Images, Benchmarks, Specifications of AMD's R9-290X
New specifications, as well as images and benchmarks of AMD's upcoming Radeon R9-290X 'Hawaii' graphics card have surfaced.
Earlier this month we reported some specifications of AMD's upcoming Radeon R9-290(X) cards, and now we bring you more. A new report from DG's Nerdy Story, which seems to have been taken offline, shows a number of pictures of the Radeon R9-290X card, as well as some benchmarks.
According to the earlier rumors, as well as newer reports, the card will supposedly feature up to 2816 stream processors, all of which will be split among 44 different clusters. In all likelihood, the Hawaii chips will feature either 32 or 64 ROPs along with about 176 TMUs. The chip will be built on the 28 nm lithography and manufactured by TSMC. It'll feature the GCN 2.0 architecture, as well as support for DirectX 11.2. The GPU is expected to be clocked at 900 MHz or above. The memory interface will be a 512-bit wide memory interface, carrying GDDR5 memory, though no reference capacity was mentioned in the report. The chip appears to carry a total of sixteen memory chips, which is where the idea of a 512-bit memory interface comes from, as 384-bit wide memory interfaces normally come with 12 memory chips.
Image Source: DG's Nerdy Story
Image Source: DG's Nerdy Story
In benchmarks, the card was earlier rumored to be faster than the GTX Titan, but less powerful than the GTX 690. The benchmarks posted on DG's Nerdy Story support this, and the card appears to be notably more powerful than the Titan. While the model name of the top graphics card is blanked out on the benchmark tables below, we'd say it's fairly safe to assume reference to either the Radeon R9-290X or the Radeon R9-290.
Image Source: DG's Nerdy Story
Image Source: DG's Nerdy Story
There was no word on what the card would cost, though we can either expect it to be priced in-line or slightly above the GTX Titan. Below it by a fair margin is also a possibility if AMD wants to destroy the competition. That said, there are numerous sources pointing towards a price of just $600. AMD will likely be showcasing the GPUs this week, with sales kicking off around mid-to-end of October, starting with the Radeon R9-290 (Hawaii Pro) and the Radeon R9-290X (Hawaii XT).
Follow us @tomshardware, on Facebook and on Google+.
That said it is interesting how these benchmarks show it pulling ahead more with AA enabled than without AA enabled. That's a good sign.
Again we must take these with a grain of salt but the PCB design looks just like a standard Sapphire designed card so it might be real.
But more seriously...We must keep an open mind, IF these benchmarks are true, coupled with a lower price, more ram, more bandwidth, better opencl performance, better overclockability (if the 7xxx is any indication), the Nvidia 780 and the Titian are effectively crippled.
More benchmarks: View this thread in the forums for the pics, the reply system hates picutures.
I remember hearing that when the 680 originally came out that it was org supposed to be the 660Ti. And the 780 is just a lowered down version of the Titan. So I get what you are saying but I have a feeling that AMD/Nvidia both do the holding out game waiting to see what the other person does then one upping them all the time.
THIS is what the CPU wars should be. I mean we can be fan boys over each brand but thsi is what we want form AMD/INTEL.
There is no price war nor will there be any on the CPU side any time soon since very little mainstream software can make use of the processing power already available. Until enough mainstream applications start requiring quad-cores to threaten Intel's market share, I would not expect a major market shift for another 5-10 years.
Another possibility is that applications and games with embarrassingly parallel computations will simply shove those onto OpenCL and not really care how fast (or slow) the CPU actually is as long as there are sufficient OpenCL resources available.
There is no price war nor will there be any on the CPU side any time soon since very little mainstream software can make use of the processing power already available. Until enough mainstream applications start requiring quad-cores to threaten Intel's market share, I would not expect a major market shift for another 5-10 years.
Another possibility is that applications and games with embarrassingly parallel computations will simply shove those onto OpenCL and not really care how fast (or slow) the CPU actually is as long as there are sufficient OpenCL resources available.
I'm just saying that the cpu market right now is very stagnant and a lot of that is because intel doesn't really have anyone to push them.
All I was really saying was that with amd and nvidia really pushing each other it would be nice if AMD had the same ability with their CPU line since they are the only other desktop CPU out.
Time will tell with GPU's exciting to see the advancement with things. Although one has to ask themselves outside a game like star citizen what do we really need the extra power for at the current moment. Any high end AMD chip now or Nvidia chip now can pretty much do w.e it wants with the games that are out now with rare exceptions 3d/4k/and multi screens.
