Microsoft Windows 7 patch for AMD FX (Bulldozer) CPU's

[dividebyzero]

As some of you are aware, Microsoft has initmated that there would be a scheduler patch to made available to better take advantage of AMD Bulldozer's CMT (Cluster Multi Threading) architecture.

Release notes (Microsoft Knowledge base) >>here<<
Hotfix >>here<<


Feedback from users at:
OCN
TPU

I'd suggest that FX users on TS might like to add their before-and-after scores to this thread...

 

[red1776]

I apperently was typing at the same time and posted this in another thread. I just talked to Mizzou who is working on it now. i will add my results as soon as I get my hardware moved.

 

[dividebyzero]

Ah, what a difference 8 minutes makes. Don't worry, I'm guessing we'll have a third thread to add to the mix soon, if you get my drift...
Maybe the admins will notice and delete/merge as applicable.

EDIT: Just read your thread. Interesting that the MS (I presume) quote in your thread attributes SMT to Bulldozer, rather than CMT....doesn't auger well does it?

 

[red1776]

Ah, what a difference 8 minutes makes. Don't worry, I'm guessing we'll have a third thread to add to the mix soon, if you get my drift...
Maybe the admins will notice and delete as applicable.

:haha: yeah, I'm drifted. i will just link mine to this one.

dividebyzero;1118122]Ah, what a difference 8 minutes makes.

Does that mean you're 8 minutes smarter than me?? ( there is a softball right down the middle for you!:haha

 

[dividebyzero]

oes that mean you're 8 minutes smarter than me?? ( there is a softball right down the middle for you!:haha

It means that it's 9 a.m. here. Bread mix* is churning through the Hobart (dough hook) for forty minutes, which means (my second) four double-shots of espesso courtesy of the restaurant and some speed surfing of the net.

* Cornbread, Ciabatta (toasted Walnut) and Focaccia (parmesan)

 

[red1776]

It means that it's 9 a.m. here. Bread mix* is churning through the Hobart (dough hook) for forty minutes, which means (my second) four double-shots of espesso courtesy of the restaurant and some speed surfing of the net.



* Cornbread, Ciabatta (toasted Walnut) and Focaccia (parmesan)

...and linked

mmmm...so let me ask you....'do you have any reservations?
and ...3...2...1...:haha:

**after this, let's do the 'dead parrot sketch' ROFL

EDIT: Just read your thread. Interesting that the MS (I presume) quote in your thread attributes SMT to Bulldozer, rather than CMT....doesn't auger well does it?

yup, I am choosing to see that as a generic oversight!

 

[dividebyzero]

mmmm...so let me ask you....'do you have any reservations?

Yup. An over reliance upon breads for the tasting/dipping platters doesn't make use of the seasonal opportunities for sweet breads (Panettone, Stollen, Challah) which I'm about to rectify

 

[red1776]

mmmm...so let me ask you....'do you have any reservations?
and ...3...2...1...

Yup. An over reliance upon breads for the tasting/dipping platters doesn't make use of the seasonal opportunities for sweet breads (Panettone, Stollen, Challah) which I'm about to rectify

http://instantrimshot.com/

well to change the markets demand, you will need some marketing staff. I happen to know there are a bunch looking for work right now.:haha:

http://instantrimshot.com/

 

[dividebyzero]

I just received an email back from Nigel & JF. This is their strategy for us...

I believe that going forward our emphasis should be a forward looking consumer enviroment that makes use of our breads strengths. We can assure the most future focussed patron no less than a 50% improvement in IPC (Ingredients per Croissant) over our previous flagship pastry.

 

[red1776]

i just received an email back from nigel & jf. This is their strategy for us...

quote: I believe that going forward our emphasis should be a forward looking consumer enviroment that makes use of our breads strengths. We can assure the most future focussed patron no less than a 50% improvement in ipc (ingredients per croissant) over our previous flagship pastry.

ROFL!

I completely agree with the strategy. it has been said more than once that I have very forward thinking buns as well.

 

[Arris]

Focaccia. I'd kill for some good focaccia around here. Recently baked, just drizzled with olive oil and a little salt. Mmmmmmmmmmmm.

Also I'd be interested to see the Bulldozer performance improvements.
Hope it isn't a half baked fix from Microsoft *groan*

 

[Arris]

C5!
Direct hit on Flagship Pastry.
D7!
You sunk my Olive and Herb Focaccia Submarine!

 

[Mizzou]

Hope it isn't a half baked fix from Microsoft *groan*

Let's just say that it probably needs to go back in the oven for a while

The good news is that the hotfix is a quick install and equally important, can be uninstalled without drama. The only real difference I've seen so far is that the memory rating went from 7.8 to 7.9 in the Windows Experience Index (same clocks).

With the hotfix installed you can quickly see a change in how the cores are being utilized, Nero Recode normally utilizes all eight cores at 60 to 70% utilization. With the hotfix only 5 of the 8 cores are consistenly loaded at around 75 to 80% with the other cores being utilized intermittently or not at all.

While the majority of AIDA64 benchmarks show a minor improvement (negligible) video transcoding times went up considerably. Not a game changer by any means.

