The Inquirer reports clocks for R520XT

[ddogg]

Originally posted by: blckgrffn
I still think that 16x2 would be cool and I can't think of what else "Extreme" pipes might mean...

they said 16 extreme pipes for the X800XT/PE as well so i really wonder what they're upto now....it might be a 16x2!!

 

[nRollo]

Originally posted by: Ackmed
ATi did best NV, with the 8500. After their very crappy drivers were sorted out 92 months), it was a better card than the GF2, and the GF3. A GF3 Ti series was generally faster however.

Wow- you mean the 8500 actually beat the GF2 after launching a year and a half later!?!?! That must have been tough.

Of course it launched over four months after the GF3 as well, so it's competition was the Ti series which you already admit was faster.

So how is that a "win"???
:roll:

 

[fstime]

Rollo, take your own advice.

"The 5800U is ancient hardware, NO ONE CARES"

Guess what, no one cares about your GF2, TNT, GF3, or GF4.

Stop telling us how many times Nvidia has bested ATI because its "ancient hardware"

You seem to bring this up in every thread.

 

[Cookie Monster]

Originally posted by: BouZouki
Rollo, take your own advice.

"The 5800U is ancient hardware, NO ONE CARES"

Guess what, no one cares about your GF2, TNT, GF3, or GF4.

Stop telling us how many times Nvidia has bested ATI because its "ancient hardware"

You seem to bring this up in every thread.

Technically ackmed did. Hes only replying to Ackmed's post. (fanboys trying to convince each other that ATi/Nvidia is the better company)
BouZouki take a chill pill and dont start another flame war, or even worse directing petty personal attacks at forum users.

 

[BFG10K]

!!! Is that enough to beat the GTX?

That really depends on the definition of a pipe.

Hmmm.... maybe a 16x2 architecture? That would be pretty damn cool for some massive multitexturing

These days a "x2" architecture is a just a waste of transistors.

I'm sure there are good reasons why all the modern cards are using the x1 archtecture.

The most notable reason being that traditional multi-texturing is dead since the shader units just emulate it when required. Of course if you've got shader units you may as well do nifty things that multi-texturing can't dream of doing.

Of course it launched over four months after the GF3 as well, so it's competition was the Ti series which you already admit was faster.

Later drivers put the 8500 on par with a Ti4200 and these days it's even more interesting because of SM 1.4. BF2 for example won't even run on NV2x hardware and Chronicles of Riddick can render its lighting with one pass on a 8500 using 1.4 but it takes four passes on NV2x hardware.

The 8500 is a classic example of great hardware being held back by drivers when it launched, just like the original Radeon before it. When ATi got the drivers and hardware right they produced the R300.

 

[Munky]

Originally posted by: bunnyfubbles

Originally posted by: ZobarStyl

Originally posted by: bunnyfubbles
The one thing that seems strange is that they went through so many tape outs just to get to 600MHz?

This is exactly why I don't think R520 is capable of really blowing 7800GTX out of the water. If they got slightly lower yields/speeds (say, 500-550 mhz) and could match G70 performance on the first try, they would have released it and saved the heavy stuff for the refresh. You don't go through 3 tapes outs and come out with your product this late if you want to win, you do it if you were that far behind and need more mhz just to catch up.

But why would it be so hard to get to 600MHz when it seems like ATI could have pulled it off even at the much larger 130nm process? They were only 60MHz away with their X850XTPE...Then there's all the talk of 16 pipes, and the die size being the same despite a shrink from 130nm to 90nm...I'd wager that something else is on that chip that should give it an advtantage. From all this rumor and speculation it doesn't sound like it will blow anything out of the water, but it just seems unlikely that ATI would have spent so much effort without the delay paying off.

It's true that nVidia did it in the past, but thats all the more reason ATI shouldn't fail, because there is history to learn from there. If you're going to have a part that is slower why delay the inevitable? Release it sooner and just suffer the 2nd place. At least you could try and market hardware H.264...

I dunno, like I've said we'll have to wait and see, however I agree that all the talks of specs without any much if any speculation on performance doesn't look good for ATI. Those who purchased a 7800 card instead of waiting (because they couldn't) haven't made a bad decision if spending that much money on a video card isn't easy for them (the primary source of fanboys, investment they cannot truly afford)

However once we get the actual reviews it does seem (atm) as if this will be another round win for nVidia.

That's what sounds weird to me also - if they got 540 mhz on a 130nm with 16 pipes, how hard can it be to get 600mhz on 90nm with 16 pipes? There's gotta be more to this than we're suspecting, especially after a 3rd tape out.

