R360 and RV360

[Seiko]

Re: R360 wont improve in too much in clock speed

Yep, I think anyone expecting the R360 to be a far higher clocked part is going to be in for a disappointment. I don't believe for one moment ATI set the R350 at the speeds they did if they where capable of higher clocks. Sure this refresh may add a stable 20mhz more but ultimately it's not going to be much faster on a .15 process.

I disagree.

I see no point at all for refreshing and taping out a new core for something like only 20 Mhz or so more. There's a difference between being "capable" of higher clocks, and being "capble of higher clocks, while meeting OEM power consumption guidelines for the board level product."

380 core/680 DDR-I might simply be the point where they are at the power-consumption limit, even if the R350 core can be pushed a bit higher with only minor impact on yield.

What will be interesting to see, is if the R360 comes with DDR-I, or DDR-II, and how much memory it will be available with. By using "only" 128 MB DDR-II, they may be able to significantly raise core and memory clocks, and still keep within power consumption limits of the current 128 MB DDR-I 9800, and the 256 MB DDR-II 9800.

Personally, I expecting something like 450 Mhz core, and 128 MB of 400 Mhz DDR-II.

I would guess, that if ATI offers a 256 MB version of the R360, (or faster than 400 Mhz DDR-II memort) that card might indeed need some exotic cooling and generally be some "OEM Unfriendly" part...or even a part that perhaps ony ATI sells directly themselves. (Basically just a PR piece for retaining the technology leadership crown.)

...but to expect .15 to continue to deliver is daydreaming imho.

That seems to be a recurring concern with the last 2 0.15u high-end products ATI made....and ATI has delivered both times.

I think the fact that we see the 9800Pro 256 with DDR2 still not pushing 400 is a genuine indication that the .15 process is pretty much at its limit.

Again, disagree...considering that the board likely meets some OEM guidelines for power consumption, and the board as 256 MB of memory on it.

Unless of course they too opt for a monstrous cooler!

My prediction (again): we might indeed see it, but only if there is a 256 MB version of this refresh product.

Hi Joe, yep all valid opinions although I have to point out that I didn't have any concern regarding the R300 core and it's original clock speed and pegged the R350 as being a very mediocre speed increase k so it was a lucky guess . Sure some doubted that those speeds could be reached on .15 but at the end of the day ATI delivered. Unfortunately the R350 didn't raise the clock speed bar enough and has obviously begun to show the upper limits for sensible cooling. I think this has been compounded with the latest 256 meg version not increasing the clock speed even though it has been paired with DDR2!

Although the R350 no doubt helped margins and that alone will help ATI I think as the competing architectures have become more balanced and dare I say similar in basic terms we are now going to see a pure clock speed race. If I had to draw an analogy to the big chip makers ATI probably find them selves in a similar position to AMD whilst nvidia have just released the P4 which although still not that much faster is on the road for rapid and consistent speed increases.

Of course if ATI has higher margins they may still come out on top as they can still claim the best IQ a feature I think a lot of the reviews aren’t highlighting enough! The only other areas I think ATI need to address is the DoomIII performance deficit and the low end dx8.1 card. We all know when D3 is released NVIDIA cards are going to a force to be reckoned with and with so many people watching the D3 space I think it’s essential ATI compete. Again I’d like to see the rendering paths and their performance/drawbacks explained in reviews but as most sites are still only just mentioned IQ I think that will be someway off. Until then I think it’s now NVIDIA turn to ride the wave!

 

[Joe DeFuria]

Re: R360 wont improve in too much in clock speed

I think this has been compounded with the latest 256 meg version not increasing the clock speed even though it has been paired with DDR2!

Hi Seiko,

Yes, it was paired with DDR-2, but don't forget the amount of DDR-II doubled from 128 to 256 MB. In other words (IIRC), the power consumption of 256 MB of DDR-II should be roughly the same as that of 128 MB of DDR-I, given the same densities.

