ATI Radeon HD 4000 Series discussion

[Afterburner]

Anybody here has a Samsung F8 LCD TV and CF/SLI setup to test and report back....

[Xello]

I'm not sure about this microstuttering thing, so i'll request a Bottom Line here:

If i purchase another 8800GTX for my current system will there be any performance gains or won't there?

Thanks.

[GAR]

I'm not sure about this microstuttering thing, so i'll request a Bottom Line here:

If i purchase another 8800GTX for my current system will there be any performance gains or won't there?

Thanks.

Ofcourse there will be frame rate increase without a doubt, but you might notice a micro stutter effect every now and then.

[cegras]

http://www.vr-zone.com/articles/Rade..._280/5851.html

While we are away for Computex, CJ let us know that the Radeon HD 4870 X2 R700 prototype card is out and it beats the NVIDIA GeForce GTX 280 in 3DMark Vantage. The R700 card is basically made up of two RV770 GPUs with 2x256-bit memory interface to either GDDR3 or GDDR5 memories. We asked our sources about R700 in Computex and apparently AMD is going to let AIB partners to decide the specs themselves. Therefore, the partners will set their own clock speeds, PCB design, memory types, cooler solutions etc. and there will be Radeon HD 4850 X2 and 4870 X2 cards differentiate by the memory type. The R700 card apparently doing pretty well at this stage scoring about X5500 in 3DMark Vantage Extreme preset while the GeForce GTX 280 card is scoring X4800. Both sides are still working hard on optimizing their drivers for the new architecture so probably we will see the performance to improve over time.

[Xello]

Ofcourse there will be frame rate increase without a doubt, but you might notice a micro stutter effect every now and then.

I see, perhaps i misunderstood from what i'd read, it seemed like people were saying that although there was an fps increase in programs like fraps there was actually no increase!

So there is an actual juddering of the image? I guess i'll just get it and see for myself.

http://www.vr-zone.com/articles/Rade..._280/5851.html

Check page 59 mate!

[BenchZowner]

Already posted.
And tbh a bit hard to believe.
If true, then it better be reflected in games with AA/AF.
That'll be a miracle and a nice comeback.

[mr2sw21]

http://www.vr-zone.com/articles/Rade..._280/5851.html

so is that finally "Golf" finally beat the "S-class"

[Monkeywoman]

The r700 doesn’t use software or hardware to do crossfire, its two beautiful GPUs working in harmony to kick some sever ass! SOO there wont be any of this "micro-studdering"

[dsaraolu]

Very hot for HD4850

[Hu1kamania]

The r700 doesn’t use software or hardware to do crossfire, its two beautiful GPUs working in harmony to kick some sever ass! SOO there wont be any of this "micro-studdering"

Surely it must use hardware if not software? Which if so is good! Doubt anyone knows but what about if you put two of them together? "Micro-studdering" then?

I think this whole "micro-studdering" thing is blown out of proportion anyways. I ran 7800GTX and 8800GTX's in SLi and I fully plan on running a multi-gpu based system with which ever card wins the gpu wars. Micro-studdering be damned! SLi has always proved to be a enjoyable gaming experience for me.

[madcho]

15% faster !!!!!!


maybe AMD's gonna ask 800$ for it !!!

[Chickenfeed]

I've always noticed the microstutter but after running multi gpu for 6 months, I'm used ot it and it doesn't bug me. Sure it isn't ideal but I'll still take the increased performance anyday. To say multi gpu doesn't help is completely ignorant. If ATI do manage to remedy this with the 4870X2 or better yet with multicard crossfire, that would be a push in the right direction.

[STaRGaZeR]

The r700 doesn’t use software or hardware to do crossfire, its two beautiful GPUs working in harmony to kick some sever ass! SOO there wont be any of this "micro-studdering"

So what are they using?

[Xello]

I've always noticed the microstutter but after running multi gpu for 6 months, I'm used ot it and it doesn't bug me. Sure it isn't ideal but I'll still take the increased performance anyday. To say multi gpu doesn't help is completely ignorant. If ATI do manage to remedy this with the 4870X2 or better yet with multicard crossfire, that would be a push in the right direction.

It sounds similar to the effect caused by Telecine judder.

If it's down to a lack of syncronization of the frames being transferred from the gpu to the display, isn't it somehow solvable in the same way that v-sync is used? Do you still get it with v-sync on?

[adpr_02]

So what are they using?

Magic.

But really, how can they not use software OR hardware? that kinda means they're completely disconnected from each other.....

[conzymaher]

Erm, its already confirmed that the R700 is the same basic structure as the 3870X2, We can see from the cooler that it is two separate dies and it was confirmed that the PCI Express bridge is now 2.0

So it uses the similar magic / hardware / software to the 3870X2 :p

[zerazax]

Not really. A bridge of some sort is required regardless if its on-board CF or not - the chips have to communicate somehow anyways!

The real question is whether the communication is like CF today or if the GPU's themselves communicate differently - the bridge is just the medium for it to pass through

[LordEC911]

By the time R700 comes out nVidia does a GX2 trick on AMD.

Ummm... R700 is due out in July/Aug.
Nvidia will need 45/40nm to even think about doing a GX2 and that won't happen for them until sometime in '09.

Very hot for HD4850

Overclocked card with the fan and GPU activity at 0%...

Erm, its already confirmed that the R700 is the same basic structure as the 3870X2, We can see from the cooler that it is two separate dies and it was confirmed that the PCI Express bridge is now 2.0

So it uses the similar magic / hardware / software to the 3870X2 :p

As far as I know which things might have changed in the last ~20hours but when I went to bed last night, nothing was "confirmed" about R700.
We are not 100% sure that black cooler is R700, not 100% sure about a PLX2.0 chip, nor 100% sure about it being the same as 3870x2.(AMD/ATi made a statement saying it was NOT going to be two RV770 on a PCB like the R680 was)

Now if you have a source verifying any of that please post it.

[w0mbat]

rv770pro: 625mhz
rv770xt: 750mhz

& there´s no shaderclk

[GAR]

so is that finally "Golf" finally beat the "S-class"

3870x2 can beat a 9800gtx all day in benchmarks, but the 9800gtx kicks ass in most games, especially with aa/af turned on. 3dmark is garbage and misleading piece of software ever, i wish people would stop using it, its not representing the true performance of the chip.

[Nicksterr]

So @ stock clocks are we looking at a 15-25% improvement from 4850>4870 in all tests? I think I'm gonna pick up an ATI card this time, mainly cuz I play EQ2 and nvidia cards are for eq2.

[y2kbos]

GPU's themselves communicate differently

Do people really think Ati have something that complex ready to launch?

[Luka_Aveiro]

Ummm... R700 is due out in July/Aug.
Nvidia will need 45/40nm to even think about doing a GX2 and that won't happen for them until sometime in '09.

If nVidia could improve G92b shader performance by 50% just like GTX280, a 9900GX2 would be something interesting, but would certainly touch the fields of GTX280.

[BenchZowner]

rv770pro: 625mhz
rv770xt: 750mhz

& there´s no shaderclk

I can confirm these core clocks.
They're 100% correct.

[zerazax]

Who knows really - supposedly R700's been in the works for some time and R600 was just the testbed for the architecture. Whether or not it actually "works" (it could always have been scrapped for sucking) or was delayed to another generation (after all the delays of R600 for instance) is hard to say. All we do know is that R700's design was rumored quite some time before and R600 had a lot of things in it that weren't fully utilized that people have hinted may have been used in a grander scheme, but nobody knows what's changed since then