ATI 4xxx Series Thread

[Avalon]

I would be very disappointed if the HD4870 was barely as fast as a 9800GTX, considering the 3870X2 is generally marginally to moderately faster than it (minus a few games such as quake wars). I have been expecting it to at least perform on par or slightly better than a 3870X2, and if that's not the case...meh.

 

[Quiksilver]

IIRC, it's 4870 is supposed to be 30% faster than 9800GTX, to I take it 4870X2 will be about 50% faster than 9800GX2 and probably step on the toes of a GTX 260, maybe a tad faster.

 

[Quiksilver]

Meet 4870 and 4850

 

[Dkcode]

Already been posted mate.

Still though its a nice looking card.

 

[Janooo]

4870 for sale now!

 

[ghost recon88]

Originally posted by: Janooo
4870 for sale now!

Why would someone be that desperate to pay $571 USD for a watered down GDDR4 version of the card? Especially when the retail is gonna be GDDR5, and cost a whole lot less. I'm surprised to see Tomshardware actually buying an ES product, rather then waiting for the company to give them a sample. Couldn't they be sued for obtaining an ES product through some other channel other then the vendor?

 

[Nemesis 1]

The seller maybe . Not the buyer

 

[Rusin]

nothing

 

[Bateluer]

Originally posted by: Rusin

Originally posted by: ghost recon88
Why would someone be that desperate to pay $571 USD for a watered down GDDR4 version of the card?

Well someone bought typical Intel quad core engineering sample with $2400 price tag and it weren't even the enthusiast level processor..so $571 shouldn't be problem.

Few people have bought them. Also Tom's Hardware bought them so they can release scores early. At least they could tell something about how HD4850 cards will perform if they down clock it..

They may anger AMD by doing so, which would mean they won't receive final silicon to review. And that would continue on to future AMD products likely.

 

[Rusin]

Originally posted by: Bateluer
They may anger AMD by doing so, which would mean they won't receive final silicon to review. And that would continue on to future AMD products likely.

They might post it as rumour with pics.. someone sent them those pics..etc.

 

[Janooo]

Some AMD marketing propaganda

770pro goes against 9800GTX.
770xt is 30% faster.

The big question is 700 scaling.

 

[Rusin]

nothing

 

[Nemesis 1]

Wait until you guys see this game perform with a ATI 4000 series card its going to knock your sucks off.

http://www.pcgameshardware.de/...ce_Review_Mass_Effect/

 

[Rusin]

nothing

 

[taltamir]

Originally posted by: dv8silencer
approx. 15" - tri-slot

haha... you know it...

And the G280 is probably gonna come in a separate BOX with its own PSU.

 

[Nemesis 1]

Originally posted by: Rusin

Originally posted by: Nemesis 1
Wait until you guys see this game perform with a ATI 4000 series card its going to knock your sucks off.

http://www.pcgameshardware.de/...ce_Review_Mass_Effect/

Let's see those performance numbers when there's driver support for that game.

This isn't Away its meant to be played game fella. This isn't crysis.

 

[Rusin]

nothing

 

[taltamir]

driver support for what? single card GPUs?

AMD just seems to completely dominate on this game. GOOD FOR THEM... finally a game worth playing where they can keep up with nvidia.

Nvidia cards still perform well on this game... Actually both cards never go below 30 MIN fps (not average... I think the min FPS should stay above 30 for true smoothness).

Only thing I wonder about is their misguided choice to use only 2GB of system RAM. Testing all those cards on a super high end CPU, And their choice to omit the second most important test... the 1920x1200 test. As well as completely ignore the non AA/AF tests...

The current test priority today should be:
1. 1680x1050
2. 1920x1200
3. 1280x1024
4. 2560x1600

And each should be tested with and without AA/AF.

Is there any chance it makes more of a difference for a 8800GTS 512MB then a 3870 512MB due to the 3870 supposedly using less gram when AF is on? and thus might require more system ram to compensate and which is limited by the 2GB of system ram they used?
mmm... actually after asking I have to say I SERIOUSLY doubt that is possibility. It seems that the 3870 is simply faster then the G92 on mass effect at mid and low resolutions with max AA/AF.