 

[xtknight]

Problems with the process-that's why. Heat leakage, power leakage, etc. The same issues Intel had. AMD saved themselves by using SOI, I think. I'm sure there's more problems associated with migrating to a different process than we know about.

 

[Cookie Monster]

You people are forgetting something. More complex GPU = Harder to maintain high clockspeeds.

 

[bunnyfubbles]

Originally posted by: Cookie Monster
You people are forgetting something. More complex GPU = Harder to maintain high clockspeeds.

That?s the thing, all we think we know right now is that it is 16 pipes and 600MHz (for the XT no less), but we do not know how much more complex it is. Right now putting two and two together doesn't look good for ATI based on the only little and seemingly irrelevant ?information? we've been ?given?; irrelevant if it turns out to be faster, because 16pipes @ 600MHz doesn?t add up to a performance leader based on what we?re used to. ?More complex? could very well mean more pipes, but that doesn?t seem to be the case, and ?extreme? pipes? If it is only 600MHz, I doubt the ?extreme? pipes are extreme because of the clock speed, because a clock of 600 isn?t that much more impressive than 540...

 

[lein]

But the 520 and 420 are different architectures to begin with. You cant compare a 540MHz x850 to lets say a 540MHz R520. Right now, I'm not counting on ATI being better than nvidia; I just want them to still be competitive in some way...

 

[DRavisher]

Originally posted by: BouZouki
Rollo, take your own advice.

"The 5800U is ancient hardware, NO ONE CARES"

Guess what, no one cares about your GF2, TNT, GF3, or GF4.

Stop telling us how many times Nvidia has bested ATI because its "ancient hardware"

You seem to bring this up in every thread.

The reason he brings it up is that ATi fanbois seem to think that ATi are teh 1337 h4x0rz!! of graphics, when they have only had one successful architecture. People make it seem like ATi have a long history of besting Nvidia, when they have done it 1 time (the 8500 was more like being kind of close to nvidia, and don't you dare come with the argument of it supporting better shader model, LOL).

 

[imported_Ged]

Originally posted by: munky

That's what sounds weird to me also - if they got 540 mhz on a 130nm with 16 pipes, how hard can it be to get 600mhz on 90nm with 16 pipes? There's gotta be more to this than we're suspecting, especially after a 3rd tape out.

SM2.0 vs. SM3.0 hardware. SM3.0 adds a lot of complexity. Someone had a quote from ATI describing how many transistors the move from SM2.0 to SM3.0 would take. It was a significant change. Not only that, ATI is moving from 24bit to FP32bit precision if I recall correctly.

Other than that, I think ATI is going to do something nifty with "scheduling" in order to increase sustained throughputs. If that's the case, that adds complexity as well.

There's obviously a reason for the lower than expected clocks though.

 

[imported_Ged]

The reason he brings it up is that ATi fanbois seem to think that ATi are teh 1337 h4x0rz!! of graphics

Actually, ATI fanbois/grlz think that NVIDIA is the "teh 1337 h4x0rz!! of graphics" due to their ()/>7i/\/\iZ47i0/\/s (optimizations).

 

[blckgrffn]

Thanks for the x2 explanation, BFG

And to back up the statements of the 8500, in my experience about 3 months after launch or so when the Cat's started hitting the street it was no longer anywhere near slower than a GF3ti and really started competing with the Ti4200 about another 3 months later. My 8500 rocked, too bad it only had 64 megs of ram or I would probably have used it another 6-12 months. It at least lasted a year for me And I bought it about a month or two after launch, when I still had to install different drivers for different games. :roll: Can't imagine why it wasn't more of an immeadiate hit I actually bought it because it was the same price at Best Buy as a GF4 MX that was a fresh new release... what a good decision that was! At least I don't shop at BB anymore.

Just as long as it is in the same range as the 7800 series and they actually have some cards soon...

 

[IceMole]

nm

 

[swatX]

Originally posted by: apoppin
Guaranteed the r520 will the the top AGP choice - at least for awhile

i am sure many gamers are in my position and won't be artificially "forced" into PCIe prematurely. . . .

ati wont be releasing the AGP part for quite a while

 

[ElFenix]

Originally posted by: Ged
There's obviously a reason for the lower than expected clocks though.

.09 hasn't turned up roses for anyone yet, as far as clocks are concerned. AMD has been treading water with clocks, intel has barely gotten anything addition with preshot, etc.

