Maybe since the cpu competition is dead they have started dumping money and research into other materials like fiber optic and Graphene to replace copper on the cpu.
6-8 physical cores
Higher clockspeed
Better IPC
Higher overclock potential
HT on all CPUs
Better IGP capable of crysis on medium/high
All CPU's unlocked
Lower load & idle power consumption
All for the low low price of £140!
This is what i expect from Haswell. Should my expectations be dashed i will look like this guy --> D:
Really ? Did they say 8 core,, I think haswell is only 6 core and 12 threads well the expensive version,, the rest are quads... 6 core will cost 600 dollars of not more.
What hype? AVX2 doubles the throughput per core and enables widespread use of the SPMD programming model. That's the same high throughput computing approach as GPGPU, except on the CPU cores so it does not suffer from any heterogeneous overhead! And TSX offers long awaited hardware transactional memory and lock elision support, which are considered indispensable for efficient multi-core scaling. No hype there either.lol, is possible to die by hype? BenchPress is near of doing it XD
Please connect the dots. The Intel paper mentions ispc, which recently got support for AVX2.dude...
there is nothing about AVX2, TSX in the link...
SIMD on CPU is not like GPGPU except for less complex algorithms. Both approaches have pros and cons. For example a Haswell core is still limited to 2 threads and a few YMM registers, while a GPU "core" (e.g. a SIMD array) can manage tens of threads w/ low overhead per single op while being able to keep ALUs busy most of the time with so many threads to choose from.What hype? AVX2 doubles the throughput per core and enables widespread use of the SPMD programming model. That's the same high throughput computing approach as GPGPU, except on the CPU cores so it does not suffer from any heterogeneous overhead! And TSX offers long awaited hardware transactional memory and lock elision support, which are considered indispensable for efficient multi-core scaling. No hype there.
What hype? AVX2 doubles the throughput per core and enables widespread use of the SPMD programming model. That's the same high throughput computing approach as GPGPU, except on the CPU cores so it does not suffer from any heterogeneous overhead! And TSX offers long awaited hardware transactional memory and lock elision support, which are considered indispensable for efficient multi-core scaling. No hype there either.
Please connect the dots. The Intel paper mentions ispc, which recently got support for AVX2.
That wasn't JF-AMD's job.i understand JF because that was his job...but...you?
what's the impetus for Intel to significantly improve performance in anything other than incremental steps?
AVX/AVX2 are in the same category so probably not. AVX3 for Skylake and Skymont might as they will start using 512-bit FPUs.Will Windows 7 require another patch to handle the AVX2 instructions similar to SP1 which was required for AVX?
AVX2 doubles Integer Vector ops but keeps Floating Point Vector ops the same.
It doesn't exactly provide double float throughput in real world. It would be around no speedup to 1.5x speed up in run of the mill code. The ability to shoot out more integer vector ops is more beneficial than having the ability to do a multiply and an add within the same operation.AVX2 brings FMA, which would double float throughput if we get 2 256-bit units.
Actually CPUs are more efficient at complex algorithms than GPUs...SIMD on CPU is not like GPGPU except for less complex algorithms.
Running many threads is not an advantage. Each thread only gets a tiny amount of cache space and the locality is poor due to the constant switching. Also, CPUs can use the L1 cache to store plenty of temporary variables so having few registers isn't an issue, while GPUs have to lower their thread count when running out of registers and thus their scheduling options diminish and they rapidly lose performance.Both approaches have pros and cons. For example a Haswell core is still limited to 2 threads and a few YMM registers, while a GPU "core" (e.g. a SIMD array) can manage tens of threads w/ low overhead per single op while being able to keep ALUs busy most of the time with so many threads to choose from.
Have any products with hardware ASF support been announced yet?Researchers found ~20% and higher improvements in typical algorithms when using simulations of ASF on K8/K10 Opterons.
No, absolutely not.Will Windows 7 require another patch to handle the AVX2 instructions similar to SP1 which was required for AVX?
Yes I'm sure. I don't work for Intel in any way.dude, are you sure that you don't work at intel?
