9700 OC to 450/800 a GFX Killer

[jbirney]

IMO, 5% would be very disappointing

I would also have to agree. Given that the FX has about an 200 mhz core speed advatange I am hoping more for a mere 5%. Then factor in that new cards are always suppose to be faster than "older" cards so I would expect it to be much faster than the 9700pro.

[shmall]

Originally posted by Shadowx
But for comparison if nv30 comes at 450mhz of speed or 500 we get to see how a similar clock 9700 will perform and know if the r350 really can compete with nv30. If the nv30 blows a r300 at almost the same speed and r350 been a r300 upgrade then r350/nv30 migth be close. if not then r350 can beat nv30 by a fair large margin. What do you think?


I heard that the nv30 will run at 500/1000... so you will need to bump up that 9700 a bit more yet for a comparison....


when are 9900pro coming out??

Simon.

[ReDeeMeR]

Originally posted by -=DVS=-
Sure it runs fast with Dustbuster on lol :rolleyes: :D realy Nvidia should do something about that , but high cloak's are not everything its still limited by 128 bus , ******* why nvidia could not have made it with 256 bus :mad:

I hear ya!!! :mad:

[Shadowx]

http://www.nordichardware.se/artiklar/Intervjuer/English/nVidia/index.php

In this Nvidia interview They say that Maximun Pc Benchmarcs are real so maybe the 128bit will infact hold back the GFX.

[ZoinKs!]

Originally posted by [eNv]-LORD-eX-Bu
... but what I don't understand is, why make your high-end part .15 and your mainstream .13?:confused:
Because they don't want their part to be delayed half a year. It is simpler to make a slower, less complex mainstream part on a new process. It is more difficult to make a very fast, very complex, high-end new part on a new process.

Consider the nv30 as an example. First nVidia had delays designing the part, mainly due to 128 bit precision. Then tsmc had trouble getting .13 manufacturing going. And then, even after the nv30 chip itself is ready, there is a complex, 12 layer pcb required. The result is a fall part becomes a spring part, and the competition has taken the lead.

ATI wants to get both design and manufacture using the .13 process ironed out using a relatively simple mainstream part. Then they'll be ready to bring out the r400.

I'm curious to see how far they can take .15 with the r350, and even more curious to see what they'll do with the r400.

[T-Spoon]

Originally posted by ZoinKs!
Because they don't want their part to be delayed half a year. It is simpler to make a slower, less complex mainstream part on a new process. It is more difficult to make a very fast, very complex, high-end new part on a new process.


Yeah, just look at other manufacturers.

The SIS Xabre600 is manufactured on 0.13micron process and is a low-end/mainstream product.

The Trident XP4 is also manufactured on 0.13 process and is also a low-end/mainstream product.

So it's not very strange for ATi to make their low-end/mainstream product a 0.13 part.

[StealthHawk]

Originally posted by T-Spoon
Yeah, just look at other manufacturers.

The SIS Xabre600 is manufactured on 0.13micron process and is a low-end/mainstream product.

The Trident XP4 is also manufactured on 0.13 process and is also a low-end/mainstream product.

So it's not very strange for ATi to make their low-end/mainstream product a 0.13 part.

Xabre is also a low transistor part with terrible performance(once you turn up quality to include trilinear).

Trident's cards aren't even out yet. this part is even more delayed than gfFX :angel2:

[T-Spoon]

Originally posted by StealthHawk
Xabre is also a low transistor part with terrible performance(once you turn up quality to include trilinear).

Trident's cards aren't even out yet. this part is even more delayed than gfFX :angel2:

The Trident XP4 card has been reviewed by ExtremeTech eventhough it was an early model. It sucked though.

I was just pointing out that low-end cards are more likely to be produced on a smaller process than higher end cards, because of low transistorcounts.

The fact that the SIS and Trident cards are performing badly is not a sign or indication that ATi will also perform badly on 0.13 with their low-end/mainstream RV350-part.

