View Full Version : Official R520. R530, and R515 info at Anandtech
http://www.anandtech.com/video/showdoc.aspx?i=2532
So here is how the line-up goes:
X1800XT 16pipe 600/700 - $599
X1800XL 16pipe 550/625 - $499
X1800 Pro 16pipe 500/500 - $449
X1800LE 12pipe 450/450 - $345
256bit external bus w/ internal 512-bit ring bus
256mb and 512mb config
XT and XL have 2 slot cooling
---
X1600XT 12pipe 600/700
X1600 Pro 12 pipe 500/400
128bit external bus w/ internal 256-bit ring bus
128mb, 256mb, and 512mb config
---
X300 Pro 4pipe 550/500
X300LE 4pipe 450/400
64 and 128bit memory bus
128mb and 256mb config
All cards will support H.264 decoding.
Ninjaman09
09-13-05, 11:08 PM
Cool, thanks for the info! Looking forward to seeing some benchmarks. :D
GlowStick
09-13-05, 11:46 PM
Looks good, cant wait for benches.
Thing is it seems like it's pretty on par to a 7800GTX, except for almost $150 more?
It does have more onboard memory than a 7800gtx though... Bring on the R580!
Roliath
09-14-05, 12:25 AM
I really want to see some benches comparing it the 7800GTX...
I can't freakin wait!
X1800 XT 16 600MHz 700MHz 512MB GDDR3 $599
That's pretty cheap for a 512MB card too, considering thats what 7800GTX's started out at.
Yea, not too bad for 512mb. Too bad they couldn't get 32 pipes out of that bad boy. 32 pipes with those clockspeeds would have been something to see :D
Hey guys do we have any info on the vertex units? I heard 10... dunno if thats true or not.
Thing is it seems like it's pretty on par to a 7800GTX, except for almost $150 more?
The GTX debuted at the same price. If they are available in good quanties then the price should drop just like it did on the GTX.
Performance wise, it's probably gonna win in everything but OGL. 1400mhz w/ the 512bit ring bus should provide a huge supply of bandwidth when compared to the GTX which is pretty bandwidth limited. The card also has a much higher pixel fill than the GTX (6880mp/s vs. 9600mp/s). The performance difference between the 7800GTX and 1800XT is going to be small, thats for sure. I expect the 1800XT to have a slight edge at everything, but 2048 where the GTX's 24 pipeline design should win. Its really too close to call a winner before we see the benchmarks.
EDIT:
Hey guys do we have any info on the vertex units? I heard 10... dunno if thats true or not.
The Anandtech article didn't mention it. The word on the street (rage3d) is 10. That would explain the 1k advantage in 3dmark05 which is very vertex bound.
coldpower27
09-14-05, 12:45 AM
Well the R520 seems to have the edge in output texel fillrate 9.6 GP vs 6.88 GP, more memory bandwidth due to 1.4GHZ memory, and most likely more vertex shader power due to high clocks and at least 8 VS, should be. Hard ot say performance numbers at this point in time. Though it looks competitive on paper.
fivefeet8
09-14-05, 01:09 AM
The card also has a much higher pixel fill than the GTX (6880mp/s vs. 9600mp/s).
7800gtx:
24 pps x 430 mhz = 10320 mp/s
x1800xt:
16 pps x 600 mhz = 9600 mp/s
7800gtx:
24 pps x 430 mhz = 10320 mp/s
x1800xt:
16 pps x 600 = 9600 mp/s
bah, take the KO edition :D 11,760 mp/s :D :D
7800gtx:
24 pps x 430 mhz = 10320 mp/s
x1800xt:
16 pps x 600 mhz = 9600 mp/s
The 7800GTX only has 16 ROP (raster operations) per clock.
16 ROP x 430 = 6880 mp/s
Too bad they couldn't get 32 pipes out of that bad boy. 32 pipes with those clockspeeds would have been something to see :D
No kidding. I wonder how the ATI fansites are reacting to the news. It seems like they'd be upset over such a long wait and then only getting a 16-pipe card. Plus, it doesn't look like it's going to be all that much faster than a card that has already been out for quite a while. I'm sure they'll be excited about finally getting some next gen features but you'd think they'd have to be at least a little disappointed in these specs. I guess we could still be surprised when we see the actual (official) results though.
Edit: 512M is good though. I'd prefer to have 512M cards...just because. :)
sandeep
09-14-05, 02:05 AM
The 7800GTX only has 16 ROP (raster operations) per clock.
16 ROP x 430 = 6880 mp/s
Why would you take ROPS...thats only the output but 24 pixel is whats doing the working. It processes 24 pixel per clock! Your calculations are favouring ATI. ;)
fivefeet8
09-14-05, 03:11 AM
The 7800GTX only has 16 ROP (raster operations) per clock.
