R420 and R423 details from HardOCP

[Sabastian]

Well if nvidia can run a .13 micron part with 222million transistors @ 400 mhz ATi ought to be able to get a 160-180 million transistor part to run at 500mhz on the .13 micron process. Surely ATi was able to manage 380mhz with a 110 million transistor chip on the .15 micron process. I think though that the 16 pipeline variety of the R420 will be over 200 million transistors. ATi has a history of making cooler running processors compared to nVidia at the same process and relatively same transistor count but in this case they might be running into some limitations and implement similar cooling a la nVidia, maybe.

 

[Unknown Soldier]

So if this is correct and the R420 does comeout with 16 pipes and 600 Mhz Clock with 1200 Mhz GDDR3, how much memory bandwidth are we looking at??

I know the NV40 with 16 pipes and 400Mhz Clock with 550 Mhz GDDR3 gives a memory Bandwidth of 35.2GB/sec(as per Dave's preview).

The "1200 Mhz" cited by HardOCP (if true), is actually 600 MHz GDDR3, or 50 Mhz more than the 550 on the 6800 Ultra.

600 Mhz GDDR3 on a 256 bit wide bus would yield 38.4 GB/sec.

Oops ye .. I was meant to change that and forgot. 1200Mhz to 600 Mhz I mean.

US

 

[Evildeus]

Did ATI have a shader day event yesterday? Anyone reporting on it??

I think it's under NDA :?

 

[Tim]

Will this require me to get a new power supply (400watt right now) cause they don't make silent power supplys any bigger than this and i really don't want a crap load of noise in my silent case .

The Tagan 480W is quiet (and very good).

 

[jvd]

Will this require me to get a new power supply (400watt right now) cause they don't make silent power supplys any bigger than this and i really don't want a crap load of noise in my silent case .

The Tagan 480W is quiet (and very good).

yea but how much is it , 4 months ago i got the ion power supply and it cost me 80$ not ready to add another 100$ to my system for a video card. You know what i mean ?

 

[Richteralan]

Will this require me to get a new power supply (400watt right now) cause they don't make silent power supplys any bigger than this and i really don't want a crap load of noise in my silent case .

The Tagan 480W is quiet (and very good).

yea but how much is it , 4 months ago i got the ion power supply and it cost me 80$ not ready to add another 100$ to my system for a video card. You know what i mean ?

I got mine Tagan 480W for like 88.99USD free shipping.

 

[Pete]

I'd like to know how many people here think running a 160-180M chip at 600MHz is possible.

nVidia did back in January, according to an NV40 slide. I wonder if low-k is enough to put ATi up to 500-600MHz.

 

[jvd]

Will this require me to get a new power supply (400watt right now) cause they don't make silent power supplys any bigger than this and i really don't want a crap load of noise in my silent case .

The Tagan 480W is quiet (and very good).

yea but how much is it , 4 months ago i got the ion power supply and it cost me 80$ not ready to add another 100$ to my system for a video card. You know what i mean ?

I got mine Tagan 480W for like 88.99USD free shipping.

which means in a 5 month time period i would have spent 170$ on the same part .

Mabye the non pro wont need a 480watt supply .

 

[Bjorn]

Mabye the non pro wont need a 480watt supply .

From what it seems, the NV40 doesn't really need one either. Some people have been using 350W PSU's and that seemed to work fine. And some site had measured the power usage and it wasn't that much more then a 9800 XT if i remember it correct.

I think Nvidia is a bit conservative with this recomendation, maybe because of upcoming parts.

 

[jvd]

Mabye the non pro wont need a 480watt supply .

From what it seems, the NV40 doesn't really need one either. Some people have been using 350W PSU's and that seemed to work fine. And some site had measured the power usage and it wasn't that much more then a 9800 XT if i remember it correct.

I think Nvidia is a bit conservative with this recomendation, maybe because of upcoming parts.

mabye not but i have alot of stuff in my pc running off my powersupply including my water pump.

So i don't really want to start smelling something and finding out my 400$ new vide ocard died.

 

[PeterAce]

Well if nvidia can run a .13 micron part with 222million transistors @ 400 mhz ATi ought to be able to get a 160-180 million transistor part to run at 500mhz on the .13 micron process. Surely ATi was able to manage 380mhz with a 110 million transistor chip on the .15 micron process. I think though that the 16 pipeline variety of the R420 will be over 200 million transistors. ATi has a history of making cooler running processors compared to nVidia at the same process and relatively same transistor count but in this case they might be running into some limitations and implement similar cooling a la nVidia, maybe.

ATI manged better then than that, the R360 (9800XT) is qualified at 432 Mhz @ .15 micron process (412MHz default clockspeed because of OEM requirements).