 

[apoppin]

Originally posted by: swatX

Originally posted by: apoppin
Guaranteed the r520 will the the top AGP choice - at least for awhile

i am sure many gamers are in my position and won't be artificially "forced" into PCIe prematurely. . . .

ati wont be releasing the AGP part for quite a while

you mean nVidia won't be releasing their AGP part for quite awhile




and i'm in no hurry - no current game chokes my 9800xt - at my resolutions of 10x7 - yet

 

[mwmorph]

Originally posted by: apoppin

Originally posted by: swatX

Originally posted by: apoppin
Guaranteed the r520 will the the top AGP choice - at least for awhile

i am sure many gamers are in my position and won't be artificially "forced" into PCIe prematurely. . . .

ati wont be releasing the AGP part for quite a while

you mean nVidia won't be releasing their AGP part for quite awhile




and i'm in no hurry - no current game chokes my 9800xt - at my resolutions of 10x7 - yet

just wait till you play fear. the demo kills my oced 9800pro at 800x600 med-high.

 

[apoppin]

Originally posted by: mwmorph

Originally posted by: apoppin

Originally posted by: swatX

Originally posted by: apoppin
Guaranteed the r520 will the the top AGP choice - at least for awhile

i am sure many gamers are in my position and won't be artificially "forced" into PCIe prematurely. . . .

ati wont be releasing the AGP part for quite a while

you mean nVidia won't be releasing their AGP part for quite awhile




and i'm in no hurry - no current game chokes my 9800xt - at my resolutions of 10x7 - yet

just wait till you play fear. the demo kills my oced 9800pro at 800x600 med-high.

i said CURRENT game

and i played FEAR beta multiplayer demo on my 9800xt - managed playable frame rates at 10x7 with most details on [turn off soft shadows!] and and med settings . . . i think the 256MB vRAM helps the xt over the 9800p [a lot[

 

[mwmorph]

Originally posted by: apoppin

Originally posted by: mwmorph

Originally posted by: apoppin

Originally posted by: swatX

Originally posted by: apoppin
Guaranteed the r520 will the the top AGP choice - at least for awhile

i am sure many gamers are in my position and won't be artificially "forced" into PCIe prematurely. . . .

ati wont be releasing the AGP part for quite a while

you mean nVidia won't be releasing their AGP part for quite awhile




and i'm in no hurry - no current game chokes my 9800xt - at my resolutions of 10x7 - yet

just wait till you play fear. the demo kills my oced 9800pro at 800x600 med-high.

i said CURRENT game

and i played FEAR beta multiplayer demo on my 9800xt - managed playable frame rates at 10x7 with most details on [turn off soft shadows!] and and med settings . . . i think the 256MB vRAM helps the xt over the 9800p [a lot[

great, now i have to break out the fraps to do some benchmarking in fear. shouldnt be all thatm uch a difference as far as i see, my 9800pro is flashed and oced so it should be more on par at 10x7 high/med textures. Probably is a Softshadow issue.

 

[hop1hop2]

maybe ATI was gonna release close to NVIDIA but then they did that "real" launch so ATI's stocking up on cards.

 

[Pete]

Originally posted by: ddogg
they said 16 extreme pipes for the X800XT/PE as well so i really wonder what they're upto now....it might be a 16x2!!

Actually, R420 was rumored to have 8 "extreme" pipes, which turned out to be 16 "regular" (for ATI) ones.

Originally posted by: Ged
SM3.0 adds a lot of complexity. Someone had a quote from ATI describing how many transistors the move from SM2.0 to SM3.0 would take. It was a significant change. Not only that, ATI is moving from 24bit to FP32bit precision if I recall correctly.

Hmmm, I remember Eric Dremers saying FP24 -> FP32 would require 50% more transistors. NV's CEO said SM3 in NV40 cost them 70M transistors in a conference call with market analysts, but that's simply comparing NV's to ATI's public transistor counts ... and their transistors are often said to be counted differently. So I'm not sure how much we can expect either SM3 (and FP blends) or FP32 to boost transistor count, though cumulatively the improvements will surely be significant.

Other than that, I think ATI is going to do something nifty with "scheduling" in order to increase sustained throughputs. If that's the case, that adds complexity as well.

Yep, heard that, too (first with Huddy's "Save the Nanosecond" notes, and recently with rumors concerning an improved memory bus).

Still, after all this gossip and so close to the purported launch, I'm not sure of anything beyond that R520 will support SM3 and FP blends. I'm left to hope for a pleasant surprise.