Microsoft Hotfix For AMD "Bulldozer" Based Processors

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
From the front page:

Tis the season of resolutions! And a top resolution for 2012 is to get organized. For people, a central part of getting organized is learning how to improve scheduling your time to fit all the tasks at hand. Improving scheduling is not just a human thing – from an operating system (OS) viewpoint, better scheduling means that threads, processes or data flows are given more efficient access to system-level resources.

Our new AMD OpteronTM 6200 and 4200 Series processors are on the road to achieving optimal efficiency in 2012 with the recent release by our OS partners of improved scheduling techniques for their popular operating systems.

Let’s take a first look at Windows®. Currently, the CPU scheduling techniques used by Windows Server 2008 R2 SP1 are not optimized for the module architecture of the AMD Opteron 6200 and 4200 Series processors. This means that certain multi-threaded workloads will not be optimally distributed between cores, which can result in decreased system performance.

Microsoft has just posted a hotfix for Windows Server 2008 R2 SP1 documented in KB2645595 that enlightens the OS kernel to the AMD Opteron 6200 and 4200 Series topology so that the kernel can do a better job of scheduling threads to the processor cores. This patch leverages work that AMD and Microsoft have done cooperatively for future versions of the Windows kernel, and can be applied as desired by users running Windows Server 2008 R2 SP1. Below you’ll find links to the patches:

http://support.microsoft.com/kb/2646060

http://support.microsoft.com/kb/2645594

What are the performance gains that can be expected from this patch? It depends. Workloads that are moderately threaded have the potential for uplift. This could include virtualization, database, or transactional environments that are “bursty” – a mixture of light and heavy transactions, or legacy applications that are by nature not very threaded. The more heavily threaded the application, the less the likely the uplift. We recommend that interested users apply the patch to a test system and see how their workload responds.

If we switch over to the Linux world, Red Hat Enterprise Linux (RHEL) 6.2 was released on Dec 6, 2011. RHEL 6.2 is another OS that is “enlightened” for the AMD Opteron 6200 and 4200 Series processors. It delivers higher levels of efficiency realized through resource management and performance optimization and enables enhanced business agility through more flexibility for virtualized and clustered environments.

AMD works closely with Microsoft, Red Hat and the Linux Open Source Community to drive enhancements for our processors at the operating system level. Our resolution is to get users to evaluate our AMD Opteron 6200 and 4200 Series processors with these new “enlightened” OSs.
 
How about along with that, you show the #'s. Just saying that would be helpful to state your case AMD.

The Opteron's are not obviously intended for desktop use, but if this shines the turd up good on one end, hopefully the same changes could be done on the desktop with the FX line.
 
I was able to get the Core parking disable hotfix but I cant get the Scheduler fix
 
I was able to get the Core parking disable hotfix but I cant get the Scheduler fix

Just follow the directions and link for the HotFix. Once you request the HotFix, MS will send you an email with details and a download link.
 
Don't forget the Windows 7 users. This is for the AMD FX CPUs also.

http://blogs.amd.com/play/2012/01/1...-processor-using-windows®-7-scheduler-update/

Still not much of a difference. Even Win 8 beta shows at max 8% in certain apps/games. Still not enough for me to give up my 2600k/3930k for any AMD FX - Bulldozer system anytime soon.

In initial testing of the upcoming Windows 8 operating system, we’ve seen performance improvements of up to 10% in some applications, when compared to Windows 7.

:D
 
Just follow the directions and link for the HotFix. Once you request the HotFix, MS will send you an email with details and a download link.

That was the one with the request hotfix info, the other hotfix says its in windows update, which it is not.
 
One of the pages shows this link, the other one doesn't.

This is what it says on the 2645594 update page.
How to obtain this update
This update is available from the following Microsoft Update website:
http://update.microsoft.com
 
If you just look at the design there's no way they are going to be able to keep up with anything Intel has now or in the future as long as AMD sticks with this bulldozer crap. Back to Intel for the foreseeable future.
 
If you just look at the design there's no way they are going to be able to keep up with anything Intel has now or in the future as long as AMD sticks with this bulldozer crap. Back to Intel for the foreseeable future.

Go back to your Troll cave.
 
If you just look at the design there's no way they are going to be able to keep up with anything Intel has now or in the future as long as AMD sticks with this bulldozer crap. Back to Intel for the foreseeable future.

Says the guy with multiple advanced degrees in computer engineering and integrated circuit design. Oh, wait.
 
Go back to your Troll cave.

I think we all said it when Intel released the P4 initially. AMD all the way. Then, Intel ramped up the clock speed and it made a huge difference.

