WOOT R300 at 400 mhz already!!!

[druga runda]

A Gigabyte board at Anand

http://www.anandtech.com/video/showdoc.html?i=1686&p=4

Can you believe it!?!


GPU to test the effectiveness of the cooling system. Being that the R300 is a brand new core and most overclocking utilities don't support brand new architectures until they have actually been available for a while, we didn't expect to be able to overclock the GV-R9700Pro. Much to our surprise, the latest version of Powerstrip (3.20) properly detects and supports overclocking the R300.

Powerstrip reported the core clock speed properly at 325MHz and reported the actual operating frequency of the memory at 310MHz (remember that the memory actually runs at 310MHz even though it transfers data twice per clock, which is why it is sometimes called 620MHz memory).

We played around with the settings and managed to overclock the R300 to an incredible 400MHz, an increase of 23%. The memory was only able to hit 337MHz DDR (effectively 674MHz), an increase of 8% but the core overclock was by far the most impressive of the two numbers.



One of the reasons that the overclock was so impressive is because of the sheer size of the R300 GPU itself. Remember that we're talking about a 110M transistor GPU built on a 0.15-micron process; we were shocked when ATI announced that they would be able to hit 325MHz on retail boards, but when we managed to reach 400MHz on the GV-R9700Pro we were amazed to the point of extreme skepticism.

In order to put our worries to rest, we then tried overclocking ATI's Radeon 9700 Pro card - the same card we used in our review. While we couldn't hit 400MHz on this card, we managed to get as high as 378MHz reliably. Without any additional cooling, this gave us a good deal of faith in ATI's ability to deliver an even higher clocked version of the R300 on their current 0.15-micron process. Should NVIDIA make an impressive showing of NV30 before the end of this year, it would not be out of ATI's reach to release a 350 - 400MHz R300 with faster memory to compete without any significant changes to the core.

It's clear that ATI is well aware of this potential of the R300, which is why they told us months ago that when the NV30 launches they will have something even faster up their sleeves. ATI doesn't need a 0.13-micron process to ramp up the clock speed of the R300, but they will need faster memory. Our astounding over****ing successes don't kill the need for a 0.13-micron R300, since ATI will eventually need to outfit each of their 8 rendering pipelines with a second texture unit once the move to faster DDR/DDR-II memory is made. What these results do mean however is that if the need arises, ATI can squeeze a bit more out of the current R300 (and so can yo






:eek: :eek: :eek:

Well this is getting better every day... I think Nvidia has a problem on their hands, bigger than it seemed at first.

WOW!

[pelly]

Yes...it's a fantastic card based on .15....
For the size, it is a great engineering success and a powerful card.

However, I highly doubt that this .15 card ( or any for that matter ) can compete with a well-engineered .13 part regardless of how high they set the core speed. Anand realizes this and draws attention to the fact that memory speeds will not be shifted to DDR-II until the shift to .13 is made.

The main limitation at this point seems to be memory bandwidth and die size; the lack of additional memory bandwidth and die size constraints prevented ATI from going to two texture units per pipeline, which would definitely increase performance across the board.

Our astounding over****ing successes don't kill the need for a 0.13-micron R300, since ATI will eventually need to outfit each of their 8 rendering pipelines with a second texture unit once the move to faster DDR/DDR-II memory is made.

[Bigus Dickus]

Originally posted by pelly
Yes...it's a fantastic card based on .15....
For the size, it is a great engineering success and a powerful card.

However, I highly doubt that this .15 card ( or any for that matter ) can compete with a well-engineered .13 part regardless of how high they set the core speed. Anand realizes this and draws attention to the fact that memory speeds will not be shifted to DDR-II until the shift to .13 is made.



Anand never says that DDR-II will wait for .13u. ATI could do it before, or after... no one knows. But, they have the ability to do it whenver they feel the need.

[Megatron]

ATI already has stated that they believe there is still plenty of headroom on the .15 process over and above what they have achieved with the 9700.