16 ROP x 430 = 6880 mp/s
Then how come the Anandtech article states:
"600MHz core clock will give it a lower fill rate than the GeForce 7800 GTX (24-pipes at 430MHz)"
And on the Nvidia website. The Pixel Fillrate of the 7800gtx is listed at 10320.
nutball
09-14-05, 03:31 AM
Then how come the Anandtech article states:
"600MHz core clock will give it a lower fill rate than the GeForce 7800 GTX (24-pipes at 430MHz)"
And on the Nvidia website. The Pixel Fillrate of the 7800gtx is listed at 10320.
Do some research.
fivefeet8
09-14-05, 03:42 AM
Do some research.
Heh.. Ah I see. It's the texel fillrate that is 10320. So the R520 has the same texel and pixel fillrate then. A bit confusing when some review sites don't state whether they are looking at texels or pixels when they say "fill rate".
But then how many ROP's does the R520 have? Wasn't it recently discussed that the G70 could have different internal clocks for it's ROP's too?
Do some research.
There are two fill rates.
#1: One refers to the number of pixels the GPU can output (ROPs x clock speed).
#2: One refers to the number of pixels (EDIT: this called Texels (http://en.wikipedia.org/wiki/Texel_(graphics))?) the GPU can work on (Pixel pipes x clock speed).
More and more #2 is becoming the more important number.
I'll also add that theoretical peaks are nice to throw around for marketing, but the sustained rates and real world performance are where it's at. It is possible, in theory, to have a much lower peak and yet still have the upper hand because you have a more efficient architecture.
fivefeet8
09-14-05, 04:24 AM
(EDIT: this called Texels (http://en.wikipedia.org/wiki/Texel_(graphics))?)
I think beyond3d has a good definition of both:
"Basically you shouldn't worry to much about whether it says texels or pixels. Just keep in mind that pixels usually refers to values written to the frame buffer (that are the pixels shown on your screen, pixel = picture element). Texels refer to the pixels that construct a texture . The word texels comes from texture elements." - Link (http://www.beyond3d.com/articles/fillrate/index4.php)
The GTX debuted at the same price. If they are available in good quanties then the price should drop just like it did on the GTX.
True, but the big difference is that the GTX didn't have any competition when it was released (and still doesn't) and was immediately available in large quantities, while the R520 will have to compete with a cheaper and more than likely higher performing GTX.
True, but the big difference is that the GTX didn't have any competition when it was released (and still doesn't) while the R520 will have to compete with a cheaper, and more than likely higher performing GTX.
Yes, but X1800XT and XL have 512mb memory.
Nvidia wanted to milk the enthusiasts with the 600$ launch price and I guess their 512mb version will cost 600$ too.
There are two fill rates.
#1: One refers to the number of pixels the GPU can output (ROPs x clock speed).
#2: One refers to the number of pixels (EDIT: this called Texels (http://en.wikipedia.org/wiki/Texel_(graphics))?) the GPU can work on (Pixel pipes x clock speed).
More and more #2 is becoming the more important number.
I'll also add that theoretical peaks are nice to throw around for marketing, but the sustained rates and real world performance are where it's at. It is possible, in theory, to have a much lower peak and yet still have the upper hand because you have a more efficient architecture.
I'm not saying that having a high pixel fill is really important. Its just one of the advantages of the X1800XT.
But then how many ROP's does the R520 have? Wasn't it recently discussed that the G70 could have different internal clocks for it's ROP's too?
16 like the 7800GTX. It just runs at a much higher clock speed.
Yes, but X1800XT and XL have 512mb memory.
Nvidia wanted to milk the enthusiasts with the 600$ launch price and I guess their 512mb version will cost 600$ too.
Having a rumoured 512MB memory doesn't mean it'll perform any better, but it certainly is a marketing ploy for the uninformed.
Having a rumoured 512MB memory doesn't mean it'll perform any better, but it certainly is a marketing ploy for the uninformed.
I guess you are more informed. :o
512mb is no rumor and it will give you the chance to play FEAR, BF2 and many other games coming without stuttering or having to buy another gig of system ram.
Don't take my word for it,just search this forum. ;)
JonathanM
09-14-05, 11:49 AM
Wow, the X1800 Pro seems to be the one to avoid. For only 50 bucks more you can get the XL with 50MHz more on the core, 125 MHz faster mem and 256 MB more than the Pro.
IF the products ship at their MSRP (and that's a big if considering what happened last generation) Anyone getting the Pro has definitly not done their research...
vBulletin® v3.7.1, Copyright ©2000-2012, Jelsoft Enterprises Ltd.