I'm not going AMD on Bulldozer because I'm happy with Intel right now. If AMD pulls ahead and really blows Intel out (like with Athlon 64), then I'll go AMD. I hold no allegiance to any brand - I go for the best performance.
 
precisely. Right now the base design is the issue. Intel hit the wall the p-4 in terms of power because the base design was a wreck. AMD is in that boat right now. For the troll commenter...you don't know me..luckily my clients do. Before you start name calling try asking why someone has a certain opinion instead of acting like you have a lack of critical thinking.
 
Good job BRO, lets just derail this thread into another "lets all go buy Intel" fiasco . Forget talking about the hot-fix or the issues downloading it, or personal benchmarks and performance gain. Stay constructive my friends, trolling irrelevant bullshit about how you don't like something gets old fast. Every time I come back to this thread hoping to find a direct download link to the scheduler hot-fix, or any other helpful information, I get to read about you guys bashing BD and your worthless opinions that nobody asked you for. If you guys want to run your mouths about how you don't like something, go start your own thread about it. Stop polluting everyone's threads with this garbage.
 
In initial testing of the upcoming Windows 8 operating system, we’ve seen performance improvements of up to 10% in some applications, when compared to Windows 7.

:D

Nice,

The max out of the 340~360+ apps tested over last 4 weeks was 8.4%. So rounded down to 8% What programs/games are you seeing a 10% increase over Win7?

I can put together another FX8150 PC going. I tore down the 2nd and 3rd FX 8150s where taken about after 3 weeks worth of testing.

I could not recommend a reason to upgrade from the PCs that where already deployed in the eng group. We saw a significant downgrade over our sample i5-2400/2500 workstations that where placed in April/May of 2011. Running mainly 3d modeling/cad/eng programs in eng group.
 
One of the pages shows this link, the other one doesn't.

This is what it says on the 2645594 update page.
How to obtain this update
This update is available from the following Microsoft Update website:
http://update.microsoft.com

Don't know why it's not on MS Update. I have had this hotfix since Dec 6th, 2011. You might also check MSDN or Technet forums for a link.
 
precisely. Right now the base design is the issue. Intel hit the wall the p-4 in terms of power because the base design was a wreck. AMD is in that boat right now. For the troll commenter...you don't know me..luckily my clients do. Before you start name calling try asking why someone has a certain opinion instead of acting like you have a lack of critical thinking.

We already all know that Bulldozer isn't as powerful as Sandy Bridge. AMD states that users that have intensive multi-threaded applications will see no benefits from this fix. This fix is for the rest of users to help squeeze out a small performance boost. I would rather that they did this instead of doing nothing at all, and it's good to see they're actively working on improving performance on their hardware.

Nowhere in this thread (yet) did anyone say this will magically make Bulldozer better than Sandy Bridge. Bringing that up is what makes you a troll.

Nice,

The max out of the 340~360+ apps tested over last 4 weeks was 8.4%. So rounded down to 8% What programs/games are you seeing a 10% increase over Win7?

I can put together another FX8150 PC going. I tore down the 2nd and 3rd FX 8150s where taken about after 3 weeks worth of testing.

I could not recommend a reason to upgrade from the PCs that where already deployed in the eng group. We saw a significant downgrade over our sample i5-2400/2500 workstations that where placed in April/May of 2011. Running mainly 3d modeling/cad/eng programs in eng group.

Average of 8.4% means that there were some applications that achieved a performance boost of greater than 8.4% and some less than 8.4%. That makes his claim of "up to 10% in certain applications" completely plausible.
 
I think since they point you to windows update the scheduler is an automatic download after the core parking fix is installed. It may take a reboot or two for the update to happen but it should eventually.

I also imagine it's the same file that was accidentally released a while ago that was taken offline.
 
I for one would totally laugh my ass off if they come out with 8 and suddenly BD's are faster then SB's. That would totally be sweet revenge to all the idiot Intel fanboys trolling the forums here constantly.
 
I for one would totally laugh my ass off if they come out with 8 and suddenly BD's are faster then SB's. That would totally be sweet revenge to all the idiot Intel fanboys trolling the forums here constantly.

Not going to happen. But it would be funny and a miracle IF it did, which it won't.
 
I for one would totally laugh my ass off if they come out with 8 and suddenly BD's are faster then SB's. That would totally be sweet revenge to all the idiot Intel fanboys trolling the forums here constantly.

I swear, it's like they work for Intel. I don't even bother clicking on the Intel forum, I don't own one so why should I care about it?
How could you waste so much time and energy on a forum that is completely irrelevant to you? I'm not on the Ferrari forums telling
people they are stupid because they could have bought a Corvette ZR1 for less money and similar performance. I just don't understand the mentality.
 
Last edited:
I swear, it's like they work for Intel. I don't even bother clicking on the Intel forum, I don't own one so why should I care about it?