It's a smart move from ATi to use the value VPU as their first 0.13 product instead of trying to make a hi-end (and very complex) 0.13 VPU without any experience. This way they will probably avoid the problems that nV has had with GFFX.

The 0.15 process still has headroom for some improvement and the RV350 will be cheaper than a 9500 because of the die-shrink while (probably) performing better.

[StealthHawk]

Originally posted by T-Spoon
The Trident XP4 card has been reviewed by ExtremeTech eventhough it was an early model. It sucked though.

the point being that it still ain't out in retail. we had preliminary benchmarks of this card over 4 months ago. we had a "review" in early December.

maybe it's not a coincidence that the XP4 is on .13? that's all i was trying to suggest.

more likely it is that Trident's engineers, budget, or software engineers just don't cut the mustard.

[T-Spoon]

Originally posted by StealthHawk

more likely it is that Trident's engineers, budget, or software engineers just don't cut the mustard.

That's my point exactly. :)

[thcdru2k]

the 9700 doesn't need to oc to 450/800 to kill the gffx. i think it already did

[mreman4k]

More than likely the GFFX will surpass the r300 only to be tied if not surpassed again by a cheaper r350. This also means that prices on all those r300s are coming down.:eek:

[YeuEmMaiMai]

simple


The problems were not worked out with .13u to Ati's satisfaction when the R350 taped out and they wanted to be SURE that they can answer Nvidia Nv30. After all ATi states that the NV30 would be countered by the R350 is conditions warrent.
All I got to say is thank God ATi did not try .13u on their high end part like Nvidia did.

AS for NV30 working at 700Mhz maybe with some form if liquid or refrigerative cooling...

question is can the 400Mhz part keep pace with Ati's card? as it seems that the 500Mhz part will be slightly faster or on par in some cases

I do not think so since it takes 500Mhz to get 350Mtriangles and 15.6B AA samples a sec so at 400Mhz it would be 262.5Mtraingles and 11.7B AA samples a sec a far cry from the R300's 325Mtriangles and 16B AA samples


IT seems like ATi is playing it smart and staying out of the "we've got XXX" and just letting the R300's AWESOME Mind blowing performance speak for it's self...

Originally posted by [eNv]-LORD-eX-Bu
remember ben6 said that nVIDIA has GFFX working at 700 Mhz engine in their labs? I don't think R300 can compete much with that, RV350 tho, that might, since it is .13 micron, maybe it will be able to clock high, but what I don't understand is, why make your high-end part .15 and your mainstream .13?:confused:

[Chalnoth]

Originally posted by YeuEmMaiMai
AS for NV30 working at 700Mhz maybe with some form if liquid or refrigerative cooling...
Yes, overclocking the NV30 much past 500MHz is going to be challenging at best. I think that at this speed, the NV30 is already far beyond where the heat/clock speed ratio is linear. While somebody might be able to do it with more exotic cooling (yes, more exotic than the FX Flow...), it's not going to happen often.

But, due to manufacturing constraints, I would expect overclocks in the 550MHz range to be relatively common.

[Renzo]

Given that the FX has about an 200 mhz core speed advatange

Is it advantage really? Perhaps the GPU isn't as powerful as R300 in clock-to-clock comparison (like P4 vs AXP) and it has to run at high frequencies to compete with R300. Anyway the performance is sum of all parts and I think 128bit memorybus isn't enough even at high frequencies.

Will the GFFX be well balanced remains to be seen, hopefully we'll see some results soon :retard:

[shmall]

Looks like I will get a 9700pro after seeing the GFFX(ultra) benches here:

http://www.tecchannel.de/hardware/1109/

look how much behind the FX is behind the 9700pro in FSAA x4.


http://www.tecchannel.de/hardware/1109/images/0011963_PIC.gif


and the benchmarks that don't use FSAA are about the same for both cards.

big dissapointment for me being an nVidia fan for years.

for the price the 9700pro seems the better choice.....

look out ATI... here I come...... :)


Simon.

