SUBMAN1
10-02-07, 11:22 AM
I told you that I was skipping the 8800 line and the x2900 line. 250 Watts of power draw is just crazy!
Anyway, the test boards from ATI for the new RV670 is drawing only half that - about similar to a x1900 XTX again:
-S
RV670 clocks are not yet defined
And at the third stroke it'll be 800MHz, or thereabouts
By Theo Valich (http://www.subsim.com/articles/flameAuthor/gb/inquirer/news/2007/10/02/rv670-clocks-defined): Tuesday, 02 October 2007, 11:00 AM
AS WE PREVIOUSLY disclosed in this article, (http://www.theinquirer.net/gb/inquirer/news/2007/10/01/ati-rv670-beat-r600) RV670 is a vastly improved R600 chip. And if you are wondering how a chip with 256-bit memory controller can outdo a 512-bit one, the answer can be found by looking at the clocks.
As it stands right now, neither the clocks not the TDP have been finalised, and this will not happen for the next couple of weeks, until the final round of speed binning completes and ATI issues a final production BIOS for the card.
However, partners now have physical prototype cards that work on different clocks, and we can now say that from one side, we have RV670XT, followed by cheaper RV670Pro.
Prototype board with RV670XT markings, that was leaked by several vendors sported a TDP of 132 Watts, and had GPU clocked all the way to 825 MHz, with GDDR4 memory working at 1.2 GHz (2.4 GT/s), while RV670 Pro prototype came with GPU clocked at 750 MHz, and GDDR3 memory was working at 900 MHz (1.8 GT/s). This board has TDP of around 104W.
As you know, original Radeon HD 2900XT had GPU clocked at 743MHz, with GDDR3 memory clocked at 826MHz. Now, even the RV670Pro has higher GPU and memory clock, so you can imagine the performance difference. We have to remind you that specs are not final, and the clocks of Gemini board (two RV670 GPUs on single PCB) and final clocks of XT/Pro are not set in stone. Same applies for overclocked version of RV670 with 2-slot cooling. GPU clock of 900MHz or 940MHz should not be unreachable, and when put under water, you should expect quite stellar results.
Both boards are PCI Express 2.0, but in order to keep the compatibility with 1.0 and 1.1 boards, both boards carry 6-pin PEG connector. There have been some issues with PCIe 2.0 boards when plugged in 1.1 motherboards (read: did not work), so Daamit is playing it safe and RV670 does not surpass the 75W draw from the PCIe slot - as required by 1st gen PCIe x16 electrical spec.
Several readers came back to us after the original article and asked about video capabilities. Unlike R600 and the whole soggy affair about R600 has UVD on several marketing slides (in fact, it did not), RV670 had enough room on a wafer to put full video encoding/decoding engine known to the world as UVD or Universal Video Decode.
We hope we managed to shed some light in this situation. Nevertheless, our standard disclosure applies: RV670 is a living and breathing project, with nothing being set in stone. Yet. µ
http://www.theinquirer.net/gb/inquirer/news/2007/10/02/rv670-clocks-defined
Anyway, the test boards from ATI for the new RV670 is drawing only half that - about similar to a x1900 XTX again:
-S
RV670 clocks are not yet defined
And at the third stroke it'll be 800MHz, or thereabouts
By Theo Valich (http://www.subsim.com/articles/flameAuthor/gb/inquirer/news/2007/10/02/rv670-clocks-defined): Tuesday, 02 October 2007, 11:00 AM
AS WE PREVIOUSLY disclosed in this article, (http://www.theinquirer.net/gb/inquirer/news/2007/10/01/ati-rv670-beat-r600) RV670 is a vastly improved R600 chip. And if you are wondering how a chip with 256-bit memory controller can outdo a 512-bit one, the answer can be found by looking at the clocks.
As it stands right now, neither the clocks not the TDP have been finalised, and this will not happen for the next couple of weeks, until the final round of speed binning completes and ATI issues a final production BIOS for the card.
However, partners now have physical prototype cards that work on different clocks, and we can now say that from one side, we have RV670XT, followed by cheaper RV670Pro.
Prototype board with RV670XT markings, that was leaked by several vendors sported a TDP of 132 Watts, and had GPU clocked all the way to 825 MHz, with GDDR4 memory working at 1.2 GHz (2.4 GT/s), while RV670 Pro prototype came with GPU clocked at 750 MHz, and GDDR3 memory was working at 900 MHz (1.8 GT/s). This board has TDP of around 104W.
As you know, original Radeon HD 2900XT had GPU clocked at 743MHz, with GDDR3 memory clocked at 826MHz. Now, even the RV670Pro has higher GPU and memory clock, so you can imagine the performance difference. We have to remind you that specs are not final, and the clocks of Gemini board (two RV670 GPUs on single PCB) and final clocks of XT/Pro are not set in stone. Same applies for overclocked version of RV670 with 2-slot cooling. GPU clock of 900MHz or 940MHz should not be unreachable, and when put under water, you should expect quite stellar results.
Both boards are PCI Express 2.0, but in order to keep the compatibility with 1.0 and 1.1 boards, both boards carry 6-pin PEG connector. There have been some issues with PCIe 2.0 boards when plugged in 1.1 motherboards (read: did not work), so Daamit is playing it safe and RV670 does not surpass the 75W draw from the PCIe slot - as required by 1st gen PCIe x16 electrical spec.
Several readers came back to us after the original article and asked about video capabilities. Unlike R600 and the whole soggy affair about R600 has UVD on several marketing slides (in fact, it did not), RV670 had enough room on a wafer to put full video encoding/decoding engine known to the world as UVD or Universal Video Decode.
We hope we managed to shed some light in this situation. Nevertheless, our standard disclosure applies: RV670 is a living and breathing project, with nothing being set in stone. Yet. µ
http://www.theinquirer.net/gb/inquirer/news/2007/10/02/rv670-clocks-defined