ATI Radeon HD 4000 Series discussion

[Hornet331]

no, g80 had 32, g92 had 64

sorry but thats not ture, g80 had 64 TMUs and 32 TFUs -> G92 had 64TMUs and 64TFUs, get your facts right.


btw anyone can bench crysis at everything high on 800x600 (4580 scores ~avg. 56)

[ORBR]

http://xfastest.com/viewthread.php?t...&extra=&page=1


And yet, it is the FASTEST single GPU in texture fillrate when compared to the Digit life scores.
(RV770 gets higher than the 64TMU G92s, touche)

57 FPS in Crysis @800x600 Dx10 Vista 64?

why 800x600?
maybe they are testing the cpu?

[ORBR]

(4580 scores ~avg. 56)

4580?

this is a typo's fest?

[Hornet331]

http://xfastest.com/viewthread.php?t...&extra=&page=1


And yet, it is the FASTEST single GPU in texture fillrate when compared to the Digit life scores.
(RV770 gets higher than the 64TMU G92s, touche)

lol you dont really expect that the 4850 has a teratexel fillrate?

[whocaresbg]

see below in the same site ..

The card scores 56 FPS avg. on DX10;800x600 .

If someone could test on a similar rig with HD3870CF/9800GTX or HD3870X2 .

57 FPS in Crysis @800x600 Dx10 Vista 64?

why 800x600?
maybe they are testing the cpu?

benchmark_gpu .

[bowman]

lol you dont really expect that the 4850 has a teratexel fillrate?

Wait for an engineering sample CPU to show 1THz in CPU-Z, and then for someone to post heralding this marvel of engineering.

[Hornet331]

Wait for an engineering sample CPU to show 1THz in CPU-Z, and then for someone to post heralding this marvel of engineering.

hehe

[Genius_pt]

Honestly that crysis benchmark at 800x600 is intriguing me. Why the hell they tested it at that resolution and not at very high setting?

What did they want to show?

[Extelleron]

http://xfastest.com/viewthread.php?t...&extra=&page=1


And yet, it is the FASTEST single GPU in texture fillrate when compared to the Digit life scores.
(RV770 gets higher than the 64TMU G92s, touche)

If those are true scores, then it is likely that RV770 either has 40 TMUs or architectural improvements IMO.

Unless there have been improvements in the Vantage Feature Scores with updated drivers of course.

Looking at Firingsquad's Vantage review: http://www.firingsquad.com/hardware/...ance/page8.asp

That score would give HD 4850 ~1.86x the texture power of HD 3870, when the theoretical jump w/ 32 TMUs @ 625MHz is 1.61x. The jump with 40 TMUs @ 625MHz would be 2.01x, which seems likely given the 1.86x increase in texture fillrate. HD 4850 might have trouble getting all of its theoretical fillrate since it is likely bandwidth limited.

[Macadamia]

Honestly that crysis benchmark at 800x600 is intriguing me. Why the hell they tested it at that resolution and not at very high setting?

What did they want to show?

THEY HAVE THE CARDS AND ARE NOT UNDER NDA. And they will not post the scores because they want to protec- oops.

If those are true scores, then it is likely that RV770 either has 40 TMUs or architectural improvements IMO.

Unless there have been improvements in the Vantage Feature Scores with updated drivers of course.

Looking at Firingsquad's Vantage review: http://www.firingsquad.com/hardware/...ance/page8.asp

That score would give HD 4850 ~1.86x the texture power of HD 3870, when the theoretical jump w/ 32 TMUs @ 625MHz is 1.61x. The jump with 40 TMUs @ 625MHz would be 2.01x, which seems likely given the 1.86x increase in texture fillrate. HD 4850 might have trouble getting all of its theoretical fillrate since it is likely bandwidth limited.

I concur, actually. But even on the safe side, 32 TMUs would be great already. Memory is SERIOUSLY limiting the 4850's performance from my standpoint.

[spoof]

see below in the same site ..

The card scores 56 FPS avg. on DX10;800x600 .

If someone could test on a similar rig with HD3870CF/9800GTX or HD3870X2 .



benchmark_gpu .