 

[anthrax]

Well if nvidia can run a .13 micron part with 222million transistors @ 400 mhz ATi ought to be able to get a 160-180 million transistor part to run at 500mhz on the .13 micron process. Surely ATi was able to manage 380mhz with a 110 million transistor chip on the .15 micron process. I think though that the 16 pipeline variety of the R420 will be over 200 million transistors. ATi has a history of making cooler running processors compared to nVidia at the same process and relatively same transistor count but in this case they might be running into some limitations and implement similar cooling a la nVidia, maybe.

I wouldn't be very surprized at 180 million transitors running at 600 Mhz........Chip technology from a year at IBM could pack mulitple cores to make a 3.2 billion transitor chip running at 1 GHz.....

link http://www.redbooks.ibm.com/redbooks/SG246947/css/SG246947_50.html

 

[reever]

Well if nvidia can run a .13 micron part with 222million transistors @ 400 mhz ATi ought to be able to get a 160-180 million transistor part to run at 500mhz on the .13 micron process. Surely ATi was able to manage 380mhz with a 110 million transistor chip on the .15 micron process. I think though that the 16 pipeline variety of the R420 will be over 200 million transistors. ATi has a history of making cooler running processors compared to nVidia at the same process and relatively same transistor count but in this case they might be running into some limitations and implement similar cooling a la nVidia, maybe.

I wouldn't be very surprized at 180 million transitors running at 600 Mhz........Chip technology from a year at IBM could pack mulitple cores to make a 3.2 billion transitor chip running at 1 GHz.....

link http://www.redbooks.ibm.com/redbooks/SG246947/css/SG246947_50.html

That's just a weeee bit different

 

[Stryyder]

I have seen 9600xt's overclocked to 580mhz so I think we should be optimistic rather than pessamistic anout the clocks on the R4XX

 

[Pete]

AT OCing roundup: their four 9600XTs ranged from 530-608(!)MHz, while their five 5700Us ranged from 510-555MHz. More importantly, the 9600XTs OC'ed to a minimum of 530MHz, while the 5700Us seemed to bunch up around 535MHz. A not-insignificant difference, particularly given that the 5700U had a molex connector.

The engineering differences between NV36 and NV40 may be too great compared to RV360 vs. R420, so extrapolating NV40 vs. R420 clocks from NV36 vs. RV360 peak speeds may be a stretch.

 

[Stryyder]

AT OCing roundup: their four 9600XTs ranged from 530-608(!)MHz, while their five 5700Us ranged from 510-555MHz. More importantly, the 9600XTs OC'ed to a minimum of 530MHz, while the 5700Us seemed to bunch up around 535MHz. A not-insignificant difference, particularly given that the 5700U had a molex connector.

The engineering differences between NV36 and NV40 may be too great compared to RV360 vs. R420, so extrapolating NV40 vs. R420 clocks from NV36 vs. RV360 peak speeds may be a stretch.

Anybody know what the transistor count for both cards are (9600xt vs 5700U's) ? ? ?

 

[Kaotik]

AT OCing roundup: their four 9600XTs ranged from 530-608(!)MHz, while their five 5700Us ranged from 510-555MHz. More importantly, the 9600XTs OC'ed to a minimum of 530MHz, while the 5700Us seemed to bunch up around 535MHz. A not-insignificant difference, particularly given that the 5700U had a molex connector.

The engineering differences between NV36 and NV40 may be too great compared to RV360 vs. R420, so extrapolating NV40 vs. R420 clocks from NV36 vs. RV360 peak speeds may be a stretch.

Anybody know what the transistor count for both cards are (9600xt vs 5700U's) ? ? ?

9600XT ~75m 5700U ~82m (afaik)

 

[AlphaWolf]

Well if nvidia can run a .13 micron part with 222million transistors @ 400 mhz ATi ought to be able to get a 160-180 million transistor part to run at 500mhz on the .13 micron process. Surely ATi was able to manage 380mhz with a 110 million transistor chip on the .15 micron process. I think though that the 16 pipeline variety of the R420 will be over 200 million transistors. ATi has a history of making cooler running processors compared to nVidia at the same process and relatively same transistor count but in this case they might be running into some limitations and implement similar cooling a la nVidia, maybe.

ATI managed 432 mhz on a 110 million transistors with the 9800XT.

 

[Stryyder]

Well if nvidia can run a .13 micron part with 222million transistors @ 400 mhz ATi ought to be able to get a 160-180 million transistor part to run at 500mhz on the .13 micron process. Surely ATi was able to manage 380mhz with a 110 million transistor chip on the .15 micron process. I think though that the 16 pipeline variety of the R420 will be over 200 million transistors. ATi has a history of making cooler running processors compared to nVidia at the same process and relatively same transistor count but in this case they might be running into some limitations and implement similar cooling a la nVidia, maybe.

ATI managed 432 mhz on a 110 million transistors with the 9800XT.

The 9800xt isnt on low k

 

[MadMaxtrox]

500mhz Core is currently a given and 600mhz is the target, but yield will be the biggest issue.

Looking at the memory bandwidth, you just have to choose where you want your bottleneck