Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
AMD teases new FX-series processors
https://twitter.com/amd_roy/status/480034598554845184
So what is it? Based on AMD's server roadmap, this chip is probably to Vishera as Richland is to Trinity: still Piledriver cores, but a more optimized chip. In other words, maybe 5% faster in the same TDP and cost. Now, given the choice between 5% faster or not 5% faster, you'd rather have 5% faster. But still, it's not a huge deal.
AMD could have made a chip with 8 Steamroller cores by now if they wanted to. But based on their server roadmap, they probably don't, or perhaps rather, don't want to do so badly enough to justify the cost. It's worth noting that Steamroller cores make more sense in servers than desktops due to their improved energy efficiency but lower clock speeds, so there's no way that they'd release a desktop version of the chip but not a server version. There have been rumors that AMD was going to make a server chip with 16 Steamroller cores, but such a chip would probably be clocked so low as to be mediocre in desktops for most purposes--and such a chip isn't on AMD's 2014 server roadmaps.
Comments
Well, with Devil's Canyon turning out like it is after the Intel "5Ghz on Air" hype, all AMD has to really do here is not throw out another 2-step-back Bulldozer performance fiasco and they will come out looking like heroes, even if the chip is still a good deal slower than Intel.
What about people who want to pay rather less than $250 for a CPU?
Yup. There are a lot of budgets out there where spending $250 on a CPU is just not a possibility. AMD's CPU's at the ~$100 price point make an excellent bang for the buck.
Sure, they may not compete with a 3570k, but you know what? It's fast enough for most nearly everything. A few games will CPU bottleneck on a modern AMD, but only a very few.
My car isn't a Mustang Shelby GT, it can't do 200mph+. But it at least goes the speed limit (and a bit more, with the wind at my back), and it cost a hell of a lot less.
Agreed, one of the reasons why i have stuck with AMD for my CPU's is because of the price point, much as i would like to get the equivalent Intel CPU, the cost is an issue, i am at the moment still using an AMD phenom I quadcore at 2.6ghz, my next build will probably be the 8 core AMD processor, simply because its the best i can afford
I think AMD really depends on what you are doing. If you are doing a lot of zipping, encoding, and calculating PI (typical benchmarks for CPUs). Than an Intel trounces AMD. If you are doing something massively parallel like Rendering or calculating AI or calculating physics; AMD trounces Intel. Finally if you are in a game. No developer in their right mind would develop a game for a 3.5 Ghz quad core without first trying to make more cores usable or offloading to a GPGPU. It would be a tiny market for a game that can't be played on a Pile Driver based CPU.
You are quite spot on all the way up to the "No developer in their right mind...". The thing is, working with a quad 3.5Ghz processor is simply a matter of scaling up from working with a (say) quad 2.6 Ghz processor. The gain is direct.
To spread the load between the GPGPU or another core requires a totally different approach. It requires you to actually put in the effort to make it work. For example, I applaud that AMD is trying to do it with it's Mantle, but it's something that needs to be fixed on a core level.
Even the newest console games, built with multiple AMD cores in mind, when played on a PC with Mantle enabled perform better with Intel CPUs.
Of course, a Pile Driver will probably allow you to run all the current games well but a CPU is not something replaced as easily a GPU so why not get something slightly more future-proof? So far the 1155 socket mobos have been pretty great no matter the manufacturer and should last a while.
GPGPU actually doesn't make much sense for games outside of traditional graphics purposes. For starters, the video card tends to be busy on graphics when gaming. Furthermore, even if that weren't the case, GPUs are very finicky about what they can do well, so even though most of the work to program a game could be done on a GPU, the overwhelming majority of stuff that wasn't traditionally done on the GPU in the graphics pipeline will run massively faster on a simple x86 CPU--even a low end model.
That's not to say that GPU compute can never be useful for gaming. It can have some niche uses and it wouldn't be completely shocking if there are eventually quite a few games that do some GPU physics computations outside of the normal graphics pipeline. Planet Explorers says that they're using OpenCL, and gives me the impression that this is instead of the normal graphics pipeline. It's only that most stuff that games need to do simply doesn't fit what GPUs are good at.
In Game development, a better CPU has not really progressed to the same degree as a GPU. If we exclude things related purely to the GPU or could be executed better on the GPU, than the real advancements were to new input gathering and AI. Over the last decade I have not seen these aspects improve at all in games. So really I only see the most recent advancements in CPU for games to be related to physics and preparing data for the GPU. The GPU related code was also reduced in DirectX 10, and will be again in DirectX 12. So really its only physics that could share calculations on the GPU.