R350 in .13?

tt_22 said:
Randell said:
why not - makes sense to me.

test 0.13 on a smaller, less complicated core.
Has this been done ever before?

Intel does it all the time. Just look at their cpu launch history. The new generation core is often lauched on an older, more mature process. Example: P4 was originally introduced on 0.18 (Williamette core?), then transitioned to 0.13 (Northwood core?) as they gained more experience with the newer process.

IIRC, Intel first made the transition to 0.13 with the PIII Mobile.
 
Randell said:
tt_22 said:
Randell said:
tt_22 said:
ATI still have the R9700s to improve the 0.15 core on. I don't expect them to drop the R9700 line very soon :)

what do you think the R350 is, if not an improved R300 core?

I was thinking improved in case of yields and corespeed. R350 will be an improved R300 core, but with some additional features, like 2 TMUs?, VS3.0?, PS3.0? Even if the R350s are 0.13 they can still improve the R300s, maybe release R9700 ultra with 400 core.

How would that help? They would just be incurring production costs and polluting their product line with too many similar products. I'm sure part of ATI's reasoning is that they wanted to ensure any FX beating part was not delayed and was prodcued on a reliable process, if core speed is up I'm sure signal integrity is part of the current issues at 0.13 which is less of an issue at lower clock speeds (e.g. value part speeds).

All my speculation of course.

Sorry abt all the quoting here.

I think we all agree ATI releases its next card to counter the NV30.
With a little more cooling and some ddr2 memory the R9700 would be more than capable of that.

I don't see that ATI would take such a risk by moving the R350 to 0.13. Leaving the R350 on 0.15, you could also risk it to be slower than the R300 considering it is supposed to be more advanced.
 
tt_22 said:
I think we all agree ATI releases its next card to counter the NV30.
With a little more cooling and some ddr2 memory the R9700 would be more than capable of that.

I don't see that ATI would take such a risk by moving the R350 to 0.13. Leaving the R350 on 0.15, you could also risk it to be slower than the R300 considering it is supposed to be more advanced.

hmm we agree, except I bebelive the R300 needs fillrate not bandwidth, so higher clock speed is more improtant IMO than DDR2 (more expense, not much gain over 256bit DDR at this moment in time).
 
At this point ATI is in the driver seat and an r300 die shrink using a troublesome process may not be worth their while, especially with r400 waiting in the wings. With RV350 the stripped down version there isn't much reason to shrink r300 or 350 (at least right now) unless the process has improved to the point where the engineering and design costs could be recovered over the short time frame such a product would be desireable.
 
Hmm...I think you are "mystifying" technology, tt_22.

DDR-II is desirable because it allows higher clock speeds than DDR (DDR-"I"), not because it is called DDR-II. With 500 MHz DDR modules available, and with R300 core designs looking competitive at just above 400 MHz, it seems simply a matter of which is cheaper for the necessary performance (I'd guess DDR, not DDR-II).

0.13 is desirable because at the the same complexity, the same clock speed, and, most importantly I think, the same level of maturity, you get each chip for reduced cost. The choice for process seems to me determined by the complexity the process allows at a target clock speed and the level of maturity of the process, and what you can achieve within those constraints.

ATI has demonstrated they can achieve quite a bit on 0.15 given these concerns. 0.13 is not a magic wand that can be waved, it is a tradeoff of the above, where a greater complexity would tend to be allowed, but with a lower maturity at this point.
The R350 need not be slower than the R300, even if more features are offered...the question for the R350 is about where its optimization (redesign of some elements to achieve better performance or allow higher clock speeds) and enhancements (adding or changing features for improved functionality at a given clock speed) balance, and whether any "extraneous" functionality was dropped in the process of this balancing.
 
I believe that the r350 will be the r300 at speeds of 400+ (hopefully 450mhz) prob a better designed core allowing for that extra speed . Also there will be to versions out. a 256meg card and a 128 meg card.

Its just a refresh people. Thats all ati needs at this point and I'm sure they've been aware of that fact since the geforce fx was delayed . I can see them all slaving away trying to have the r400 ready to strike at any moment than adding to much to the r350. Thats just my two sense. I mean Why stick around to beat the tar out of an opponent that you've already beaten when you could be preparing for the next one ?
 
According to digitimes the R350 is :
- 0.15, using a lowcost 10 layer pcb
- feature low power consumption, to target both desktop and notebook

Is this plain wrong Hellbinder?
 
They probably just want to limit it's overclocking potential to keep it from canibalizing on the R400.

that's why the thicker process.

then you can shift the existing products to become more attractive.

9900 - > high-end
9700 - > value
9500Pro - > mid-range
????(RV350) - > low end
9500 - > disappear (OEM only)
 
I dunno .


I believe the r400 will be around 600mhz on the .13 and feature 10-12x1 config . come with 256 megs of ram and out perform the r300 by 2xs .

so i see this

r400 -400$
r350 - 300$
9700pro-250$
9700- 200%
9500pro 100$ (or whatever replaces it)
 
tamattack said:
Intel does it all the time. Just look at their cpu launch history. The new generation core is often lauched on an older, more mature process. Example: P4 was originally introduced on 0.18 (Williamette core?), then transitioned to 0.13 (Northwood core?) as they gained more experience with the newer process.

IIRC, Intel first made the transition to 0.13 with the PIII Mobile.

Intel has their own fabs and is hands on for the whole production process. Intel has engineers to work out the problems. Intel has more money. Until the last few years Intel had no competition (and 0.13 or not, the PIII was still slower than the Athlon C). Need I go on?
 
Nagorak said:
tamattack said:
Intel does it all the time. Just look at their cpu launch history. The new generation core is often lauched on an older, more mature process. Example: P4 was originally introduced on 0.18 (Williamette core?), then transitioned to 0.13 (Northwood core?) as they gained more experience with the newer process.

IIRC, Intel first made the transition to 0.13 with the PIII Mobile.

Intel has their own fabs and is hands on for the whole production process. Intel has engineers to work out the problems. Intel has more money. Until the last few years Intel had no competition (and 0.13 or not, the PIII was still slower than the Athlon C). Need I go on?

I agree 100%. Intel are probably the the best in the world when it comes to IC production. They threw a huge amount of resources on the 0.13 transition and still they had problems for months.

Patrik
 
glappkaeft / Nagorak: no argument from me... but what's the relevancy? If the company with the best fabs in the world finds it beneficial to operate this way (ie: new product on older process, slow rampup of products on new process), then why wouldn't it make sense for a fabless company to do the same?
 
Intel is the best in world at Die Shrinking because they got endless pockets and huge engineering teams. They have the two key things to be successful, people and lots and lots of money..

Sales:

press20020723.gif
 
Back
Top