[Shadowx]

Well i hope NV dont start to go down, 3dfx miss two cycles by performance or delay and went under. With r350 comming out rigth now NV30 and NV35 dont look good. I think NV35 will need liquid nitrogen for cooling and a 256bit bus to compete with r350 NV is just behind. And i really think r350 will be 425-450/800-900 and that looks tuff to beat by NV35 given NV30 performance. What do you think will NV catch up with NV35 or will it take NV40 to do it?

[Chalnoth]

Originally posted by Shadowx
Well i hope NV dont start to go down, 3dfx miss two cycles by performance or delay and went under. With r350 comming out rigth now NV30 and NV35 dont look good. I think NV35 will need liquid nitrogen for cooling and a 256bit bus to compete with r350 NV is just behind. And i really think r350 will be 425-450/800-900 and that looks tuff to beat by NV35 given NV30 performance. What do you think will NV catch up with NV35 or will it take NV40 to do it?
The primary reason for the heat problem of the NV30 appears to be that nVidia couldn't make TSMC's low-k dielectric process work. By this fall, that should certainly be fixed. Thus, the NV35 should run much cooler than the NV30 does (at the same clock speed), and will likely be clocked a little bit higher.

From the reaction across most of the web, it is very likely that nVidia will realize very quickly that placing such an extreme cooling solution on a video card is not a good idea (whether or not this idea is sent home will depend on sales, of course).

[mikechai]

The report from xbitlabs will certainly give an idea of what the R350 performance is like.
It will have 25~35% increase in performance as compare to the R9700 Pro.
It is certainly very good for ATI as NV35 is no where near completion.
________
hot box vaporizers (http://hotboxvaporizers.com)

[Shadowx]

__________________________________________________ __
The primary reason for the heat problem of the NV30 appears to be that nVidia couldn't make TSMC's low-k dielectric process work. By this fall, that should certainly be fixed. Thus, the NV35 should run much cooler than the NV30 does (at the same clock speed), and will likely be clocked a little bit higher.

From the reaction across most of the web, it is very likely that nVidia will realize very quickly that placing such an extreme cooling solution on a video card is not a good idea (whether or not this idea is sent home will depend on sales, of course).
__________________________________________________ __

But it looks like NV30 will compete with r350 FEB/MARCH (NV30 is not shipping yet), so NV35 will compete with r400 so NV is one cycle behind when will they catch up ATI NV40? "IF" ATI has problems with .13 process on r400? or will it take them Longer:confused:

[Chalnoth]

We'll see, Shadowx. I've said in the past that I really do not believe that ATI can keep up their current pace. Die limitations will catch up to them soon. And it will take a while for nVidia to drop out of the race, even if ATI somehow does manage to stay ahead. nVidia just has more resources than ATI, and so I seriously doubt they will be behind for long.

As a side note, I think far too many people are putting the FX in much too negative of a light. I really think that the high-end FX beats the 9700 Pro in performance, while the non-Ultra is very slightly slower. Though the FSAA is certainly not as good, the FX has other benefits, such as more comprehensive-quality anisotropic filtering. That and the FX certainly has further to go in performance from driver enhancements than ATI, meaning that I feel that the non-Ultra FX will surpass the 9700 Pro in performance rather soon.

[Onde Pik]

Originally posted by Chalnoth
We'll see, Shadowx. I've said in the past that I really do not believe that ATI can keep up their current pace. Die limitations will catch up to them soon. And it will take a while for nVidia to drop out of the race, even if ATI somehow does manage to stay ahead. nVidia just has more resources than ATI, and so I seriously doubt they will be behind for long.

