Slashdot Mirror
AMD Unveils New Family of GPUs: Radeon R5, R7, R9 With BF 4 Preorder Bundle
MojoKid writes "AMD has just announced a full suite of new GPUs based on its Graphics Core Next (GCN) architecture. The Radeon R5, R7, and R9 families are the new product lines aimed at mainstream, performance, and high-end gaming, respectively. Specs on the new cards are still limited, but we know that the highest-end R9 290X is a six-billion transistor GPU with more than 300GB/s of memory bandwidth and prominent support for 4K gaming. The R5 series will start at $89, with 1GB of RAM. The R7 260X will hit $139 with 2GB of RAM, the R9 270X and 280X appear to replace the current Radeon 7950 and 7970 with price points at $199 and $299, and 2GB/3GB of RAM, and then the R9 290X, at an unannounced price point and 4GB of RAM. AMD is also offering a limited preorder pack, that offers Battlefield 4 license combined with the graphics cards, which should go on sale in the very near future. Finally, AMD is also debuting a new positional and 3D spatial audio engine in conjunction with GenAudio dubbed 'AstoundSound,' but they're only making it available on the R9 290X, R9 280X, and the R9 270X."
Wow, that $89 R5 actually looks surprisingly attractive. If the benchmarks hold up, I might think about replacing the old power-hungry card I've got in my main desktop machine right now - I'd probably save energy and get better performance to boot.
that work reliably for more than the current crop of just released games, I don't care how much faster these chips are. I've had too many glitches with radeon drivers over the years to consider them again. Their opengl is horrible, and CCC is a bloated pos.
Personally I would've gone for a mention of Mantle, the proprietary API they are introducing that sidesteps OpenGL and DirectX. I don't really know what it does yet, haven't found good coverage, but DICE's Battlefield 4 is mentioned as using it, and the description I've read said it enabled a faster rate of calling Draw calls.
http://www.xbitlabs.com/news/graphics/display/20130924210043_AMD_Unveils_Next_Generation_Radeon_R9_290X_Graphics_Card.html
Where did you buy it from that you paid that much? I bought a 2GB 7850 over a year ago for around that price.
AMD has totally ruined the future of Nvidia and Intel in the AAA/console-port gaming space. Working with partners at EA/DICE, AMD has created a 'to-the-metal' API for GPU programming on the PS4, Xbox One, and any PC (Linux or Windows) with AMD's GCN technology. GCN is the AMD architecture in 7000 series cards, 8000 series, and the coming new discrete GPU cards later this year and onwards into the foreseeable future. It is also the GPU design in all future AMD CPUs with integrated graphics.
GCN is *not* the architecture of any Intel or Nvidia products, neither now or in the future. Nvidia and Intel will be stuck with only openGL or directX versions of games, and these versions will be much slower/feature incomplete compared to 'Mantle' versions ported form the consoles.
OpenGL and DirectX are OBSOLETE methods of controlling rendering for future AAA games. Both of these APIs/drivers have massive state overheads, and can never be made efficient for the mixed rendering/GPGPU methods required for the new games engines of late 2013 and later.
While some nerds with a better memory than brainpower will dribble about 'Glide' (the proprietary API from now defunct 3DFX), the correct comparison is x86 vs 68000 (the Motorola CPU design). GCN is actually an ISA (instruction set architecture) like the x86 ISA. Intel and Nvidia are like TI and Motorola at the time of emerging competing 16-bit CPU designs that finally led to the dominance of x64. Nvidia is Motorola. Intel is TI. TI's 16-bit CPU designs were no-hopers. Motorola was widely seen as superior to Intel at the time. When Intel was chosen for the first PC, it was game-over for Motorola.
Using OpenGL or DirectX to 'program' a modern GPU is like using Fortran to program the CPU. Using 'Mantle' on the other hand is like using 'C'. However, because 'Mantle' closely connects to the GCN 'metal', it is almost impossible to envisage a version of Mantle for competing GPU architectures.
Of course, ATI customers with 6000 series cards or earlier (or Zacate, Llano, or Richland APUs) are as out-of-luck as Intel and Nvidia GPU users. AMD is only supporting GCN, because older GPU designs from AMD use a different GPU ISA.
