AMD RV770 refresh -> RV790

[strike]Performance-wise RV770 is 2.36 times RV670XT while GTX280 is 1.84 RV670XT. Power-wise I have no clue, sorry.[/strike]
Sorry, I misunderstood Jawed here. He was talking about Pixel-Shader-Test, I was referring to Perlin Noise from 3DMark 06.
 
Last edited by a moderator:
Performance-wise RV770 is 2.36 times RV670XT while GTX280 is 1.84 RV670XT. Power-wise I have no clue, sorry.
Hang on, you're saying that the Pixel Shader test, which is heavy on texturing, is running faster on HD4870 than GTX280? That's a surprise.

I wonder if GTX280 is throttling it due to temperatures? Is GTX280 much faster than 9800GTX?

I wonder if Fallacy will try this and the 3DMk01 test that no-X referred to...

Jawed
 
Hang on, you're saying that the Pixel Shader test, which is heavy on texturing, is running faster on HD4870 than GTX280? That's a surprise.

I wonder if GTX280 is throttling it due to temperatures? Is GTX280 much faster than 9800GTX?

I wonder if Fallacy will try this and the 3DMk01 test that no-X referred to...

Jawed
Nope - sorry. I was referring to your numbers and links posted, which (i mistakenly thought) indicated you were talking about "Perlin Noise" from 3DMark06.

"Pixel Shader" actually runs at 2.92 times RV670XT on GTX280 and 2.27 on RV770XT.
 
Last edited by a moderator:
9800GTX+ says hi. ;)

Also waving a red flag at me are these pictures on chiphell. A dual cooler for the 5850 on a really long board.

Surely they realise this part will spend most of its life well below US$150. Does also give credence to the rumors of additional shaders ie 960. Thankfully only one 6 pin though, so not everyone has gone insane ;)

According to other threads apparently samples are with AIBs now, so leaks should start soon en masse.
 
Also waving a red flag at me are these pictures on chiphell. A dual cooler for the 5850 on a really long board.

Surely they realise this part will spend most of its life well below US$150. Does also give credence to the rumors of additional shaders ie 960. Thankfully only one 6 pin though, so not everyone has gone insane ;)

According to other threads apparently samples are with AIBs now, so leaks should start soon en masse.
Isnt that the recently released FirePro?
 
Ha ha, kindof a relief it is fake. Funny how much higher quality the images were than the genuine leaks Next time that guy should put way less effort in so that the result is more believable...

Guess in real life probably try to reuse the board and cooler design of the 4800 series as much as possible to save costs.
 
Chiphell admin does seem to have a sample. Or at least we're in the progress of coaxing him to. :D


But he did legitly mention on RV790XT replacing the 300USD slot. I don't think there is going to be an X2 this time too, just one (slightly) bigger chip serving both mainstream (Pro) and enthusiast (XT).
 
I wonder if Fallacy will try this and the 3DMk01 test that no-X referred to...

Jawed

Sorry for the late answer, i´m currently a little bit busy ;).

I´ve done some tests with 3DMark01 an 06 on a GeForce GTX 260 (65nm). And yes 3DMark01 is really power demanding and gets close to the results of FurMark.

Furmark: 166 Watt
3DMark01 (Mult-texturing-tests): 150 Watt
3DMark06 (Batch-Tests): 125 Watt

Hope i´ll have a little bit more time for testings in a few days, so that i can keep an eye on some other applications.
 
FurMark is a real video card nuker. A friend of mine got an old 2900XT attached to a pretty solid water cooling rig, with overvolted and overclocked GPU and FurMark virtually boiled the water in the loop like never before, but the thing held pretty well all the torture.
 
If it's still a single 6 pin connector with 1 gig of memory, I may get one to replace my 4870 with 512 meg. While it violates my general rule of thumb to not buy a graphics card more often than once a year, I've been running into the occasional wall with AA/AF and 512 megs.

And I REALLY can't justify replacing a 4870 with another 4870 just to get more memory. :)

Regards,
SB
 
VR-Zone got a hold of the same info I did about the HD4890. I'm expecting more info to leak soon (today or in the next couple of days) cause this wasn't all the info they were given if I'm correct.
 
13% on the core and 8% on the memory is not going to make this competitive with GTX285. Not unless they double the number of Zs per RBE - which is clearly not going to happen.

Jawed
 
Yup, but it didn't say anything about the number of simd's on it. Probably, there's where the punch lies.

I'd be really disappointed if they launched a new chip on 40 nm w/o any increase in shader count. IIRC, rv7xx was pad limited so it would be wasting die space if it wasn't increased.
 
Back
Top