AMD Is Restructuring Again, Losing 7% Of Employees

[nightmarex]

HSA, OpenCL and the like are only useful if the problem domain maps well to the architecture of a GPU. A lot o apps do of course but HSA won't solve all of AMD's problems. Hell AMD has sponsored or other wised help developers port their code bases to the GPUs they sell and that still hasn't moved sales forward.

In part AMD's problem are one of a negative impression in the marketplace. That impression isn't based on any real issues as their hardware is plenty fast enough for general usage. Frankly it isn't easy to remake your image in the minds of the buying public.

Honestly though I have to feel for Ms. Su as she gets thrusted into remaking the business soon after taking on the job of running the company. Seriously depressing to have to lay off 7% of your work force as your first major initiative. I have to wonder if she was moved into the position due to this very problem.

I don't think they will solve AMD problems but computers in general are pathetically impractical. I don't see 'c' compilers jumping to stack addresses xor'ing registers and using esp to speed things along I see generic code with many unnecessary ops. Interpreters are even worse. However, say I need a program that iterates cyphers, I made one and testing all the combinations within a constrain is something that can be scored individual and multi paralleled something a GPU shines at. However did I write it that way? No because writing for opencl kernels is fucking nasty. If there was a "send task" -> stupid worthless pointless bullshit existance operation without setting frame work when so much is still staged on high frequency low latency cpu's I would be happy. I see HSA as a step towards this, regardless of Intel or AMD I want progress. So personally The AMD vs Intel is all moot to me, but Intel don't seem to be driving innovation like AMD towards things that are useful past brute ICP/cycle efficiency (important but driving nails with a hammer is slower than a nail gun no matter how efficient the hammer). The world will wake up and look upon me and my ilk favorable until then we are asshole's and zealots who make little sense to the unlearned.... sorry if you're one of them. Keep using hammers as shovels and fly the Intel flag high.

[gutigen]

I don't think they will solve AMD problems but computers in general are pathetically impractical. I don't see 'c' compilers jumping to stack addresses xor'ing registers and using esp to speed things along I see generic code with many unnecessary ops. Interpreters are even worse. However, say I need a program that iterates cyphers, I made one and testing all the combinations within a constrain is something that can be scored individual and multi paralleled something a GPU shines at. However did I write it that way? No because writing for opencl kernels is fucking nasty. If there was a "send task" -> stupid worthless pointless bullshit existance operation without setting frame work when so much is still staged on high frequency low latency cpu's I would be happy. I see HSA as a step towards this, regardless of Intel or AMD I want progress. So personally The AMD vs Intel is all moot to me, but Intel don't seem to be driving innovation like AMD towards things that are useful past brute ICP/cycle efficiency (important but driving nails with a hammer is slower than a nail gun no matter how efficient the hammer). The world will wake up and look upon me and my ilk favorable until then we are asshole's and zealots who make little sense to the unlearned.... sorry if you're one of them. Keep using hammers as shovels and fly the Intel flag high.

You lost me at "stack addresses", but I support innovation with all my heart and I think that open source is a key for any kind of computing evolution. AMD happens to not only innovate at many fronts, but also supports open source unlike those assholes from Apple-like, walled garden green company or those blue ones who if not for EU and US anti-monopoly laws would just eat AMD with brute force and anticompetitive politics.

[profoundWHALE]

Which is no wonder. It's ridiculously difficult to buy AMD based servers.

HP: 3 out of 23 servers are using AMD processors.
Lenovo: not a single AMD offering.
Dell is 100 % Intel. Even the low performance servers use Intel Celeron (!!) and Pentium instead of much more powerful AMD processors for the same price.

This does not make any sense at all! But Intel is known not to play fair. They have been convicted many times of abusing monopoly power and there is no reason to believe that they've stopped this.

In conclusion: market share does not say anything at all about the viability of server processors.

When they do use AMD server CPUs they're usually the really cheap crap ones, which doesn't help AMD for brand perception.