Phynaz
Lifer
- Mar 13, 2006
- 10,140
- 819
- 126
I dare say Dirk got his priorities correct, the board is full of idiots looking for headlines such as "AMD in tablet/smartphone" for short-term stock boost.
The mixture of operation lengths was because I aligned instructions in groups of 4 to the 32-byte blocks. As there is always a delay clock between two uses of registers, dependencies, false or otherwise, shouldn't matter.It is not so good to use 32 Bit operations in 64 Bit code since that might cause false dependencies, though you spread the operations widly over registers.
Changed to 64 Bit only, no false dependencies, I also changed the immediate operations to register operations to not run into decoding issues by larger immediate values.
not sure if i rember its march and i believe a bo if that helps
Uhm, you sure know that B0 is newer than A1 ?i called my boss he said we got the wrong ones. he said as gimpr stated we will get a1 stepping and a new bios for the boards
Uhm, you sure know that B0 is newer than A1 ?
Hope you will get B1.
and we haven't seen whatever Dirk planned for AMD in 2012, 2013, 2014.
Well then try at least Native Specialist (AMD CPU Identification), available here:yes some confusion b1 is the correct stepping. bios for board is now inhouse
waitng for cpus. was told not to post some benches or clockspeeds.() I have an authority issue. I have a few i want to run.
Well then try at least Native Specialist (AMD CPU Identification), available here:
http://cbid.softnology.biz/
And remove the clock information with paint/photoshop.
Maybe that program is a bit better than CPU-Z.
Alternatively, try HWinfo32:
http://www.hwinfo.com/download32.html
The latest beta has some kind of Bulldozer support:
Improved support of AMD Bulldozer.
Uhm, you sure know that B0 is newer than A1 ?
Hope you will get B1.
You registered five years ago and just made your first two posts?
Is that a record?
You registered five years ago and just made your first two posts?
Is that a record?
You registered five years ago and just made your first two posts?
Is that a record?
Intel can force their way in, by making tiny low-power Atoms, while everyone else is stuck on larger processes w/o finfets. When everybody else catches up, they will be another one or two nodes ahead, again being able to make something smaller than everybody else.And so how is Intel any better?
That's not exactly what I mean (edited). A new Windows on ARM won't be Windows CE (rejoice!). But, how well will it compare to the x86 version? There isn't enough information to really know. In fact, there likely won't be enough to know until 6-12 months after it has been out, since 3rd-party software so defines the Windows world. Real Windows on ARM has promise, but is also quite a wildcard, as far as having comparable x86 and ARM hardware are concerned.Windows 8 has already been demoed on ARM.
i am bringing a gpu to work to run vantage. i dont know if nvidia has drivers that work with 2008r2.
Maybe it does not matter, but false dependencies prevent scheduler from pulling instructions ahead. If there is no false dependency it can pull them ahead using shadow registers. The problem is if the BD does not work exactly as you expect but it could do more ops/cycle you might have difficulties to observe this with that test. Especially as the starting point of the sequence maybe shifted (the alignment won't do that alone, it is just for keeping fetch and I-cache fine).The mixture of operation lengths was because I aligned instructions in groups of 4 to the 32-byte blocks. As there is always a delay clock between two uses of registers, dependencies, false or otherwise, shouldn't matter.
I was not to clear in this point, it is to reduce code length to ensure more code is in the decode windows. It's not about ability to decode immediates that is right.AMD processors have always been able to decode "enough" immediates, because the decoders are symmetric.
Windows 8 has already been demoed on ARM.
Martimus got was I was trying (failing) to communicate.
My point isn't to debate whether or not bulldozer is aimed squarely at server markets, my point was that the company itself appears to be intentionally directing itself away from that being their bread and butter.
And I agree, getting rid of Dirk is just ridiculous on so many levels if you've closely followed DEC, Dirk, and AMD at the technological level since the early nineties.
That they did to Dirk what needed to have been done to Hector is just appalling...or maybe not. We haven't seen Llano, we haven't seen Bulldozer, and we haven't seen whatever Dirk planned for AMD in 2012, 2013, 2014. Maybe things really were about to suck big-time and even a BoD that let Hector run the show for years could not stand idly by letting Dirk do whatever it is that he did and we have yet to realize.
This thread simply will not give up the ghost.
Not until Bulldozer is released, and even then there maybe be a couple of thread arcs that continue for a bit. Plus, in a year from now, someone who is out of the loop will post here...