1000 core cpu http://uk.news.yahoo.com/21/20101228/tsc-new-ultra-fast-processor-unveiled-4b158bc.html Some how i don't see this getting to the retail market place for a very long time thought given the cores need pre-programming but we could be hopeful they develop a consumer system for it?
Not sure they will be anymore expensive than todays prices if they come to the market in a consumer form, the same could have been said if 10 years ago when we had pentium 1s about todays cpus...the prices will always be comparable to the price of today i would imagine as it was comparable to the chips 10 years ago.
And here is a weblink: http://www.tomshardware.com/news/ibm-transistor-graphene-100ghz-material,9611.html
All these things don't really help in a standard home-PC-environment. The limiting factors are the connections and data-transfer-rates between the different parts of a standard PC.
Not really. That can be improved if need be, I mean just look how SATA 6Gb/s magically appeared when SSD's arrived, and thats already old tech due to OCZs HSDL... It's the CPU's performance that needs to ramp up that triggers the rest. You can always bust bold on more PCI-E lanes/busses, just look at the N200 chips on the SR-2...
We're talking of CPU-RAM bandwidth today of what? 25GB/s. This is allready limited by the controller, the latency of the RAM etc, and the CPU itself could deliver even more actually, if it wasn't limited by the bus it uses to connect to the other parts. However, even those 25GB/s is allready 20 times as fast as the HDSL-stuff shown by OCZ.
But as soon as DDR3 is 'too slow' we'll move to DDR5 I bet, which would be a massive speed bump. And HSDL is the fastest cable connection standard out there currently... All you'd need to do to make it faster is to add more lanes to it...
Each lane is delivering a maximum of 500MB/s (PCIe 2.0) currently. Usually we're looking at 32 lanes alltogether, each lane consisting of a pair of "wires"... and these lanes have to be physically available, so it's not possible to just add as much lanes as needed, as you simply have'nt enough space to route and print them onto the motherboard etc. Nevertheless... let's look at the currently fastest HDD-solution. The Revodrive x2 uses a x4 PCIe 2.0 interface, which would allow for 2GB/s according to the PCIe-standards. It does deliver way less however. Not even half that speed actally, if you look at the reviews, so there's other limiting factors inplace, that have to be dealt with before we're looking to increase the speed of anything else. The upcomiong PCIe 3.0 will be offering 1GB/s per lane. So now with 32 lanes available, we would have the possibility allready for high-speed HDDs. 16 lanes for the GPU, 1x audio, 1x ethernet, a few here nad there, so laet's say we have 8 lanes left for storage. We could have 4x 2GB/s connections for HDDs available, if the HDDS would actually deliver that bandwidth, but even the currently fastest available storage is far from that. We don't even use as much speed as possible currently. No sir, the CPU isn't the limiting factor at all here at the moment.
Then if that is the case why are people getting excited about the new generation of Intel CPS's? As going by what your saying they wouldn't have an effect on the performance of a system...
I can't wait until motherboards will have optical interconnection between the CPU, memory and other parts.
don't forget AMD's last generation graphics card 5870 had a wooping 1600 cores, each with its own decode-execute. and nvidia' current Fermi have 512 cores of available GPGPU. it's nothing special. unless that's 1000 x86 processors, which it isn't. FPGA in consumer products will be interesting. they are not excessively expensive. £2000 for a top of the range dev board with most bell and whistles. a Virtex 5 only cost about £30 per chip. it's just currently, ASIC approach are much, much faster with less power requirement than any FPGA. waiting until we reach size limit? FPGA and ASIC are on exactly same size, ASIC will always be faster, FPGA have a lot of overhead, thus making it slow and power hungry.
CPUs can allways be improoved in specific areas, but for most areas the only thing that will speed up things is an increased frequency per core and higher speed of interconnections. Basically every software not made for multithreading doesn't make any use of additional cores. The new SandyBridge parts aren't exactly more powerful then the currently avaiblabe high-end i7-CPUs, but they're more efficient per clock. The i7 2600 for example is pretty much comparable to the i7 970, but it has 30 Watt less TDP with a not too shabby IGP ontop and it only costs half the price aswell. Additionally, you can easily overclock an $300 i7 2600k to 4.0 GHz on air-cooling, which makes it competitive to an $1000 i7 980x There's a reason, why we don't see more cores on CPUs allthough it's technically possible. the Cell-processor from IBM has 8 cores running for quiet some time now allready, the new server-processors from AMD have 12 cores even, but they don't speed things up, as the software is single-threaded most of the time. To improove performance further we first need to improove the speed of storage to feed more data per second, then we need to improove software to make use of the allready available cores and then, sometime in the not so near future we can think about adding more cores for CPUs used at home by Billy and Bob.
I think you may have got confused somewhere, I didn't say I wanted more cores, I want a 100GHz chip... Using Graphene transistors it is possible, and will allow a speed bump for all aspects of the system, not just the CPU...