As a side note, I think far too many people are putting the FX in much too negative of a light. I really think that the high-end FX beats the 9700 Pro in performance, while the non-Ultra is very slightly slower. Though the FSAA is certainly not as good, the FX has other benefits, such as more comprehensive-quality anisotropic filtering. That and the FX certainly has further to go in performance from driver enhancements than ATI, meaning that I feel that the non-Ultra FX will surpass the 9700 Pro in performance rather soon.

LOL!

[Sazar]

Originally posted by [eNv]-LORD-eX-Bu
remember ben6 said that nVIDIA has GFFX working at 700 Mhz engine in their labs? I don't think R300 can compete much with that, RV350 tho, that might, since it is .13 micron, maybe it will be able to clock high, but what I don't understand is, why make your high-end part .15 and your mainstream .13?:confused:

um.. if you care to check the specs of the gddrII memory @ samsungs website you may notice that the design specs say 500mhz... thats the designed max speed of the product... hence if you are running it @700 you think that may be a little on the HIGH side?

also because 500 is the listed max speed (or max recommended... lets be fair) do you REALLY think nvidia will try and put a product out there running @ those speeds when there is a possibility it may not be as stable as it is a little slower?

as for chalnoth... other than some of your blindly nvidia pr related comments you are correct...

the product is not a failure and I do expect performance to be made up as time goes...

however it will still have a 128bit memory bandwidth and its IQ is not going to visibly be affected greatly... which is where MY greatest disappointment with the product lies... I was not very happy with the IQ @ all with this product..

ati has taped out products on the 0.13 micron process and are still taking nvidia products to the limit with their 'old' technology so where exactly will their die limitations arise? they are on the same die size as the gf FX and have EXACTLy the same number of 0.13 micron parts retailing as of this date (feb 1st 2003)...

the non-ultra part of the nv30 lineup struggles to keep up with the 9700 NON-pro hence your statements there concerning the products is not entirely correct...

now... when you have a high LOD and have AA and FSAA @ the levels people EXPECT to play @... instead of what nvidia CLAIMS that people should play @... you will notice that the 9700pro takes a bigger increase than the 20-30% that the drivers are expcected to increase performance...

furthermore... the card has been out for a while I would say... where has the driver team been all this time?

blowing smoke-screens all over the place cannot hide the fact that this product has not lived up to its nvidia-distributed hype.. nor even to the standards you had proclaimed for it in the weeks leading up to the benchmark releases...

[Hellbinder]

As a side note, I think far too many people are putting the FX in much too negative of a light. I really think that the high-end FX beats the 9700 Pro in performance, while the non-Ultra is very slightly slower. Though the FSAA is certainly not as good, the FX has other benefits, such as more comprehensive-quality anisotropic filtering.


1. I completely disagree with your assesment of Aniso Quality.

2. I MORE that completely disagree with your assesment of performance.

If you set both cards to Equal IQ levels the LEAST ATi wins by is around 30 FPS. And That is not talking using performance Ansio. Which i admit is great for some games, and horrible for others.

Once you factor in the average overclock capable on the 9700pro WITHOUT risking The downclocking which a similar overclock would create on the GFFX. Such as my 351/344. Then there really is no contest in about 80% of the benchmarks (i think i am being generous). I gain an average between 13-16 FPS over my defaut clock. It is especially noticable in FSAA situations.

Yes you could bring up exotic cooling solutions etc.. but i am talking joe user with some skillz droping either card in their system.

[Shadowx]

quote:
--------------------------------------------------------------------------------

As a side note, I think far too many people are putting the FX in much too negative of a light. I really think that the high-end FX beats the 9700 Pro in performance, while the non-Ultra is very slightly slower. Though the FSAA is certainly not as good, the FX has other benefits, such as more comprehensive-quality anisotropic filtering.

--------------------------------------------------------------------------------


The non ultra is not slightly slower it will be 10 to 20% slower and the benefits of GFFX "more programable shades" are in a hardwear that does not have the power to used them in a practicall way"FPS"