For a new card, I would give accelero any day for proper hybrid cooling like r9-295x2 has. Accelero Advantages: - It cools GPU really well Accelero Disadvantages: - 3 slot solution limits use of boards for CFX configurations - poor memory / VRM cooling - all heat stays inside case and require additional airflow management - in CFX cooling performance is greatly degraded Hybrid Advantages: - It cools GPU really well - only 2 slots taken allowing broader selection of boards for CFX - adequate memory /VRM cooling - Big part of heat goes out from case allowing use of smaller cases with worse airflow - cooling performance does not suffer in CFX configurations much Hybrid Disadvantages: - for each card there is required mounting slot for radiator and therefore may limit case selection Side note, My accelero works good only due to memory/VRM cooling modifications I did, and even so I do not like temperature VRMs goes to.
Yes it does. One does not exclude the other. And a bad ass STOCK hybrid solution is simply more impressive than one of those shiny, cocky nVidia aircoolers. AMD took a lot of flak for their silly dustbuster coolers last year so doing something about that is also a sound marketing strategy.
You can get a Xeon X5650 that use the same socket like the 920(so you don't have to change anything), Now they are giving them for like $70 , a 32nm part, 6 cores, 12 threads, can be OC around 4.2GHz, also slightly newer architecture compared to the 920.
since 7xxx series imo AMD has done really really good with drivers. Not that I had any massive problems with 4xxx or 5xxx either. But xfire and sli is hit or miss when it comes to driver support from both teams and we must remember the ones that have problems are way more vocal then the ones without any.
I can get artifacts on a amd card or nvidia card when I oc them bit too much. Not had any artifacing issues with 7950 or 290x
Yeah everything is rumor until we see solid evidence. Still you have to admit it would be great if they were 30% faster than a 780 Ti.
I have some nice rumors (my made). There is short story of HD7950, r9-285 and r9-290x: When r9-290x came it had 512bit memory controller instead 384bit like HD7950 but that controller took 20% lower area and doubled pixel filling by doubling ROPs without significantly increasing GPU area taken. When r9-285 came it was basically cloned GPU from HD7950 with addition of TrueAudio, 2x improvement for tesselation and maybe some minor tweaks here and there, but got new HW level texture compression allowing to get around 90% of HD7950 memory bandwidth with only 256bit bus. (roughly 35% compression improvement) Now that means that if AMD makes GPU with stuff they learned, they can pair it with 384bit memory interface which will act like 512bit one due to compression (384*1.35 = 518.4). That 3GB of vRAM will in comparison to today's standard act like 4GB (3 * 1.35 = 4.05) That they can use saved space and entire card TDP to have more powerful GPU which will have at least 2x as high pixel fill rate than HD7950/70 per TMU and Twice as high tesselation power allowing for greater details. And if they happen to keep 512bit memory interface it will have bandwidth of 691bit bus (512 * 1.35 = 691.2), paired with 4GB of vram acting similar to 5.4GB. ->This I see likely for 20nm GPU next year. And if you increase memory bandwidth by 35% on GPU which was not bandwidth starved then you should increase GPU power at least by same amount. Story of the numbers short, any transistors which AMD saves on memory controller and any TDP saved by not using additional memory bricks should go into more powerful GPU. Right now GPU with 384bit memory interface can have 15% increased performance over R9-290x but be like 5% cheaper to make. GPU with 512bit memory interface would gain nothing because it does not need higher bandwidth and is at TDP limit already. At 20nm it may gain maybe 35-45%. And this bold thing is like only solution AMD can use against GTX980 if it really performs like presumed. (I lied it is long.)