I think demalion made some additional points on "why" ATI has reason not to "push" core clocks faster than "needed".

I say similar in basic terms we are now going to see a pure clock speed race.

I pretty much agree. And while nVidia has the advantage of being on a more advanced process, I'd say ATI has the advantage of better "power consumption" engineering.

If I had to draw an analogy to the big chip makers ATI probably find them selves in a similar position to AMD whilst nvidia have just released the P4 which although still not that much faster is on the road for rapid and consistent speed increases.

I see it differently. I don't see anything inherent in either arcthitecture that makes either one more conducive to fast clock rates. (Whereas the P4 has more stages and is thus designed for higher clock rates, relative to Athlon).

The position that nVidia is in now is a little unclear: is NV40 their next "top chip", or will they refresh the NV35? In that sense, nVidia is like AMD with the Athlon 64....NV40 could be great for them...if they can complete the design and production on time and on spec. Larger reward, but larger risk.

If Althon64 was out in time and on spec to compete with Northwood, AMD would be in a much better position. But it looks like they'll now be competing primarily with Prescott, which might not be so good.

We all know when D3 is released NVIDIA cards are going to a force to be reckoned with and with so many people watching the D3 space I think it’s essential ATI compete.

Agreed. Which is why I'm so distateful of how ID managed the Doom3 benchmark comparisons. What will be interesting to find out...is what will make a better Doom3 renderer?! A "new and more complex / flexible" NV40....or a much faster "current gen" core, like a NV35 refresh, or 0.13u R3xx (Loci).

In other words, NV40 might be a great part for the future, but it's not clear at all if Doom3 will be able to show it's promise.

Until then I think it’s now NVIDIA turn to ride the wave!

It all depends on the status and characteristics of NV40, IMO. If NV40 is on time (for fall '03,) and it is a better Doom3 renderer than a "super fast R3xx/NV3x type core", then yes, nVidia will have a good run.

If NV40 misses one of those, anything can happen!

 

[8ender]

My thoughts on these two new cores are simply that they are going to be released later this year and will fill out the low (RV360) and Mid end(R360) with Loci being released afterwards as ATI's new flagship card.

Likely the only improvements in these new tapeouts are to increase yields, make these cards cheaper to produce, and to make it easier to achieve higher clock speeds using less power.

This way, when Loci is released they can simply move the entire product line down a notch eliminating the only non DX9 part (R9000)

I mean its not like theres going to be another DirectX release anytime soon. If most of you product line already supports the latest well, simply make a faster card while making the older cards easier to produce.

To reiterate, I expect the ATI Winter/Spring Product line to look like this

Top End: Radeon X (R400 Loci?)
Mid End: Radeon 9800 (R360)
Low End: Radeon 9600 (RV360)

 

[Tim]

An interesting tidbit of information to throw into the mix here is the fact that there is another R3XX card from ATI waiting in the wings. We are told that this VPU will show a significant performance increase over any chipsets on the market including NV35 as well as R350. Expect to see this new ATI card somewhere between mid to late summer if it stays on schedule.

If this is true R360 is more than just a higher clocked R350 (or the R390/R420 will be released a lot sooner than expected), the If this is true the R360 must be more than just a higher clocked R350 (or the R390/R420 will be released a lot sooner than expected), the performance of a 450MHz R350 will perform very close to the 5900 Ultra.

But to beat the NV35 with a significant margin they need core changes, I would not be surprised if the R360 was 8x2.
performance of a 450/425 R350 will perform very close to the 5900 Ultra.

 

[Seiko]

But to beat the NV35 with a significant margin they need core changes, I would not be surprised if the R360 was 8x2.

I dont think we'll be seeing a 8*2 arrangment as I dont think it will help too much in future generations.

450/425 R350 will perform very close to the 5900 Ultra.