That's nowhere near comparable. It is a known fact that AVX2 features 256-bit integer vector instructions, hence doubling the peak throughput. And Haswell will also get fused multiply-add instructions, which "significantly increases peak flops". In fact it's easy to deduce that floating-point throughput will double as well. And to top it off it's also a known fact that AVX2 adds parallel memory access support (gather).you are remembering JF-AMD... IPC increases, IPC increases... AVX2 doubles, AVX2 doubles
. Meanwhile I'd challenge you to find any reputable person who thinks Haswell will not double peak vector throughput or thinks TSX isn't going to play a key role in multi-threaded scaling.
Yes I'm sure. I don't work for Intel in any way.
In fact I don't care about companies or brands. I only care about technology. If AMD was advancing DLP extraction on the CPU or was launching products with ASF support, I would praise them for that. Unfortunately they are stagnating while Intel is pushing forward. So I have no choice but to talk about Intel's technology.
That's nowhere near comparable. It is a known fact that AVX2 features 256-bit integer vector instructions, hence doubling the peak throughput. And Haswell will also get fused multiply-add instructions, which "significantly increases peak flops". In fact it's easy to deduce that floating-point throughput will double as well. And to top it off it's also a known fact that AVX2 adds parallel memory access support (gather).
So please stop acting as if I'm making stuff up. Whether you like it or not, there's lots of evidence to support that Haswell will be quite revolutionary. In contrast, the AMD representative's claims of IPC increases were largely unfounded. It quickly became clear that single-threaded IPC could suffer, and if it didn't then there would be reduced utilization (read: waste) during multi-threaded operation. So claims of IPC increases received a lot of skepticism. Meanwhile I'd challenge you to find any reputable person who thinks Haswell will not double peak vector throughput or thinks TSX isn't going to play a key role in multi-threaded scaling.
What makes you think Skylake will have 512-bit execution units?AVX3 for Skylake and Skymont might as they will start using 512-bit FPUs.
So let me get this straight, you are going to wait another year, until intel releases a chip with FMA, then praise them for it. Meanwhile AMD will have had a chip out with FMA4 for 1 1/2 years, and one with FMA3 for well over 6 months....
To be fair to Intel, aren't they kind of stagnated by lack of competition and lack of programers taking advantage of their advanced technology?
I mean thank (whatever power you believe in) that they are pushing ahead with the AZ plant....they could easily just re-fresh the same crap at higher base clocks if they really wanted to.
there is never a lack of technology.
Lets see if we recall the technology burst from 2004->2005 just 1 year, has more then double the advances we have now.
Do we need that much power? some do, some dont.
But is that why you end up getting high tier enthusiast line?
Who here was on a EE cpu and "required" that power?
Most of us got EE because it was an upgrade to the previous EE with a noticable amount...
Oh and incase ur wondering... i graduated hexcores..
The hexcore i have is awesome, has run great, keeps me competitive at clocks for 2 yrs now!!!
2 FREAKEN YEARS!
Is it too much to ask for a 8 core version?
Werent we suposed to have 8 cores by now?
What the hell happened to moor's law?
Oh it broke the moment AMD ran from the field..
there is never a lack of technology.
Lets see if we recall the technology burst from 2004->2005 just 1 year, has more then double the advances we have now.
Do we need that much power? some do, some dont.
But is that why you end up getting high tier enthusiast line?
Who here was on a EE cpu and "required" that power?
Most of us got EE because it was an upgrade to the previous EE with a noticable amount...
Oh and incase ur wondering... i graduated hexcores..
The hexcore i have is awesome, has run great, keeps me competitive at clocks for 2 yrs now!!!
2 FREAKEN YEARS!
Is it too much to ask for a 8 core version?
Werent we suposed to have 8 cores by now?
What the hell happened to moor's law?
Oh it broke the moment AMD ran from the field..
You just can't stay on topic topic can you? Everything has to be AMD vs the world for you.
This might help you in your misunderstanding:
http://en.wikipedia.org/wiki/Moore's_law
no i understand what it means.
But can you really say we went up double since the highest line last year?