The ATI Radeon 5XXX Thread

[largon]

^No point in taking random forum posts as gospel.

[Jowy Atreides]

Just read 2160 tflops here:http://translate.google.com/translat...istory_state0=
2160 Tflops would be 675mhz core. I hope this is the 5850.

2160 teraflops?


....


....

[Chad Boga]

I really do hope AMD releases Havok sometime with the HD5000 series... It will totally slaugther Nvidias closed propriety PhysX

AMD can't release Havok because Havok is an independent company, bought by Intel a little time ago.

[FischOderAal]

Just read 2160 tflops here:http://translate.google.com/translat...istory_state0=
2160 Tflops would be 675mhz core. I hope this is the 5850.

This is just a forum member telling what he has read somewhere. He has no inside knowledge whatsoever

[Jowy Atreides]

This is just a forum member telling what he has read somewhere. He has no inside knowledge whatsoever

Yeah, the fact he thinks any modern card has 2160 Tflops is fook'in hilarious.

That's 2160x more than a 4850

Quite a jump in performance, eh

[largon]

AMD can't release Havok because Havok is an independent company, bought by Intel a little time ago.

*cough*
Maybe, just maybe he meant:

I really do hope AMD releases support for Havok GPU physics sometime with the HD5000 series

[migunani]

Just read 2160 tflops here:http://translate.google.com/translat...istory_state0=
2160 Tflops would be 675mhz core. I hope this is the 5850.

i cant believe you quote that. That just makes you look bad.

[Farinorco]

*cough*
Maybe, just maybe he meant:

Then, he was equally wrong. Because AMD don't have to support Havok GPU physics either.

AMD has to support OpenCL, DirectX11 Compute Shaders, and/or whichever GPGPU APIs they want.

Then, software developers can make software built over those APIs/libraries (OpenCL, DX11 CS, CUDA...) and any device compatible with those APIs will be compatible with the software, in the same way than graphical software built over OpenGL and Direct3D.

AMD has nothing to release about Havok. Havok (the company who developes the library, and which is owned by Intel) are who have to release "OpenCL hardware acceleration". And then, it will be compatible with every single OpenCL capable GPU, be it from NVIDIA, AMD or Intel.

[Chumbucket843]

*cough*
Maybe, just maybe he meant:

that project has been rumoured to have been cancelled. i guess it was the consoles fault.

[largon]

that project has been rumoured to have been cancelled. i guess it was the consoles fault.

Really? Considering Havok on OpenCL was demonstrated just a few months ago...

[FischOderAal]

ATI's older attempt of GPU-physics using Havok was cancelled, OpenCL is the new attempt

Old one:
http://ati.amd.com/technology/crossf...ics/index.html

[Farinorco]

that project has been rumoured to have been cancelled. i guess it was the consoles fault.

I don't think so. As Iargon says, there were demostrations of Havok running over OpenCL which included videos not long ago. You can even watch a demo video of OpenCL version of Havok Cloths with some red dressed dancers on YouTube. And there is an official announcement saying which are going to be the first Havok physics parts which are going to support GPGPU hw acceleration through OpenCL (Havok Cloths and Havok Destruction, both non-free to license).

Maybe you're talking about the former attempts to accelerate Havok with shaders in a GPU which resulted in a first physics GPU accelerated library called Havok FX, which was cancelled in the end...

EDIT: FischOderAal answered before...

[Origin_Unknown]

with the info date being tomorrow i'm getting excited! lets hope its not another 'black' situation (i think) where there was a launch date and only a picture on an amd web site (at least i think)

[Farinorco]

I don't know if we'll see any new info tomorrow. The tomorrow US presentation for press is, AFAIK, under NDA as it was the yesterday european one... I hoped for some leaks yesterday, but nothing, I suppose AMD is going all for the "top secret" route this time... and now I don't see any reason to think that tomorrow is going to be any different. There's people talking about 16th, some others 22-23th to the NDA expiring date...

Damn, I hope we get some leaks before that date, but it's not like if it was happening, from the looks of it...

[largon]

Yep, I'm not expecting any hard numbers on technical specs let alone performance from Sep 10th event. Likely it's just the usual PR BS with DX11 demos and some general yapping about the new features.

[FischOderAal]

Yep, I'm not expecting any hard numbers on technical specs let alone performance from Sep 10th event. Likely it's just the usual PR BS with DX11 demos and some general yapping about the new features.

Don't make me cry largon I was hoping for names, price, marketing jibberish specs (shaders, whopping MHz etc) and of course: PICS!

[Calmatory]

Aren't they handing out some pieces for the reviewers then?

Ohwell, they're under NDA and it's lifting in like two weeks.

[labs23]

Don't make me cry largon I was hoping for names, price, marketing jibberish specs (shaders, whopping MHz etc) and of course: PICS!

+1. I want to know some concrete specs. too.
No problem on the price, I'm already convinced at that.

[spursindonesia]

So, quite possibly, 1600 SP, 80 TMU, 32 ROP, 256 bit GDDR5 for Cypress ?

AMD HD 5850/HD 5870: 1600 SPs and 80 TMUs

[jfromeo]

Just 1 day left to clear up any doubts

[Origin_Unknown]

So, quite possibly, 1600 SP, 80 TMU, 32 ROP, 256 bit GDDR5 for Cypress ?

AMD HD 5850/HD 5870: 1600 SPs and 80 TMUs

release this month? really?

[Chrono Detector]

I like the shader count but the 256 bit memory really annoys me for some reason.

[spursindonesia]

Just 1 day left to clear up any doubts

Yeah, this silence is deafening.

release this month? really?

Perhaps if the available quantity is good, there's a chance for a reoccurence of HD 4850 prerelease wide availability in the market ??

I like the shader count but the 256 bit memory really annoys me for some reason.

Yupz, me too, though perhaps ATi's engineers have developed a better data compression technique or using faster mem chips. RV740 shows that the mArch isn't all that bandwith starved as predicted, more of mem capacity limited in huge rez.

[Calmatory]

Any info about the everlasting rumor of MCM packaged X2? Wasn't there a picture of the heatsink which had two thermal pads?

Edit: Also any info about the "new cooling technique"? The one which uses a vapor chamber as it's base. IIRC it was developed with some cooling company.

[labs23]

Perhaps if the available quantity is good, there's a chance for a reoccurence of HD 4850 prerelease wide availability in the market ??

It could be very well possible, and we should be seeing the cards before Oct.
We should know in the next 24hrs., maybe.