The difference is that AMD does not cripple their gaming cards from being able to run production apps.
While the FireGLs are designed with production in mind (top of the crop silicon much like server CPUs), a HD7970 eats a Titan up in OpenCL (hence why AMD is so popular with the various Coin mining types which heavily use Open CL) since NVidia makes their gaming GPUs do pretty poor in OpenCL. NVidia wants you to spend a few $K on their Quadros instead.
http://www.tomshardware.com/reviews/geforce-gtx-titan-opencl-cuda-workstation,3474-19.html
I don't see why AMD will change that now unless they get greedy.
The difference is that AMD does not cripple their gaming cards from being able to run production apps.
While the FireGLs are designed with production in mind (top of the crop silicon much like server CPUs), a HD7970 eats a Titan up in OpenCL (hence why AMD is so popular with the various Coin mining types which heavily use Open CL) since NVidia makes their gaming GPUs do pretty poor in OpenCL. NVidia wants you to spend a few $K on their Quadros instead.
http://www.tomshardware.com/reviews/geforce-gtx-titan-opencl-cuda-workstation,3474-19.html
I don't see why AMD will change that now unless they get greedy.
But you still run into things that are either opencl or cuda favored or this game is favored by this amd card or this nvidia card there isn't a constant.
Look at how things turned out with tomb raider with nvidia cards. upon release the cards had a hard time same could be said for amd cards with metro last light light its not a clear yes or no answer.
Although going back to production work I think you would need to find a production environment where both cuda and opencl could be tested to really give it a go.
Also, saying OpenCL is poorly ran on nvidia cards is really a mute point thats like saying Physx is poorly ran on AMD cards. Both companies utilize different tech. It's fair when you can take a apples to apples comparison Havok versus Physx versus CUDA versus OPEN CL in a application or game that uses it.
Hopefully AMD fixes their various driver woes with this launch, as well. If they ever get their software support as good as Nvidia's, consumers will really win big time.
Hopefully AMD fixes their various driver woes with this launch, as well. If they ever get their software support as good as Nvidia's, consumers will really win big time.
And this wasn't arrogant of AMD? http://www.youtube.com/watch?v=eH6XayaLTw8
Companies only push the envelop when they feel the pressure. AMD needs to push the envelop to survive due to financial reasons. Nvidia has never felt the pressure money wise so they react to the product rather then their internal pressure.
I am referring primarily to:
-Nvidia releasing and pricing GK104 as if it were a high end card
-Nvidia releasing GK110 absurdly late and launching it with the laughable price of $1000 (Titan)
Companies only push the envelop when they feel the pressure. AMD needs to push the envelop to survive due to financial reasons. Nvidia has never felt the pressure money wise so they react to the product rather then their internal pressure.
So what your saying is that because of said
The biggest problem for CPUs is the lack of mainstream software that actually needs more processing power than is already available from both AMD and Intel. Right now, the only things that require that much processing power are professional-oriented suites and the highest-end games that cater mostly to twitch-gamers.
Without CPU-intensive mainstream applications to make most people wish they had faster CPUs the same way we did 10 years ago when we still had to wait for word processors to complete page formatting changes or spreadsheets to re-calculate after updating cells, there is little low/mid-range market demand to justify increasing processing power offer in those segments at the expense of more profitable higher-end segments.
If a mainstream killer app that cannot be OpenCL'd and requires the equivalent of an i7-4770 or FX-8350 to be somewhat usable comes out, things might shuffle quickly to accommodate that. But at this point in time, I cannot imagine anything that would require that much processing power while being a nearly universal must-have like h264 is - with nearly everyone using Youtube, Netflix and other video streaming services on a regular basis, most people wouldn't put up with a computer or tablet that lacks sufficient processing power to handle h264 for long. h265 and 4k might be the next biggest mainstream processing time sink on the horizon but by the time they become mainstream, most IGPs will decode that in hardware and make it a non-issue.
So, personally, I blame software for the stagnation between AMD and Intel: without CPU-intensive mainstream software to force average users to demand or desire faster CPUs, most of them will settle low-end CPUs that are already more than enough for most of their other everyday tasks and the gap between low-end and high-end will only continue to grow wider both on prices and performance: price pressure at the low-end and price inflation at the high-end due to marginalization and hollowing out of the mid-range segments.
If you want a "CPU war," first you need to create demand for such processing power.