Yes agreed, in fact I'd wager to say that an equally clocked R350 will beat NV35 quite comfortably. Unfortunately I dont believe ATI can get to the same clock level and as such need to switch to .13 asap.

 

[Mariner]

I don't expect the forthcoming R360 will be anything more than a clock-ramped respin of R350.

Perhaps they will do one or two tweaks, increase the voltage and improve the cooling solution so the chip can be clocked near 450MHz with faster memory, of course. How high would a current R350 clock with a cooling solution similar to that of the NV35?

Ultimately, this R360 would be a 'bragging-rights' chip so the extra expense of the cooling + faster memory wouldn't be too important - it is due to be quickly superceded by the much faster Loci 0.13 micron respin according to one or two people on this board. I'd expect Loci to be more than just a speedbumped 0.13 chip, though, if it is to complete with NV40 as seems to be the consensus.

 

[Joe DeFuria]

My thoughts on these two new cores are simply that they are going to be released later this year and will fill out the low (RV360) and Mid end(R360) with Loci being released afterwards as ATI's new flagship card.

Likely the only improvements in these new tapeouts are to increase yields, make these cards cheaper to produce, and to make it easier to achieve higher clock speeds using less power.....

Top End: Radeon X (R400 Loci?)
Mid End: Radeon 9800 (R360)
Low End: Radeon 9600 (RV360)

I had thought of that possibility...however, ATI seems to be indicating that at least the R360 will be significantly faster than the R350. That being said, I am expecting R360 to ultimately be a "just" a faster R350. So ATI could discontinue the R350 chip, and ust use the "slower" binned R360 chips (clocked the same as the R350) in their current 9800 products. Thus:

Insane End: Radeon 9900 (R360...something like 450/425).
High End: Radeon 9800, using R360, clocked at 380/340)

Now, RV360 is an interesting question. The current RV350 seems to compete very well against the NV31, with the possible exception of Doom. (And Doom performance might be OK, depending on drivers.) So I agree that it is a possibility that RV360 could just be a respin tageting the same specs as RV350, just for better yield or power consumption.

Though I do also see the possibility for upping the core clocks a bit as well.

 

[Tim]

I had thought of that possibility...however, ATI seems to be indicating that at least the R360 will be significantly faster than the R350.

Not only significantly faster than the R350 but significantly faster than the NV35 as well. That is what leads me to believe that R360 most be more than just a higher clocked R350.

 

[Joe DeFuria]

Not only significantly faster than the R350 but significantly faster than the NV35 as well. That is what leads me to believe that R360 most be more than just a higher clocked R350.

Yes, well, whenever one IHV says "significantly faster" than a competitor's product, you have to take that with a huge lump of salt.

I don't rule out the possibility of some "significant" architectural tweaks, but i'm inclined to think not.

 

[RussSchultz]

Porting the design to .13u would presumably make it "significantly faster".

Since they've already got their analog stuff ported (needed for the RV350), the digital stuff should be pretty easy assuming the design flow, tools, and yield have gotten as good as people say they have.

 

[Joe DeFuria]

Porting the design to .13u would presumably make it "significantly faster".

Agreed...though we're talking about the R360, which is still 0.15u.

 

[MuFu]

Yes, it was paired with DDR-2, but don't forget the amount of DDR-II doubled from 128 to 256 MB. In other words (IIRC), the power consumption of 256 MB of DDR-II should be roughly the same as that of 128 MB of DDR-I, given the same densities.

Joe, I think 16 x 128b GDDR2 might actually consumer more that twice the power of 8 x 128b DDR-I at this point in time.

To reiterate, I expect the ATI Winter/Spring Product line to look like this

Top End: Radeon X (R400 Loci?)
Mid End: Radeon 9800 (R360)
Low End: Radeon 9600 (RV360)

I can't see that happening - R360 is too expensive to produce. I think it'll be Loci + RV400, with the RV360 dropping down into the low-end. This tallies with Orton mentioning the R400 cycle would see DX9 brought to the mainstream.