 

[golem]

Originally posted by: taltamir
driver support for what? single card GPUs?

AMD just seems to completely dominate on this game. GOOD FOR THEM... finally a game worth playing where they can keep up with nvidia.

Nvidia cards still perform well on this game... Actually both cards never go below 30 MIN fps (not average... I think the min FPS should stay above 30 for true smoothness).

Only thing I wonder about is their misguided choice to use only 2GB of system RAM. Testing all those cards on a super high end CPU, And their choice to omit the second most important test... the 1920x1200 test. As well as completely ignore the non AA/AF tests...

The current test priority today should be:
1. 1680x1050
2. 1920x1200
3. 1280x1024
4. 2560x1600

And each should be tested with and without AA/AF.

Is there any chance it makes more of a difference for a 8800GTS 512MB then a 3870 512MB due to the 3870 supposedly using less gram when AF is on? and thus might require more system ram to compensate and which is limited by the 2GB of system ram they used?
mmm... actually after asking I have to say I SERIOUSLY doubt that is possibility. It seems that the 3870 is simply faster then the G92 on mass effect at mid and low resolutions with max AA/AF.

Those numbers just look wrong, at least for some of the cards. A 9600GT shouldn't be faster than a 9800GTX at the lower rez test.

I'm looking at the pcgameshardware.de numbers.

 

[Rusin]

nothing

 

[Nemesis 1]

This game has been ported from Xbox. So does that maybe help some out on the AA. If you don't know google is your friend.

 

[Rusin]

nothing

 

[taltamir]

didnt ATI design the xbox video card? that would explain the disparity then, if the game engine has been designed with AMD type GPU in mind...

Originally posted by: Rusin
The reason why there aren't any AA scores is that this game uses Unreal3-engine and AMD's graphics cards don't support it. AMD-card + Unreal3 based game + no AA seems to work well. Problem is that Nvidia cards don't seem to take big performance hit when enabling AA. For example in Unreal Tournament 3 game is still playable at 1920x1200 8xAA 16xAF with 8800 GTS 512..and fps numbers on 0xAA aren't as much higher that one could imagine.

We have seen these driver issues so many times. Radeon cards sucked..it self.. in Lost planet DX9 first, but new driver release made really strong player (well they still lose badly in DX10 mode). 8600 GTS were first closer to 7600 GT in some game and after driver update it was on par with 7900 GTO etc.

oops, I thought they test only with AA... not only without AA. I misread that.
If AA is impossible then it is impossible. No big deal. That means that the only thing I don't like about this test is:
1. Lack of ram, but this shouldn't really change the relative performance.
2. CPU of choice.
3. Lack of higher resolutions.

that is a pretty serious list of "only".

 

[golem]

Originally posted by: Nemesis 1
This game has been ported from Xbox. So does that maybe help some out on the AA. If you don't know google is your friend.

That may explain why ATI is doing better, but it doesn't explain why in the lower rez test, a 9600gt beats a 9800GTX. Isn't a 9800GTX and 8800 GTS/512 very close to being the same card? Their scores should be much closer.

I'm not saying this one thing invalidates all the benchmarks, but it does indicate that they should be taken with a grain of salt until more test are done.

 

[taltamir]

isn't the 9600GT the only card to increase the PCIe bandwidth? remember that whole deal with people accusing nvidia of "cheating" because the 9600GT increases the PCIe bandwidth with 174+ driver and an nforce mobo? Maybe the 9600GT being a 9800GTX with half the shaders and lower clockspeed is compensated for by the ... no wait that makes no sense, the 9600GT also has a lower bus if I recall correctly, with less then 256bit that the 9800GTX has. It makes no sense for it to come on top.

Maybe multiple testes checked the cards on different systems and this was the result?

Or maybe we have something similar to the NWN2 (also aurora engine) bug where the 8800GTX suffered from an infinite loop while trying to render water reflections, causing severe slowdowns and much lower performance compared to lower end nvidia cards.