A lot of the people you say work for Intel were former AMD fanboys. I certainly was. I have had my hand in purchasing more than 100 AMD CPUs over the last 12 years.

How could you waste so much time and energy on a forum that is completely irrelevant to you?

I am very concerned what AMD does. I would have purchased a BD instead of the i7 970 (that I paid $356.02 for new) if the BD was going to be better than the 970 enough for the potential added cost of BD over just dropping in a new cpu on my lga1366 system and selling of my i7 920 CPU.
 
Last edited:
I think since they point you to windows update the scheduler is an automatic download after the core parking fix is installed. It may take a reboot or two for the update to happen but it should eventually.

I also imagine it's the same file that was accidentally released a while ago that was taken offline.


Well, thats all good but you cant install the Core parking update unless you have KB2645594 installed, which is the link that says to go to update for.

Which specifically in the core parking fix says
Important If you apply this update, you cannot revert the settings by uninstalling this update. This update should only be installed on computers that have KB2645594 installed.

Which of course is This one which says is found in updates.

Guess I could just plop the old scheduler update and see if that works of course.
 
One of the pages shows this link, the other one doesn't.

This is what it says on the 2645594 update page.
How to obtain this update
This update is available from the following Microsoft Update website:
http://update.microsoft.com

same problem here. Got the KB2646060 but there's no link for the other one and it doesn't show up in windows updates, too.

jreinlie4 said:
I swear, it's like they work for Intel. I don't even bother clicking on the Intel forum, I don't own one so why should I care about it?
How could you waste so much time and energy on a forum that is completely irrelevant to you? I'm not on the Ferrari forums telling
people they are stupid because they could have bought a Corvette ZR1 for less money and similar performance. I just don't understand the mentality.

this ^^ - soooo much ^^ +1000

it's like as they'd have too much spare time as all their optimized benchmarks finished like a second faster...

nah j/k but you know what I mean. Seen this babble in every forum in the net from here to Azerbaijan and am very bored of it, too.

imagine, I just came here bc of this thread and: the same babble. my, if there wasn't anything more important. it's just *yawn* and a bit annoying, too.

Ppl like this should be banned from the respective subforums because they talk the same shit each and everyday not caring if anyone actually wants to read their redundant "information" (anybody already knows) or not...

An ignore filter in forums would be a blessing. Or an off-topic marker. It's srsly a PITA to read all that irrelevant bs over and over again.

So enough rant from my side ;)
let the mods do their jobs ;)
 
I swear, it's like they work for Intel. I don't even bother clicking on the Intel forum, I don't own one so why should I care about it?
How could you waste so much time and energy on a forum that is completely irrelevant to you? I'm not on the Ferrari forums telling
people they are stupid because they could have bought a Corvette ZR1 for less money and similar performance. I just don't understand the mentality.

I agree im new here and it's alreagy getting old,somebody ask a question about AMD this, all you get back is you should buy this intel.......on and on.....ok enough already is just a cpu not end of the world............
 
Give it up guys, as much as we wanted Bulldozer to be competitive, it just isn't so. You all act like the moon landing conspiracy theorists. lol ;)
 
Just wait for piledriver. It's basically the fixed version of bulldozer. These patches better be ready by then
 
Just wait for piledriver. It's basically the fixed version of bulldozer. These patches better be ready by then

I expect the patches to not be needed then. I mean when AMD fixes the slow high latency cache the impact from threads switching physical cores will be lessened.
 
Last edited:
I expect the patches to not be needed then. I mean when AMD fixes the slow high latency cache the impact from threads switching physical cores will be lessened.

I believe the main problem the patch is addressing is the fact that formerly, 2 programs would occupy 2 cores on the same Module, sharing the L2 cache of one module. With the patch, it allows the 2 programs to run on separate modules so they don't have to unnecessarily share the L2 cache of one module. There is a diagram that easily explains it, let me see if I can dig it up.




 
I'm seeing some pretty sizeable increases in matlab when using 8 threads for operations (well basically for >4 threads I'm seeing some 11%-18% performance over win 7 with no scheduler update this puts it about 4-6% faster than when I run the same things in windows 8 dev preview.

*Matlab R2007b win8 dev preview showed a large increase over 7 without the scheduler update (on another note, the old scheduler update that leaked out was slower than using 7 without the update)

**Edit two, had some of my run sizes off in the retests, numbers make more sense now
 
Last edited:
8% is the maximum increase but what was the AVERAGE performance delta? Somebody please tell me!
Also, does this hotfix shine particularly well in only certain areas of use? If so where is it worst and where is it best?
 
Has anyone/site posted a real life comparison benchmark of the hotfix?
 
Back
Top