Analysis - ATI Radeon HD 4850 - benchmarked against mid-range and high-end GPUs

[Tarinder]

out of stock , anyways back on topic . why do any of the benchmarks not have the res of 1680x1050 ? , i wanted to see what they perform like at this res

1,680x1,050?

Every benchmark has been run at this resolution.

[Brewster0101]

I think ATI are pretty sure that their new 4870 will bring Nvidia to its knees with pricing. They chose to release the 4850 first to cause a ripple in Nvidias pond and with Nvidia answering with a 30% price reduction on their matching card, ATI must be rubbing their toes together with glee....

Next 7 - 14 days are going to be awesome for the consumer.... This is old school tactics by ATI/AMD, not interested in the crown, just cleaning up everything else.


If only AMD could do as well with their CPU's.

[Mithrandir]

Regarding game selection, to be honest I do tend to read around reviews until I find ones covering Oblivion. Oblivion on maximum settings can be very demanding, and every extra frame per second means it can take more modding. Dedicated Oblivion fans will tend to play with extreme texture packs (like QTP3) and mods that add more objects visible at distance. Such a tricked out Oblivion is probably the harshest test possible

I guess you mean running at 1,680x1,050 or beyond?

I already run Oblivion with Qarls TP3 and a host of other mods on my rigs at 1280 x 1024 very nicely.

[Brewster0101]

http://www.guru3d.com/article/amd-at...-powercolor/17

Crossfire results here.... Bye bye nvidia your time is up

[Biscuit]

http://www.guru3d.com/article/amd-at...-powercolor/17

Crossfire results here.... Bye bye nvidia your time is up

lol @ the crappy crysis scores. That game engine is SUCH a joke!

[Lucio]

lol @ the crappy crysis scores. That game engine is SUCH a joke!

It does say a lot when games get some markedly different scores between card types. You have to wonder whether developers are even trying to support both brands any more, or whether they're just sticking to nVidia cards because of some behind the scenes fun & games. Of course, it could also be really simple, that you can't program for both architectures in the same way, and it might just be easier to use the nVidia tools than the ATi ones.

[Biscuit]

It does say a lot when games get some markedly different scores between card types. You have to wonder whether developers are even trying to support both brands any more, or whether they're just sticking to nVidia cards because of some behind the scenes fun & games. Of course, it could also be really simple, that you can't program for both architectures in the same way, and it might just be easier to use the nVidia tools than the ATi ones.

Maybe so but its scaling isnt much better for SLI

[sawyen]

Maybe so but its scaling isnt much better for SLI

Yea, But TWIMTBP pays hell a lot more than AMD Game! I'm sure..

Its either consumer loses, but you dine in Tiffany's

or

Consumer wins, but you dine in the kebab shop..

You pick..

[Biscuit]

Yea, But TWIMTBP pays hell a lot more than AMD Game! I'm sure..

Its either consumer loses, but you dine in Tiffany's

or

Consumer wins, but you dine in the kebab shop..

You pick..

ok... but what does that have to do witht he fact that Crysis cant scale properly for any system?

[sawyen]

LOL.. sorry, I just happened to quote the wrong person.. my bad...

but still, you do have to admit my point sort of make sense (i know not to your context, but generally as a whole).. haha

[Biscuit]

LOL.. sorry, I just happened to quote the wrong person.. my bad...

but still, you do have to admit my point sort of make sense (i know not to your context, but generally as a whole).. haha

yeah and i totally agree, its a damn shame for ATI and it does make me wonder if the same game was built twice once for each card... who would come out on top?

[chaotic_russ]

1,680x1,050?

Every benchmark has been run at this resolution.

i was sure that said 1600x1200 lol , whoops sorry

[Kotoa]

Memory timings and speed 7-7-7-20 2T @ 1066MHz 5-5-5-15 2T @ 1066MHz

Why don't both systems use the same latency RAM? Doesn't seem fair.

[Tarinder]

Because getting 5-5-5-15 DDR3 RAM is more difficult than getting 5-5-5-15 DDR2 RAM.

What's going to change in the next month is the NVIDIA testing platform, from 680i SLI to 790i SLI. Using the latter, we can eliminate most of the skew between setups.

The high-end tests are generally bound by the GPU, meaning that the subsystem doesn't play a huge role in determining performance.

[vrykyl]

As you didnt do any sli testing in the review, why didnt you just use the same system and dual boot between vista with ati drivers/nvidia drivers? That would eliminate the difference in subsystem entirely....

what possible reason would you have for benching SINGLE cards in two different setups? if accuracy of results is what you are after, use the x48/ddr3 rig for ati and nvidia single card testing, and use 790i when sli benches are needed. if all the subsytem is identical then the results are definately more valid...

slight as you may say the differences are, I will largely disregard the hexus review, simply becaue the different subsystem is entirely unneccessary, and say what you will, that makes any comparison of apples to apples unfair.

[Tarinder]

vrykyl

Because these are the same subsystems that will be used for SLI and CrossFire testing.

From previous experience, using these two different subsystems, compared to just one, represents less than the standard deviation between current test results, anyway.

We can re-use these numbers and offer a multi-GPU perspective.

ATI dropped this on everyone at the 11th hour; the reviews were supposed to be for the middle of next week, hence the preliminary look.

If we choose just one subsystem, and it's Intel-based, we'll hear claims of being biased towards NVIDIA. If it's NVIDIA-based, we'll hear claims to the contrary.

Put simply, if you're considering purchasing any of the cards, the likely scenario is that you'll be running a GeForce card on an NVIDIA platform and an ATI card on an AMD/Intel platform, to keep a multi-GPU path intact, and the results reflect that.