AMD Zambezi news, info, fans !

[JF-AMD]

For most of those benchmarks,regardless if they are real or fake, the key thing is SIMD performance. Integer SIMD is handled also by the new FPU inside bulldozer. So this will be the key for good or bad reviews. If it doesn't beat K10 in single thread or poorly thread integer SIMD workloads,it may not get so favorable reviews. If it manages to do this it will be both faster and more efficient than K10,in both single thread and multi thread benchmarks. Now,how all these "leaks" correlate to real performance is anybody's guess. At least in single thread workloads,when both halves of the flexfp are working on one thread,Bulldozer should be faster than K10 at similar clock. We don't see this in any of the leaks. It is actually slower by 15-20%,which goes against what we know about its FPU.

I am gonna take the under on this one. FP benchmarks are less interesting except in the HPC world. ~90% of what your CPU is doing all day long is integer. FP has long latencies (deeper instructions) and is really only ~10% of the work. So speeding up FP might look awesome on SPEC FP benchmarks, but for desktop apps and most server apps, it gives you little or no noticeable impact.

I don't remember who mentioned it in the last few pages (lost the multi-quote I had going), but it was about making a 4M/4T chip and if AMD would do that... I don't know, it's unlikely, but about 1/2 way through the thread I was thinking how it'd be interesting for them to make one since it would give each core total access of all module resources. I think they'd make for pretty good gaming chips. Clock them at 4GHz with maybe a 4.5GHz 4C-Turbo, weee!

No, there is a 4-core 4 memory channel interlagos, but that is a server part. It has 2 modules. There is not that much of a benefit for having a single thread running on the module. The overhead from sharing is pretty low, so you get pretty close to the same performance. Running one thread on each module vs. filling the modules means that, to run the same # of threads, you will be doubling the amount of power required because you have fired up 2 modules. You'd be better off in most cases to run it on one module, power gate the other, and get the uplift from turbo because you have more headroom.

Your actual mileage will vary, but the idea that there is a huge overhead for sharing just isn't there. In most cases it will probalby be lightly impacting, if impacting at all.

What is this ages code all about? I've seen it referenced but what is it.

AGESA is the base code that is given to board developers and OEMs to build their BIOS. The version of AGESA that is used will determine what features are exposed and any performance optimizations that might be available. But just because the feature is in AGESA does not mean that your vendor will take advantage of it, you get the features that they expose to you.

So BLT has pulled the listings for the 8150/8120/6100 from there site.. Hmmmm

I said before and I will say it again. Public sales prior to launch are not allowed per the embargo guidelines. I can't speak to how the desktop guys would handle it, but back when I used to run business development, if someone was advertising prior to launch, they were not the first guys in line to receive product. Production parts went to the folks that honored the rules.

I think it will be more interesting to see "performance per Watt" instead of "clock vs clock".
Say you have a 1100T system @ 4GHz drawing 600W from the wall in full load, what frequency will a identical FX8150 system (with same components & cpu-cooler) end up with? 5GHz?
On top of this it might even run cooler, ending up with another win: a more quiet system...

Considering that ~5% of the market at most is buying top bin parts, you can assume that ~95% are interested in price/performance or performance/watt. Base on that, you are absolutely correct, this would be the most interesting thing to consider in comparisons.