MuFu.

 

[Joe DeFuria]

Joe, I think 16 x 128b GDDR2 might actually consumer more that twice the power of 8 x 128b DDR-I at this point in time.

Wow, Really? DDR-II has no power advantages (actually disadvantages), vs DDR-I? What do you mean "at this time"...is that because DDR-II is not yet being manufactured with it's complete implementation?

I always assumed that DDR-II, since it's internal clock as 1/2 that of DDR-I, counsumed roughly half the power of DDR running at the same bus clock, all else being equal.

Do you have a source where I could read more info?

I think it'll be Loci + RV400, with the RV360 dropping down into the low-end. This tallies with Orton mentioning the R400 cycle would see DX9 brought to the mainstream.
MuFu.

Agreed...though my question is: will RV360 debut at the same time as the R360, replacing the RV350 at the mid-range, and then drop down to replace the RV280 once Loci is released?

Or will RV350 just hold in the mid-range until Loci is introduced, only being replaced and moved into the value segment by the RV360 at that time?

Speaking of RV400 -- any guesses? 0.13u stripped down version of Loci?

 

[MuFu]

Further thoughts...

RV350=>RV360 involves more changes than R350=>R360
R350 will be released in a couple of months but RV360 will not, instead being the mainstream counterpart to the RV400 and Loci.

This will cast aside the question of what name to give it, because they can use the nomenclature associated with the higher end products (as they did with 9x00). So assume the next generation is Radeon 10x00 (doubt it, but just for the sake of this speculation):

High: Radeon 10800 (Loci)
Mid: Radeon 10500 (RV400)
Low: Radeon 10200 (RV360)

As was mentioned earlier, they do tend to "sit on" their low-end cards for a while. I suspect the RV360 might be the RV350 with some cost-cutting measures taken and perhaps even further integration.

MuFu.

 

[MuFu]

Joe, I think 16 x 128b GDDR2 might actually consumer more that twice the power of 8 x 128b DDR-I at this point in time.

Wow, Really? DDR-II has no power advantages (actually disadvantages), vs DDR-I? What do you mean "at this time"...is that because DDR-II is not yet being manufactured with it's complete implementation?

Well, to be honest I'm not sure - I just don't think it's anything like half the power draw, as you mention. GDDR2 currently runs at 2.5/1.8V - 2nd generation GDDR2 (which is coming in Q2 and I believe is more compliant with the JEDEC DDR-II spec) runs at 1.8V for both core and I/O so will consume alot less power. I think the Samsung DDR-I chips ATi have been using run at 2.5/2.5V.

The thing that confuses the issue is on-die termination and I'm not too sure about the specifics myself yet. I think using ODT with GDDR2 increases power draw per chip - DDR-I has active trace termination on the PCB, so while the chips may consume less (relatively speaking), the peripheral components will consume more and bring the overall 3-rail draw up.

There are other reasons why they might have used DDR-II instead - saves a little on layout space, better signal integrity offsets the use of stub termination (which they'll have to use for some of the traces until Samsung bring out 256Mb modules), it can't be *that* much more expensive per chip than DDR-I now and the fact that they have a production qualified GDDR2 board is encouraging - good "practice", if you like. It probably has quite a bit of headroom and can be re-used for R360, saving a little on dev costs. They put quite a lot of work in to the R350 memory interface to get it working well with GDDR2, and while they probably had much higher clockspeeds in mind at the time, the enhancements will no doubt be carried over into R360, making them worth the effort. Weird way of looking at it, lol.

MuFu.

 

[Nite_Hawk]

Joe:

About ATI competing with themselves... I agree with you. I also think that they didn't give nvidia enough credit though. Back when they released the 9700pro, I remember a quote by one of the ATI employees (I think it might have even been Orton) that basically said that anyone who becomes complacent in this industry is going to get over run at some point. It "feels" to me like ATI got used to the 9700pro cash cow, and figured that if they pushed the 9800pro out farther, they could milk it that much longer. That in itself is fine, but while they were sitting on it, they should have been progressively upping the clock speed. The longer people have to wait for something, the more impressive they expect it to be. Just look at the NV30 reviews.

