HD 4870 X2 (R700) review thread

HD4870X2 to appear at NVision08

http://www.bit-tech.net/news/2008/08/04/amd-david-kirk-is-full-of-it/1

Jawed

 

I'm waiting for the photo report...

 

I'm waiting for Geo to update his sig.

Jawed

 

Heavily subsidized it may be, but it still isn't nearly as powerful, or resource rich, as a non-subsidized SoA PC gaming platform can be. Console hardware is locked for the generation of console you are looking at. Most often, console hardware is locked for 4-5 year periods--between consoles. It's precisely because consoles are so heavily subsidized in the beginning that their lifespan is extended as it is. By the end of the 4-5 year console cycle, the hardware is either no longer subsidized at all, or else it is turning a profit.

Also, subsidized or not, consoles operate inside a particular and very limited economic space--generally, the $300 and under markets. Gaming PC platforms, due to their initially more expensive costs, are far more upgradeable over the same 4-5 year periods, and are upgradeable piecemeal, too, which means for instance that merely changing out a motherboard and cpu while retaining your older components can result in performance differentials indistinguishable from having bought an entirely new PC. It's this kind of thing that you can't do with a console, because consoles were never made to provide for piecemeal hardware upgrades.

Unlike the 4-5 year console cycle, software game development operates on a 1-2 year cycle, generally, with some exceptions, of course. This means that during the developmental cycle for PC games the hardware capabilities at the game's release will have changed far more than the hardware capabilities of a console, which, unless you develop during the period immediately between old and new console introductions, will not have changed *at all.* So, again, by the very nature of the fundamental differences between the PC gaming markets and the console markets, developers have to choose which path they'll follow. Only the largest developers are able to choose both, because the demands of each market are so different.

Again, 3d PC gaming hardware develops and advances far more briskly than console hardware technology does. Some people would say, and I might be among them, that a constant frame-rate of 60 fps is sufficient, and especially if you can do that in most cases at up to 8xAA... I mean, if I can get 60 fps at 4xAA or 8xAA, then who cares about what my frame rate might be at 0xAA? 4/8xAA is going to be as completely playable as 0xAA.

I agree that the "next-gen" console is likely going to be far more powerful than the present gen. But 3d PC hardware development isn't going to sit still waiting on that to happen, but is going to continue to advance at its present rate. And that means that by the time the next-gen consoles roll out they are going to be as far behind SoA PC 3d technology as they are today. Economics is at the heart of it all.

Yes, but in what sense does AQ rely on technology advancement as IQ does? I would say that AQ generally is far less dependent on technology advancement than is IQ. To be sure, both are dependent on technology advancement in terms of the maximums possible within any given current environment. But all too often on the AQ side of things the maximums are not attempted.

But see, I'd choose to run both with as high an AA rate as was possible while still retaining a fluid and playable frame rate. That's the nice thing about IQ as opposed to AQ. With IQ you have a choice, with AQ you don't...

 

As Neliz eluded to, stuttering (low FPS) has nothing to do with micro-stuttering. Micro-stuttering is a result of uneven frame distribution, a problem which has been demonstrated to have been solved by R7xx generation hardware working in CFX (at least in the games tested thus far).

 

Look at the quoted Muropaketti numbers. It is solved. Any game which doesn't show the expected even distribution of frames is likely to be one which lacks a proper CFX profile. Alternatively, it could just be a game issue that is unlikely to ever be solved, having nothing to do with the rendering method e.g. wildly different scene contents from frame A to frame B (user moving the mouse as fast as they can).

 

the "random guy" you're referring to is none other than the infamous Sampsa (world-renowned overclocker). I'm pretty sure he didn't just make up his results... Also, IIRC he has a very close working relationship w/ATi so he is likely to have advance access to pre-release drivers.

 

In all your bickering you continue to ignore his numbers.

 

But one of his examples was using Grid with the 3870x2 and 4870x2. The frame-rate never dropped below 60 FPS with the 4870 and was in the 30 FPS range with the 3870x2.

How is the 3870x2's microstutter when it never drops below 60?

How is the 4870x2's microsttutter when it is around 30 FPS?

It offered the illusion that 4870x2's microstutter was addressed when in reality it may of been more frame-rate.

The reason why PGgames hardware's findings are important is he intentionally tried to push the hardware to receive 45 FPS and 22 FPS and there was micro-stutter.

More testing is needed.

 

At 20 to 30Fps, what your seeing isnt micro-stuttering, but old ordinary stuttering.
Micro-stuttering as what was described last year in that article is the phenomenon of seeing very high FPS numbers with fraps, but getting a slow feeling playing the game.

You could do the same tests with about every gpu and get the same results as you would get with the 4870X2. Lower the clocks on a single 4870, and go play Crysis at extremly high res, 8xaa +edge detect and game @ max. Im sure you will see loads of stuttering.

 

The lower your AVG FPS. The higher your FPS spikes can be when dealing with multi GPU. It seems obvious to me that the website in question. Lowered the AVG FPS which also would impact any FPS spikes that would occur. If you render at 60 FPS all the time. Even a 50 MS spike will be hard to see. But if your rendering at 25-30 FPS. A 100 MS pike is quite a bit more apparent. The whole issue is sitll blown out of proportion IMO. It only really shows that a single GPU delivering 30 FPS > than a multi GPU delivering 30 FPS.

 

And from a control point of view. Single GPU 30 FPS = Dual GPU 60 FPS.

Regards,
SB

 

You mean mouse and input latencies?

 

Stutter ≠ micro-stutter. Read Spyhawk's response.

You continue to demonstrate that you do not know the difference between low fps-derived stuttering, and micro-stutter. Spyhawk explains below, again.

I don't know why people aren't getting this... I've said it several times already. Maybe they'll listen to you?

 

this conversation is over.