Where are our #Google overlords when it is required? Come on! :-)

In my opinion, limitations on the CPU side will be compensated by using Open CL in the GPU. But it is just too soon for adapting SDKs, so I bet it will get much better FPS, as the GPU gets proper usage.

That's terrible for Nokia. The few chances for its survival, IMO, now are gone :-S

There is no justification in this world for such crazy massive generalization of power hog Intel CPUs in servers: Intel's CPUs are only justified for per-thread maximum performance. And that is unnecesary for 99.9% server applications.

This article shows the obvious: Excluding caches, performance per transistor in Intel x86 CPUs is very low. As example, current best performance per clock in the Intel CPUs is the AVX (Core i5/i7 -Sandy and Ivy Bridge-), delivering up to 8 FLOP per cycle with AVX SIMD opcodes (2 SIMD ALUs) while in previous generations was just 4 FLOPS per cycle with SSE2/3/4 (just 1 SIMD ALU). Thats miserable (back in 2000, the Playstation 2 was already capable of FMAC opcodes with 8 FLOPs/clock per SIMD ALU!!!). As example, similar performance with 4 FLOP per cycle with one SIMD ALU, at one fraction of waffer area.

Here is a 50$ ARM general purpose multicore-CPU example for matching 999$ performance of fastests Intel Core i7 (e.g. i7-3770K 3.9GHz (peak), 4 CPU, 8 threads, 2 SIMD ALU/CPU = 8 SIMD ALUs = 64 FLOPs/clock -> 3.9*10^9Hz * 64 FLOP/s = 249.6 GFLOPS:

• 4 x ARM OooE (e.g. Cortex A9-like) 2.0GHz with 2 SIMD FMAC-capable ALUs/CPU (ALU = 16 FLOPs/clock, i.e. 2 ALUs = 32 FLOPs/clock -> 4 * 2.0*10^9 * 2 * 16 = 256 GFLOPS
• 4 * 32KB + 4*32KB (256KB) L1 full-speed code and data cache
• 4 * 256KB (1MB) L2 half-speed cache
• 2 MB L3 half-speed cache
• 2 or 3 lane ring bus (cheaper interconnect).

For increasing integer and load/store performance, it could be achieved with pipeline and issue/execution modifications, using more functional units. The limit is to keep the OooE simple enough for avoiding wasting transistor in executing tons of instructions unnecesarily.

Because it is THE LAW. </Dredd>

It is THE LAW. </Dredd>

"clever" differential updates usually work this way (Chrome browser uses it or used it back in the day):

• Unpacked blob version 1.
• Packed blob version 1
• Unpacked blob version 2.
• Packed blob version 2.
• Differential packed blobl version 2, taking version 1 as compression base dictionary (or any other differential scheme).

And Git has can not do that yet, because it uses diff + deflate, having far less scope than, e.g. LZMA with 500MB dictionary (requiring 5GB of memory for compressing it is acceptable if it is done just once per version).

Those pads have AllWinner A10/A13 SoC (ARM Cortex A8 @1.2GHz and GPU ARM Mali 400), 512MB of RAM, and 4GB of flash. I see no reason for not having mobile phones with similar technology (the AllWinner A10/A13 is a tiny SoC) for similar price (e.g. Broadcom or Qualcomm could add 3G easily and sell their own cost-killer SoC for smartphones). IMO, is going to change everything, as everyone will be able to have an smartphone.

Because you can?

Right, but you could use QEMU for launching WINE on non-x86 platforms.

What trade secret? Published keys were already "published" by Sony, because a defective security implementation. It is like writting a number in the box and forbidding you to tell it to others. Geohot just told others what was the keys for his console, and the fact that these keys were the same for all consoles is just Sony's fault. In my opinion Sony deserves this and much more, because of fuking their customers (I still hold PS3 firmware 3.15, for using Linux, but not being able to play new games, because the *requirement* of firmware update). By the way, I have no intention of buying anything produced by Sony, including their media brands (e.g. Columbia Pictures).

You're wrong. 32-bit is 4 bytes, so 1920*1080*4 is 7.9MiB/frame, 474MiB/s at 60FPS. With 20GB/s memories is not such a big problem. Of course, a dedicated bus and memory is better.

Quite ironic to help against themselves. If there was no actual "enemy" no missile barrier should be needed at all.

I no longer buy new games since Sony blocked new games to work on PS3s running Linux (I'm stuck in firmware 3.15). Take that, you greedy overlords!