Saturday, December 17th 2011
AMD Bulldozer Threading Hotfix Pulled
Since we reported on the AMD Bulldozer hotfix, The Tech Report reports in an updated post, that the Bulldozer threading hotfix said to improve performance of the processor, has been pulled:
We've spoken with an industry source familiar with this situation, and it appears the release of this hotfix was either inadvertent, premature, or both. There is indeed a Bulldozer threading patch for Windows in the works, but it should come in two parts, not just one. The patch that was briefly released is only one portion of the total solution, and it may very well reduce performance if used on its own. We're hearing the full Windows update for Bulldozer performance optimization is scheduled for release in Q1 of 2012. For now, Bulldozer owners, the best thing to do is to sit tight and wait.It will be very interesting indeed to see how this much maligned processor benchmarks after the fully developed patch is released. It's true, actually attempting to download the hotfix and agreeing to the licence terms, at the moment, one is lead to a page that shows it as unavailable.
90 Comments on AMD Bulldozer Threading Hotfix Pulled
As the joker would say "why so serious".
Let me tell you how it is.
Intel dont need help.....you getting that?
AMD didnt really need it until windows 7 scheduling.
And I quote
"Windows 7 doesn’t understand Bulldozer’s resource allocation very well. Windows 7 “sees” eight independent CPU cores, despite the fact that each module shares scheduling and execution resources."
You starting to see the problem?
It's that Windows 7 currently knows perfectly well how to use Intel CPUs, but doesn't know how to use Bulldozer CPUs. It can end up sending two threads to the same module when other modules are available, resulting in performance drops most noticeable in software optimised for dual core CPUs.
You are basically bitching because Intel didn't fuck up.
I need to stop before I get in to it......run blibba.... run
Intel fanboys playing the victim game and saying shit here and there... srly? LOL! Buldozer will not outperform your chip, and W7 is optimized for SMT (intel design), so i don't see the point.
All i see in this news post is a benefict for AMD owners that MS will be releasing for W7 (wich all of us should be happy, not bashing each other)
Every new tech needs adoption, if no one ever optimizes things for new tech, we would still have single threaded apps and single-core CPU's worth a shit. Windows is optimized for SMT and now Buldozer brings up CMT, i don't see why not optimize it if they can.
The shared resource design is fundamentally flawed. If BD is allowed to be marketted as an 8-Core CPU, then the i7-2600k should be as well.Not sure about HTing, but I know the Athlon X2's had a Hot Fix from AMD because the chips internal schedulers were setup incorrectly for Windows. Basically since that was on a hardware level, the aptch reconfigured Windows to read the threads differently.Wow, I'm sold. Wait... no... that's retarded. If you are comparing a Phenom II X4 with a GTX580 to a i7-2600k with a 520GT, of course you're going to see the Penom II X4 perform better, with equal graphics cards, the i7 will beat the Phenom II pretty handily. And saying the CPU "means dick shit in games" proves that you have no fucking clue what you're talking about.
:roll:
:roll:
:roll:
I read these bulldozer threads for the sheer comedy.... However do understand the frustration/hope of those who bought the chip. Sad part is they'll probably get it optimized 2 days before piledrivers launch. :banghead:
To be honest it happens every time someone changes something in a way not expected. Sometimes it just needs software changes. Look at the Pentium 3 and the Pentium 4. When the p4 came out it performed more slowly than the p3 in most workloads unless you made use of SSE2 instructions. Once you did it flattened it. The p3-S 1.4 chip outperformed the p4 at some workloads until the p4 reached almost 2 ghz. This is an example of the software being "patched" or "coded to take advantage of the strengths of an architecture".
With HT there were issues when it first came out. The system would assign high priority threads to the logical core and the chip would stall out the logical core when it should have been running that thread. The system would sit in a "Waiting" period, sometimes for an actual noticeable amount of times if you had a poorly written .net program where the UI is tied into execution code or something of that nature, when this happened you could actually tell the program locked up momentarily. The patch helped aleviate this problem.
Athlon x2 had issues. A bigger issue was the first phenom TLB bug, this had fixes but the performance suffered. Phenom also had issues with the independant core clocking feature. Vista was pushing priority threads onto cores that had clocked down to 200mhz. There was a noticeable lag when you had to clock the cores up to full speed on a system critical thread.
TLDR
Change can be good but it also causes issues. If Windows is updated to the point where it can properly address bulldozers cores then this method is available to not only AMD but intel is well if they want to split cores down the middle instead of the 70/30 or 80/20 split hyper threading has now. This also paves the way for an OS aware of specialized cores. Lets say Intel or AMD actually wants to offload general computing functions to the GPU sections of their chips and the OS can do it automatically instead of having to need code compiled to do it in the first place?
Hyperthreading was designed to maximize cpu usage by always being able to feed the execution units. Instead of waiting to fully load up a new thread. Being able to switch immediately has its advantages. In the end however its a technology designed to feed a single "core". This is why I would consider SB i7 a quad instead of an 8 core chip.
In AMD's case its more like 2 cores joined at the hip, they share some resources but in actuality have 2 complete functioning pipelines and are able to complete certain operations without accessing the other "core". In intel's design unless theyve changed it recently you cant process the logical thread directly.
If Intel and AMD were monsters Intel would be a guy with 4 arms. AMD would be a guy with 4 arms, 2 heads and 1 head is disagreeing with the other one about what they should be holding onto.
Performance doesn't equal Core Amount
CPU A:
1 control unit, 512bit instruction bus, 512bit data bus, 512bit (integer) datapath with a 1024bit FPU on die in the core space
CPU B:
8 control units, 8 64bit instruction buses, 8 64bit data buses, 8 64bit integer data paths each of the 8 cores has one 128bit FPU on die in eight of the core spaces
In a well optimized benchmark for CPU A and CPU B..(It is SMP ready and CMP ready)
CPU A scores 1,000 pts
CPU B scores 1,000 pts
Following your definition of what a core is
CPU A is an eight core and CPU B is a single core
it see's HT'ing fine but cant see two SMT threads (in which each thread shares most of the hardware resources with the other thread).