[Gargr]

WOW

And ppl said 325+ Mhz on .15 is imposible or something like that??
Do you think ATI will release some 400/750Mhz R300 when NV30 is released?? Or maybe a .13 part with even higher clocks :eek:


And im getting a GF4 Ti4200 :(
i need $$$ :mad:

[SnakeEyes]

You know, that sounds entirely possible. I just saw the comment elsewhere on overclocking, where the core went pretty high when overclocked. It seems as if all that would be needed would be faster RAM and some special edition 9700Ultra Pro's could be on the market, still using the .15 process.

Other than the RAM cost, it probably wouldn't (won't?) cost ATI much either, since it mostly would involve speedbinning to find the chips that will handle the even higher than the Pro core clocks, then pairing them with the faster RAM.

I still don't know what this means for nVidia, and especially nV30. I eagerly await real preview / reviews of that card. nVidia's history says they won't sit idle any longer than they have to when competing products start to get advantages.. :)

(Who here thinks it would be nice to have cards that will run all our games at max resolutions with high AA and high aniso enabled, while getting framerates that never dip below our monitor refresh rates? Seems like this might be possible, at least until DoomIII hits.)

[jbirney]

Originally posted by pelly
However, I highly doubt that this .15 card ( or any for that matter ) can compete with a well-engineered .13 part regardless of how high they set the core speed. Anand realizes this and draws attention to the fact that memory speeds will not be shifted to DDR-II until the shift to .13 is made.

Dont automaticaly assume that .13u part will be faster than a .15u part. Yes there is a really good chance. But a part can only be as fast as its slowest link. we dont know where that slowest link is with the NV30. ATI handtweaked their design to get their speeds. Do we know for sure nV has done the same? Besides what happens if there are other bottle necks (like NV stuck on 128 bit bus) or something else?

I think the NV30 will be faster. But how much? I dont know but I doubt it will be that much faster given the fact that the R9700 can hit 400 Mhz mark.

[Megatron]

Originally posted by SnakeEyes
It seems as if all that would be needed would be faster RAM and some special edition 9700Ultra Pro's could be on the market, still using the .15 process.

Sure seems so,
Heres a snip from a thread over at the dark side started by Hellbinder:

"ATI has strong relationships with the major fab companies and we are closely monitoring the progress of the new process technologies. We believe the .15u process still has some headroom to increase clock speeds beyond what was achieved with the RADEON 9700 PRO, and coupled with the availability of DDR-II memory, higher speed derivatives of this product can be expected in the coming months."

(This is a quote from some information given to Dave Bauman prior to his review (Beyond3d).. He seems to think that it indicates that ATi will be releasing a higher clocked version of the R300 with DDR II around the time of the Nv30 release.)

[John Reynolds]

Originally posted by pelly
However, I highly doubt that this .15 card ( or any for that matter ) can compete with a well-engineered .13 part regardless of how high they set the core speed.

By can't compete, you realize that there must be a significant performance gap between the two boards for that to hold true, don't you? And I highly doubt NV30 will blow the 9700 Pro out of the water, especially since the latter has months to both improve its drivers and drop in price. A $300 9700 Pro with yet another new set of drivers that have, with other updates, boosted performance 10-15% over initial scores, probably will be able to compete against NV30.

[druga runda]

Ati will probably release a higher clocked, perhaps DDRII R300 core for NV 30 debut or close, to spoil the party in good oldfashioned way, but I still doubt that there will be much of a difference, even though Nvidia will have 2 texture untis per pipe.

Reason being is that R300 is stretching all todays games to the max, so the difference will not be so great especially with newer/ faster R300. The diff will only be seen in 4xAA and aniso at 1280 + resolutins, as all the rest will be pretty much the same. And yes R300 will have pretty mature drivers by than, so that NV30 will need to be that engineering miracle Nvidia claims it will be.

However judging by CPU's that die shrink with process refinements, could push R300 speed for 1/3 rd I'd think so what would that make it.. 600mhz?

Woooot... but seeing this quite possible if they transition to .13 process and tweak the GPU a little more, as it is now in the bayb stage on .15 process and at 400 mhz.

Interesting indeed.

I am pretty sure now that NV30 will be at least a 400 mhz part, perhaps even more?