 

[red1776]

************************The Hotfix was taken down by Microsoft today and a spokesperson for MS said that it was not supposed to be released as it was. The actual patch is a 2 part scheduling patch that will be ready for release soon. So ostensibly what we installed was only half of the equation, which makes sense given what types of programs it brought increases to , and those it didn't**********************

 

[dividebyzero]

HT4U have an analysis of the patch.
41 benchmarks : 14 improvements, 27 regressions under the patch.

Patch now pulled by Microsoft. From Tech Report :

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.

EDIT:

/ninja'ed by red !

 

[red1776]

EDIT:

/ninja'ed by red !

You are right. it must be the cooking. I was working on and set aside my injera/Teff mix when i saw that.:haha:

 

[dividebyzero]

Don't know about you, but this whole episode leaves me a little stunned.

It's bad enough that Bulldozer has been in development longer than Windows 7, and AMD wait until after the launch/reviews to approach MS about thread scheduling (mind boggling!...maybe people at AMD haven't moved over from XP...or ME ???)...but somehow manage to conspire with MS to put out a halfa##ed hotfix...I'm beginning to wonder if there is (or has been) any communication at all between Sunnyvale and Redmond. Was any testing done prior to release? -it wasn't a leak as it was publicised through it's own MS Kb page- Did AMD sign off on it ?
I'm picturing a call from MS at 7pm...Sunnyvale deserted except for a janitor who picks up the phone; "they all gone home...Microsoft?...we-l-l I'm not really the best person to as-...benefit the company you say?, well ok then, go for it...bye"

 

[red1776]

There was nothing that stunned or even surprised me about that. My first thought was that it wasn't officially posted/announced anywhere. And it did not come up when I ran a check for updates. i went back and looked one last time before I did the request for hotfix' thing because I gave it a 50/50 as to weather it would still be there.
As far as the MS getting free testing with the "leak" ...sure why not? everything else operates that way. I think there will be a bit of fan fare when the official 2 parter comes out. weather it is or not, this will/has been seen as a good will gesture from MS. (personally I think it's a relatively cheap preview testing for Windows 8) I thought it being released with none (fanfare) was suspicious to begin with.

 

[dividebyzero]

I doubt even when the finished patch is available that it would show in the Windows Update suggestions -even as optional- if the patch is CPU specific unless there was 0% chance of the fix having any adverse effects for any other processor aside from Bulldozer. All good if the fix is affected by FX series flags only, but with the vast majority of CPU's using 1C/1T or hyperthreading I could see how MS might limit the patch to a Kb link and some publicity since 1.Average Joe user might just check the box and install regardless, and 2. Optional Updates usually pertain to MS software. MS may not see it as their responsibility to optimize their scheduler to cover all processors Ex Post Facto, especially for a niche product.

 

[red1776]

Yeah maybe for a CPU, I know update 'see's' my peripherals and offers updates for them as optional. I was assuming an attempted install would say "req hardware not found" if john Doe tried it with a Phenom or i7.
I still think that MS has a lot to gain PR wise by trumpeting this a bit. The boards were littered with comments like "damn decent of MS, I might have to re-evaulte my thinking towards them" and sentiments of the like.
There was this website of an odd nature that popped up on oct 12th that was first to say what the problem was with BD, said from day one that there would be a patch in Dec, by MS, and that it would bring improvements up to 43%. It was up for a while, then brought down by ostensibly DHLS. Popped up again, was taken down again. This happened once more and is currently down. It was strangely spot on though, so I am looking forward with great interest to 'Part II' of this thing.

 

[dividebyzero]

I was figuring that MS would treat the CPU as it would any other integrated hardware. i.e. unless NOT updating will cause a conflict (graphics), then it would probably get the same treatment as controller hubs, sound, Gb LAN etc...that is to say, update at your leisure with said update requiring some input from the user to do so.
As for positive PR from publicizing the patch, it wouldn't surprise me if this whole episode eminated from MS. PR in this case is relative I think - miniscule pool of affected people (who should be largely "tech savvy" in any case), and a captive audience...what ya gonna do, game on Linux? Really wouldn't surprise me if MS (under-) delivered in any case; " Well, working the scheduling fix into the existing Windows 7 kernal is a very complicated blahblah..[ ]..but our all new Windows 8 ***SEE our FREE BETA DOWNLOAD HERE*** has built in compatibility PRE-ORDER NOW! etc, etc.."
"Up to 43%" must have been an extreme case I would think. There are already optimizations available in the wild (MS and Linux) that would suggest that number might be closer to the call sign of the NWO's Black Helicopters than attainable gain. I can't remember an instance where a CPU's performance gained 43% from a software revision offhand, and would have thought that if numbers like that were able to be demonstrated then the info would have gained traction on the net....surely Intel couldn't have gagged every BD related software developer?

AFAIW, people like Dave Kanter were talking about scheduling in Win7/Vista at launch if not before. A few people posting on forums (noteably XS, OCN, Anand etc) tried to raise the issue with John Fruehe and others only to be shot down or ignored until lighter threaded benches started surfacing, so I'm not entirely sold on the revelatory aspects of prospective patches. Would seem to be more a natural progression once BD was measured independently against Phenom II.

so I am looking forward with great interest to 'Part II' of this thing.

And so you should be. Any performance gain is a winner in this case.