With the rise of Mantle, many console games developers are going to desire that the PC market rapidly changes to AMD only, so the ported games need have only one version- the good one. Other developers, whose games do NOT need strong GPU performance, will wish to use only OpenGL on the PC, for maximum compatibility with games in the ARM space (where OpenGL ES rules).
Any PC gamer interested in high-performance would be INSANE to buy any Nvidia product from now on. 99%+ of all AAA games will originate on the new consoles released later this year- which are 100% AMD. Most casual gamers might as well choose AMD for maximum future compatibility. Intel was never really in the game. Nvidia, on the other hand, will be cursing themselves for ever starting this war (Nvidia previously paid AAA games developers to cripple AMD performance, and attempted to leverage the proprietary PhysX physics engine).
Mining bitcoins with GPUs is no longer profitable for the most part. Most profitable miners today use hardware designed specifically for bitcoin mining.
If these links are to be believed, this is just another rebadge:
http://news.softpedia.com/news/First-Sighting-of-Radeon-R9-280X-Really-a-Rebranded-HD-7970-GHz-Edition-386239.shtml
http://videocardz.com/45877/amd-radeon-r9-280x-rebranded-hd-7970-ghz-edition
http://www.techpowerup.com/191440/radeon-r9-280x-is-rebranded-hd-7970-ghz-edition.html
http://www.guru3d.com/news_story/radeon_r9_280x_could_be_a_rebadged_hd_7970_ghz_edition.html
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
The same thing that happened to the Intel i2, i4 and i6 processors.
just look at the bit rate on the memory bus. Video card manufactures use the mem bus bitrate to limit card performance so that their low end doesn't cannibalize their mid range and high end (ala 3DFX).
:)
128-bit is low end.
192-bit is your mid range card.
256-bit is your high end.
You don't need to pay attention to anything else until 256 bit. After that just sort by price on newegg and check the release date. Newer is better
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
The idea is that operating systems introduce a huge amount of overhead in the name of security. Being general purpose, they view their primary role as protecting all the other apps from your unstable app. And, lets face it, even AAA games these days are plagued with issues -- I'm really not sure I want games to have low-level access to my system. Going back to the days of Windows 98's frequent bluescreens isn't on my must-have list of features.
John Carmack has been complaining about this for years, saying this puts PCs at such a tremendous disadvantage that consoles were able to run circles around PCs when it came to raw draw calls until eventually they simply brute-forced their way past the problem.
Graphics APIs have largely gone a route that encourages keeping data and processing out of the OS. That's definitely the right call, but there are always things you'll need to touch the CPU for. I'm curious exactly how much of a benefit we'll see in modern games.
Keep soldiering on, AC
Pain is merely failure leaving the body
Wonder how long until HD manufacturers adopt 1k == 960 to "avoid customer confusion"
You mean today you just crawled out of your hole, considering AMD has all three consoles, and they're about to drop a brand new graphic architecture to the table.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
Yeah, good luck with that. Of course it is BGA, do they even make modern RAM in anything else these days? I think it is even part of the GDDR spec at this point, but there are so many pins, and the industry is all about density, so the chips are going to be *GA of some variety. So yeah, toaster oven, after you have disassembled the card to remove the melty plastic bits, then you get to put it back together and find out it doesn't work because it didn't reflow properly. I suppose with proper investment in your toaster oven controller you could get something proximate to an appropriate flow profile, but that is far from a guarantee of success. You won't have the xray, and while JTAG adapters are cheap and plentiful, good luck getting a JTAG pinout from , so even knowing if its at all correct is a stretch, and at the speeds they use as mentioned above these things are quite sensitive to the quality of the joint. If you don't get it right on your first go, you'll probably pull some pads trying to rework it. I would put my money on the card, the ram, or both being ruined as opposed to success - unless doing this sorta thing is your day job. You are right in that it can be tricky to get your hands on a small quantity of parts, but I have had good success getting samples if you put some effort into it (some things are rather simple to get with web order forms, but you won't find that for Samsung and the like- probably gotta go old school and reach into the sales channels). That being said, they probably have something like 4-8 chips (4Gb I think is towards the high end of available densities) and I've never had anyone offer me more than 5 in samples (not that I've asked).
Worth it? You'll spend more breaking stuff than it would have cost to just buy it with what you want.
FTR, soldering normal SMD packages doesn't scare me, SOIC and the like aren't so much trouble. I'd even be willing to give *GA a shot - if I was starting from a bare board and could afford to chuck a bunch of them.