8800GTX stock - 33fps AVG with the same settings. I'm not sure, i think is too slow, because at 1680* Very High i get 20fps avg@680/160/1080

[perkam]

8800GTX stock - 33fps AVG with the same settings. I'm not sure, i think is too slow, because at 1680* Very High i get 20fps avg@680/160/1080

That`s normal. Crysis has an uber-ness limiter that halves the frame rate whenever it gets close to 100. Apparently, the story goes that Crysis brings out a new patch whenever computers start getting above 100fps on any settings in the game, to preserve it`s uberness and to drive fanboys crazy over its low unoptimizied framerates for many years to come

Perkam

[clonez]

you got it

only that thse patches are called expansion-packs, that cut frames in half and offer only minimalistic graphical improvements (and crytek gets more money)

[[XC] hipno650]

i will test Crysis at 800x600 on HIGH tonight on my 8800gt and see what happens but i have a feeling it will be similar to that score. if so im not to impressed. hopefully drivers will fix things.

[Extelleron]

HD 4870 Performance vs 9800GTX

HD 4870 is ~30-50% faster @ 1920x1200 4xAA/8xAF

http://img139.imageshack.us/img139/4...ahd4800eg3.jpg

HD 4850 Performance vs 8800GT

HD 4850 is ~25-50% faster @ 1920x1200 4xAA/8xAF

http://img137.imageshack.us/img137/9...bhd4800mz1.jpg

[perkam]

HD 4870 Performance vs 9800GTX

HD 4870 is ~30-50% faster @ 1920x1200 4xAA/8xAF

http://img139.imageshack.us/im...amdslideahd4800eg3.jpg

HD 4850 Performance vs 8800GT

HD 4850 is ~25-50% faster @ 1920x1200 4xAA/8xAF

http://img137.imageshack.us/im...amdslidebhd4800mz1.jpg

Fake...unless you were trying to tell us they got 404fps in Crysis in an indirect manner that is X_X

Perkam

[Extelleron]

Fake...unless you were trying to tell us they got 404fps in Crysis in an indirect manner that is X_X

Perkam

Fixed links.

[v_rr]

Official Test Results of ATi Radeon HD 4800 Series




http://forum.donanimhaber.com/m_23897423/tm.htm

[Atal]

http://xfastest.com/viewthread.php?t...&extra=&page=1


And yet, it is the FASTEST single GPU in texture fillrate when compared to the Digit life scores.
(RV770 gets higher than the 64TMU G92s, touche)

the bottom right picture at the first post did it for me

[Stuen4y]

If those results aren't a fake nvidia will have serious problems with the price/performance relation. Only the high-end products wouldn't produce enough sales and the cards of ati will be the best choice in their price gap...

[Jowy Atreides]

reason no.1 as to why those graphs are fake:

they're an opposite version of the gtx vs 3870x2 graphs from nvidia
soooo similar in how they use scale and compare previous gen

you really think AMD just copied a leaked nvidia chart?

[Isaac MM]

Will there be any difference in performance between a 4870 in CrossFire and a 4870x2?

[Jowy Atreides]

[Stuen4y]

reason no.1 as to why those graphs are fake:

they're an opposite version of the gtx vs 3870x2 graphs from nvidia
soooo similar in how they use scale and compare previous gen

you really think AMD just copied a leaked nvidia chart?

I don't think so. The nvidia graphics were with about 100 and more percentage difference between their and the ati product, but these look more realistic. I've said they can be fake, but there is still hope in every fanATIc.

P.S. Maybe you're right about the total fake They just have pasted a 4870 pic in the other slide, in order to hide the spot.

Will there be any difference in performance between a 4870 in CrossFire and a 4870x2?

Yes there should be, because of the shared memory that 4870X2 have to use.

[Jowy Atreides]

I don't think so. The nvidia graphics were with about 100 and more percentage difference between their and the ati product, but these look more realistic. I've said they can be fake, but there is still hope in every fanATIc.

P.S. Maybe you're right about the total fake

totally faked, and badly