Nite_Hawk

 

[Joe DeFuria]

It "feels" to me like ATI got used to the 9700pro cash cow, and figured that if they pushed the 9800pro out farther, they could milk it that much longer.

I'll just have to disagree with you on this. The 9800 was launched about 6 months after the 9700 was. That's about as "soon" as you ever want to release a new follow-up high-end product. I don't see the 9800 being "pushed out further", or being "sat on" at all.

Also note that the 9800 256 MB is shipping right now. Not just launched. Many people seem to gloss over this fact.

ATI has the 9700, 9700 Pro, 9800, 9800 Pro and 9800 Pro 256 MB shipping right now.

The only actual competition from nVidia is the GeForce FX 5800 and 5800 Ultra....which nVidia basically doesn't even acknowledge exists.

In short, before you pass judgement about either schedules not being aggressive enough, or clocks not being pushed high enough...I'd wait until we see what the R360 is, and when it ships.

I will say that your posts highlights exactly one of the difficulties that ATI faces....they really don't have any actual competition in the high-end (and haven't for 9 months or so), but the perception is, they are actually behind!

 

[Tim]

Couldn’t the RV360 not be a NV34 competitor and not a mainstream part at all, it seems that the RV350 has plenty of headroom and have difficulty seeing a need for a respin.

It could also be targeted against the NV36, a product I find really interesting.

 

[Nite_Hawk]

Joe:

I think it's actually closer to 7 months (August to March) but I think we've pretty much wrapped up that part of the conversation anyway, so it doesn't really matter that much. I think you bring up a really good point though. ATI currently *is* the only company out right now with a viable highend product. There really is no reason to buy a 5800 series card at this point with the 5900 around the corner, and ATI cards performing better at the same price point. One of ATI's biggest problems is that nvidia has been extremely successful at making people believe that their cards are worth waiting for. They shot themselves in the foot with the NV30 to a certain extent, but look at how the masses of news sites have picked up the doom3 scores for the NV35. It's rather disgusting. At the same time though, the NV35 really *does* look like it's going to be competition in the highend. Still, given that nvidia seems to be the "microsoft" of videocards, ATI is going to need to provide a clearly better product to compete, otherwise the default for most people will be to go with a geforce. I'm not sure if the 9700pro has made enough of an impression yet to stop this.

Nite_Hawk

 

[Joe DeFuria]

One of ATI's biggest problems is that nvidia has been extremely successful at making people believe that their cards are worth waiting for.

Agreed. That is one of the benefits you get when you hold the leadership position for a while (Since the time of the original GeForce introduction). "Aggressive" marketing is another key.

They shot themselves in the foot with the NV30 to a certain extent, but look at how the masses of news sites have picked up the doom3 scores for the NV35. It's rather disgusting.

Again, agreed. Comes down to "aggressive" marketing in this case. Take a moment to poder what a different perception there would be of the NV35, had there been no Doom3 comparisons.

Still, given that nvidia seems to be the "microsoft" of videocards, ATI is going to need to provide a clearly better product to compete, otherwise the default for most people will be to go with a geforce.

YEs, that's a good point. Not having market leading mindshare means that your product has to be "that much better" for it to compete. The best thing for consumers, is for neither ATI nor nVidia to be perceived as the market leaders, but market co-leaders.

I believe things have been heading very much in that direction, which is a good thing.

I'm not sure if the 9700pro has made enough of an impression yet to stop this.

It certainly initiated a change from the market perspective. However, it takes more than a single "one-hit wonder" to make a permanent change. ATI must contine to deliver superior or at least equally competitive products.