[pelly]

Will the NV30 annihilate the 9700 when it comes out? No, of course not....

Will the NV30 outperform the 9700 by a measurable amount? We don't know for sure but I would assume so. Especially after the drivers are optimized...NVIDIA is famous for getting every last bit of performance out of their chipsets. If the NV30 has as much headroom for overclocking as the 9700, ATI could be in trouble until their .13 part makes it to market.

Consider this fictional scenario ( nobody knows for certain ):

NV30 is faster than 9700 : ATI raises core frequencies and memory freuqencies ( maybe even introduce DDR-II ). At this point, they are out of any major options. The design is "maxed".

now 9700 is as fast or faster than NV30 : NVIDIA raises core frequencies ( yields will get better to allow this ) and memory frequencies. At this point, NV30 could still have some headroom as it has a smaller ( more efficient ) die-size and is using DDR-II.

Back and forth until one card runs out of options...Because NV30 is using slightly newer technology, I would assume that the 9700 would run-out of options first.

As far as the .13 vs. .15 comparison speed-wise, I think we can all safely assume that this would hold-true. There are so many benefits to the smaller die size which would definitely be taken advantage of by expert companies such as ATI or NVIDIA.

[saturnotaku]

Originally posted by Megatron
(This is a quote from some information given to Dave Bauman prior to his review (Beyond3d).. He seems to think that it indicates that ATi will be releasing a higher clocked version of the R300 with DDR II around the time of the Nv30 release.)

So if the NV30 is out on schedule for the Christmas season, assuming that people have received their 9700's now and the above statement is true, the 9700 will have been out only 4 months before they release a higher-powered refresh. That doesn't seem like a very smart thing to do because there's a very good chance a lot of current 9700 owners will be pissed that ATI released a more powerful card only 4 months after they shelled out $400 for what they have now.

Unless someone can clarify, this really doesn't sound right. :confused:

[druga runda]

Doesn't have to be Christmas, but it could be Feb as NV30's will start to come out in big numbers so Ati could come out with the card that beats it???

Or at least beats it on price/performance if NV30 yields are so low.

With the previews in December of course :D

[John Reynolds]

Originally posted by saturnotaku


So if the NV30 is out on schedule for the Christmas season, assuming that people have received their 9700's now and the above statement is true, the 9700 will have been out only 4 months before they release a higher-powered refresh. That doesn't seem like a very smart thing to do because there's a very good chance a lot of current 9700 owners will be pissed that ATI released a more powerful card only 4 months after they shelled out $400 for what they have now.

Unless someone can clarify, this really doesn't sound right. :confused:

Sounds like the GF3 Ti's to me. Released in November, with the GF4's coming out roughly 5 months later.

[John Reynolds]

Originally posted by pelly
Back and forth until one card runs out of options...Because NV30 is using slightly newer technology, I would assume that the 9700 would run-out of options first.

Honestly, I think your logic is 180 degrees backwards. Unless Nvidia can significantly outperform the 9700 with their .13u part, they will run out of options sooner. TSMC and UMC won't have a 0.9u process ready next year and ATi can move to both .13u and DDR2. No, Nvidia has got to significantly outperform the 9700 or pray that ATi has serious problems redesigning the chip for the .13u process.

[saturnotaku]

Originally posted by John Reynolds
No, Nvidia has got to significantly outperform the 9700 or pray that ATi has serious problems redesigning the chip for the .13u process.

I agree. Because ATI could very easily learn from any mistakes/problems/issues NVIDIA had with the .13 process and make sure not to repeat the history. I think we know the NV30 is going to be darn good, but it's still going to be a while before we have any conclusive evidence. To quote Tom Petty:

The waiting is the hardest part...

[Jandar]

Originally posted by SnakeEyes

(Who here thinks it would be nice to have cards that will run all our games at max resolutions with high AA and high aniso enabled, while getting framerates that never dip below our monitor refresh rates? Seems like this might be possible, at least until DoomIII hits.)

Thats what a new card is all about.

I wanna crank everything to 1600X1200 with 4XAA and 16X AF on my 21" trintron and game away.

Screw playing at lower settings.

