost and performance effective I think.

Caps

edit. thats what I get for jumping ahead as Joe seems to have already said this at myself (again).

 

[jvd]

I'm guessing the 8x1 x800 se wont be coming out anymore . I'm figuring the yields are really good on the 16 part and on the 12 part. Perhaps we will see a 12x1 with a 128 bit bus as the se ?

 

[Joe DeFuria]

I'm guessing the 8x1 x800 se wont be coming out anymore . I'm figuring the yields are really good on the 16 part and on the 12 part. Perhaps we will see a 12x1 with a 128 bit bus as the se ?

That's pretty funny...because I was just thinking the same exact thing. This could be a feasible product IF, there is a Half-Life 2 benchmark that shows performance relative to say, the 9800 Pro.

 

[anaqer]

Perhaps we will see a 12x1 with a 128 bit bus as the se ?

That would be slower and more expensive than a 9800, HL2 or not. Even if someone buys a card for future shader-heavy games, it's expected to run current titles good, and I mean GOOD. With AA - not just 2x either. And you can't make such a card cheap enough to match the performance and still get a nice profit from it. I sure hope we won't ever see such a beast.
CapsLock's guess fits the picture much better IMO, and is in line with the 0.11 and "a fourth ASIC in fall" talks.

 

[jvd]

I'm guessing the 8x1 x800 se wont be coming out anymore . I'm figuring the yields are really good on the 16 part and on the 12 part. Perhaps we will see a 12x1 with a 128 bit bus as the se ?

That's pretty funny...because I was just thinking the same exact thing. This could be a feasible product IF, there is a Half-Life 2 benchmark that shows performance relative to say, the 9800 Pro.

Well if half life 2 uses gdc then you would see an increase over the 9800pro. Also the p.s 2.0 performance should be higher than the pro. Esp if they clock the core high still .
The x800pro is what 475. Figure 425 mhz for the se.

so you have a 425x12 part verses the 371x8 9800pro. I think the p.s performance would go up enough to justify this part

 

[jvd]

Perhaps we will see a 12x1 with a 128 bit bus as the se ?

That would be slower and more expensive than a 9800, HL2 or not. Even if someone buys a card for future shader-heavy games, it's expected to run current titles good, and I mean GOOD. With AA - not just 2x either. And you can't make such a card cheap enough to match the performance and still get a nice profit from it. I sure hope we won't ever see such a beast.
CapsLock's guess fits the picture much better IMO, and is in line with the 0.11 and "a fourth ASIC in fall" talks.

It will run current games better or the same as a 8x1 on a 128 bit bus but will run future shader limited games faster because of the higher clock speed and more pipelines .

 

[Joe DeFuria]

Even if someone buys a card for future shader-heavy games, it's expected to run current titles good, and I mean GOOD.

If it's running "current titles" at 100+ FPS, is that not good enough?

CapsLock's guess fits the picture much better IMO, and is in line with the 0.11 and "a fourth ASIC in fall" talks.

Of course...that's my guess too.

The question is...what does ATI do if the theoretical RV420 is late, or ATI otherwise needs to compete with something at the $300 price point to combat an nVidia product sooner rather than later?

The RV420 isn't here yet....so in it's absece, what would you do for the $250-$300 market segment, if you needed a "new" product in there right now?