KitGuru KitGuru.net – Tech News | Hardware News | Hardware Reviews | IOS | Mobile | Gaming | Graphics Cards
Sadly AMD's HD6850 launch was marred a little by European pricing issues and again later when it was found out that some publications had a version of the card with 1120 unified shaders. Some of the early HD 6850 samples were built with pre-production ASICs that have the higher number of 1120 Stream Processors enabled, like the HD 6870. These ASICs were supplied by the manufacturer as Pro (HD 6850) parts together with a BIOS for them running at Pro clocks.
There has been somewhat of a frenzy online saying that all reviews are worthless and that everyone needs to reissue their content. Thankfully our own testing wasn't tainted by the shader discrepancies and the results today show just how important these missing 160 unified shaders can be in terms of outright performance.
So how badly tainted are the reviews featuring HD6850's with 1120 shaders onboard?
The synthetic tests are improved considerably, especially Tessmark when running at moderate settings (a 36.8% increase). Alien V Predator and Lost Planet 2 also show huge gains (9.6 and 12.2%) from the extra 160 unified shaders.
KitGuru says: The HD6850 and HD6870 are both worthy gold award winners, and we hope this article today has helped clear up performance differences between the cards. AMD partners seem genuinely embarassed by this mistake and are offering to reissue samples to publications who wish to retest.
Fascinating stuff Zardon. thanks for the info, I feel more educated now 😉
Excellent, the 960 performance is great, but I can see how the 1120 shaders can enhance the results, thats for sure.
I ordered a 6870 on Friday, good reading thanks 🙂
If a few of these get leaked outside review sites they would be worth some money in the future, as a product never released 🙂
Very interesting, and ive never seen a ‘shader’ based review before. they really do make a difference, depending on the engine.
fascinating read, thank you.
3dmark gets a nice boost of a 1000 points or so, nice.
Heh, shame they didnt start selling these !
This is a lesson in shader power and how important it is. would have been awesome if they had sold these in stores for the same price 🙂 a quick OC and its 6870 goodness at less money.
Great to get facts rather than speculation. testing is really useful.
Can AMD hardlock cards from being overclocked? I thought they could. they should have sold t he 6850 at the same price with 1120 shaders, but hardlocked it so it couldnt be overclocked.
Only downside with that Garry is that people would be able to hack the bioses to get the clock speeds up. then they basically could get a 6870 for 140 quid rather than 200. but i like the idea. a locked 775 core 6850 with shader power of 6870 which would be clocked at 900mhz and able to be overclocked. i think they would sell more, but it might stop sales of 6870 long term.
The Tessmark result is strange. How can you get a 36.8% performance increase with 16.7% more shaders? Must be something else at play.
Absolutely – no idea, but ive asked the program author to comment.
Amazes me at times how many sites do little research, I saw one other writer saying kitguru had run tests with a 1120 shader HD6850 in previous reviews. Yet here we are looking at this article and the 960 shader results in this match the sapphire and XFX reviews published last week for the 960 cards. the 1120 cards are clearly much better performers. Thanks for the info, very good read.
I think the news here on kitguru, while controversial and funny at times tends to damage the amazing quality of the reviews.
Thanks, my HD6850 arrived and my results are much the same as the 960 core version in this article. Bit of a bummer they didn’t let a few 1120 shader versions out into the wild as the differences are quiet noticeable.