[pelly]

JohnReynolds
TSMC and UMC won't have a 0.9u process ready next year

http://www.theinquirer.net/?article=5055



A REPORT IN English-language newspaper the Taipei Times, said that while major chip foundry Taiwan Semiconductor Manufacturing (TSMC) will be ready to produce 90 nanometer technology in the fourth quarter of this year, demand for such products is well nigh non-existent.

As for the .13 process, it will surely take much less time for ATI to produce a part this size. In this case, NVIDIA was the pioneer. This is a good thing, although it always makes them the "lab rat". They have to deal with all the initial fab problems that will likely be solved by the time ATI looks to make a .13 part.

[John Reynolds]

Originally posted by pelly
JohnReynolds


http://www.theinquirer.net/?article=5055





As for the .13 process, it will surely take much less time for ATI to produce a part this size. In this case, NVIDIA was the pioneer. This is a good thing, although it always makes them the "lab rat". They have to deal with all the initial fab problems that will likely be solved by the time ATI looks to make a .13 part.

I highly, highly doubt that link is accurate. But let's try this: I predict that the first NV30 refresh will still be fabbed at .13u. You agree or not?

[pelly]

I certainly agree, although I suspect that the shelf-life of that update will be surprisingly short.

Here's a question for you:

Will ATI play the role of pioneer and try to .09 or will they stick to the next step of .13?

If they stick at .13, imagine what a leap NVIDIA has on them since they've been designing at that size for a fair amount of time whereas ATI will be noobies.

;)

[Bigus Dickus]

Originally posted by saturnotaku


So if the NV30 is out on schedule for the Christmas season, assuming that people have received their 9700's now and the above statement is true, the 9700 will have been out only 4 months before they release a higher-powered refresh. That doesn't seem like a very smart thing to do because there's a very good chance a lot of current 9700 owners will be pissed that ATI released a more powerful card only 4 months after they shelled out $400 for what they have now.

Unless someone can clarify, this really doesn't sound right. :confused: Yet people here think this is precisely what NVIDIA will do... release the NV30 around year's end and have a "refresh" (some people are crazy enough to even suppose the NV35) ready for a spring release.

[Bigus Dickus]

Originally posted by SnakeEyes
(Who here thinks it would be nice to have cards that will run all our games at max resolutions with high AA and high aniso enabled, while getting framerates that never dip below our monitor refresh rates? Seems like this might be possible, at least until DoomIII hits.) *holds up hand*

[pelly]

Amen...

*the annoying brat who always raises two-hands*

:p

[John Reynolds]

Originally posted by pelly
I certainly agree, although I suspect that the shelf-life of that update will be surprisingly short.

Here's a question for you:

Will ATI play the role of pioneer and try to .09 or will they stick to the next step of .13?

If they stick at .13, imagine what a leap NVIDIA has on them since they've been designing at that size for a fair amount of time whereas ATI will be noobies.

;)

Again, I think your logic is wrong. Having a finished, completed architecture and moving it to a process that has had time to mature is far easier than creating a new architecture on a new process. The fab's tools will have matured, their yields will have improved, and the actual design and layout of the chip was worked out for its initial creation.

Oh, and to answer your question. Do you mean with their 9700 refresh (possibly the R350 we're hearing rumors of) or R400?

[DadGT]

I am yet again amazed at people's belief that moving from .15 to .13 um is a reasonably simple change. If AMD had problems and had to significantly revise the athlon from one layout .13 um to another to get speeds up, do you really think ATI can just reuse the current R300 chip design? No, they will have to alter many thing because things like cross talk and stray capacitence change when switching sizes. It won't be worth it. Not with the product lifetimes we are now seeing.

And who really thinks nVidia, after having problems switching to .13 um will help ATI solve any of their problems? That's like saying Intel will help AMD get the .09 um hammers out the door next year. Any help they get for not being first to .13 um will be from TMSC. But TMSC builds the patterns they are given. Things like cross talk between the transistors is an ATI design issue that TMSC can give only limited help on.

Oh, and my source for AMD having two design revisions for the .13 um Athlons was the HardOCP review of the 2400+ and 2600+ CPUs.