AMD Bulldozer Registry Fix Coming - 40% Performance Boost

[Super XP]

Interesting. There is most definitely a Windows 7 AMD FX – software patch in the works.

In regards to the Bulldozer Design - The User Mode Scheduler is issuing blocking calls on the execution of threads. It’s not a kernel level scheduler problem. It’s the UMS. Ok… need a new cpu scheduling algorithm.

I asked some guys doing firmware hacks to check out the Bulldozer problem.
Good news one the coders that worked on the Panasonic GH1 camera hack has agreed to help write a registry patch.

AMD FX-8150 Passmark CPU Score has jumped from 8500 to 13000 using REG Patch. But it is not stable yet.

UPDATE:

The boys over at Kubuntu has been working on a AMD Bulldozer fix that sort-of applies across several platforms. The idea came from ArchLinux that run ARM multiple processors. AMD FX-Bulldozer is neither 4 core or 8 core… that’s the problem. The ARM Cortex-A9 8 core processor had a similar problem that was overcome with a software patch. The problem right now is that the patch has to be re-registered with each Windows start. Look for a 40% performance boost if this works…. more to come.

There is most definitely a Windows 7 AMD FX – software patch in the works. By most estimates the AMD Bulldozer FX is underperforming by 40-70% in most Windows 7 benchmarks. By forcing Windows 7 to recognize 8 cpu cores a huge performance hit has happened. The Bulldozer FX-8xxx design… really isn’t 8 cores, it’s a 4 core CPU with an extra integer pipeline on each core. If the FX-8xxx series scale according to the 4 and 6 core Bulldozer design than there is a serious bug in Windows 7 that is crippling the FX-8150 performance.

Preliminary data from Passmark CPU test show the FX-4100 with a score of 5900, and the FX-6100 with a score of 7700. The FX-8150 should be posting in the 11,400 area but several test have the FX-8150 down in 8000 score.

LINK:
http://quinetiam.com/?p=2356

 

[Irony]

I fully intend on getting a bulldozer whenever all of their bugs are worked out. It almost seems like AMD was shoving it out onto the market just so that they could have something to show for their deadline. I do hope AMD gets their act together soon, or else I might soon defect...:shadedshu

 

[pantherx12]

Good news everyone!

I hope this is true anyway lol

Getting a bulldozer chip anyways and if this is true then it actually makes it worth it ( Just getting it for play at the moment)



I imagine IPC will still suck though ( at-least match phenom maybe)

 

[Steevo]

Yep, AMD purposely didn't write a fix for the big new CPU so that users would experience less them expected performance. It is part of their plan to lose market share and get bought by Intel.


I would hope this was true, but it most likely isn't. And it won't increase performance 40%, testing on Windows 8 shows a whopping 10%, at best. and Windows 8 is aware of the cores/threads for scheduling cascading functions.

 

[pantherx12]

Yep, AMD purposely didn't write a fix for the big new CPU so that users would experience less them expected performance. It is part of their plan to lose market share and get bought by Intel.


I would hope this was true, but it most likely isn't. And it won't increase performance 40%, testing on Windows 8 shows a whopping 10%, at best. and Windows 8 is aware of the cores/threads for scheduling cascading functions.

I know it's only one benchmark "AMD FX-8150 Passmark CPU Score has jumped from 8500 to 13000 using REG Patch"

That's over 10% by a fair bit : ]

 

[lilhasselhoffer]

If, and I mean a very big if, this is true then I was wrong. BD has some magical power left to discover. Boy, I would love to be wrong.

At the same time, wouldn't AMD be pimping the work these people are doing if it were legit? Beating the 2500k would be excellent for everyone except Intel.


Whenever you can show me something more concrete than one test (unverified), and a link to a camera based fix, then I'll take note. For the time being, I'll wait quietly. Magical assumptions of 40-70% underperformance are exactly that, ungrounded assumptions.

 

[LAN_deRf_HA]

I don't really understand the issue here. Would this improve 8 threaded Intel performance too?

 

[streetfighter 2]

Desktops are dead - FTA

Wow, can't argue with top-notch investigative journalism like that. I think I'm gonna go get another beer to prepare for the FUD storm . . .

 

[Steevo]

AMD Dual core optimizer.

AMD can and will write the software if they found a use for it. Unfortunately either.


They don't since changing the threading procedure for Windows 7 will cause system instability or other issues.

They can't, and don't care about the users that purchase their products enough to hire someone else to do it.



If you will all remember the TLB bug, it was caused by almost the same thing (a thread schedule miss as it is waiting on information from the next scheduled thread behind it to be complete to use that information), so changing the way 7 interacts with the CPU will cause system instability as the thread scheduler will be given old and thus invalid handles. Even if this was a software patch that happened in the high level windows code it would still be hundreds of threads behind what was actually happening in the CPU cores. The fix for the TLB bug was to ignore the content in the TLB and take a 10% performance hit.

Code is fed to the CPU sequentially by named core/pipeline until a flush or cache miss occurs, the overhead of windows watching each thread execute on each core and make adjustments would bring a system to a grinding halt. Over 50% of the processor cycles and hundreds of lines of code would have to be in place to monitor it all.

The dual core optimizer just prevented executables from using standard timing RDTSC for scheduling threads and slots for threads.

 

[kid41212003]

If Steevo himself says Bulldozer is suck, then it's pretty much true.

No sarcasm intended.

 

[Goodman]

Yep, AMD purposely didn't write a fix for the big new CPU so that users would experience less them expected performance. It is part of their plan to lose market share and get bought by Intel.


I would hope this was true, but it most likely isn't. And it won't increase performance 40%, testing on Windows 8 shows a whopping 10%, at best. and Windows 8 is aware of the cores/threads for scheduling cascading functions.

I know it's only one benchmark "AMD FX-8150 Passmark CPU Score has jumped from 8500 to 13000 using REG Patch"

That's over 10% by a fair bit : ]

I agree with Steevo if there is to be any boost by windows reg fix or what ever it would be 10% at best or more likely 5-7% using Windows 8 which is Beta right now so pretty much a slim version of Windows , wait till it is retail with aero & all the other extras in it & you will see very little differences between Win7 & Win8... (performance wise)

 

[cdawall]

I want to see it...

 

[pantherx12]

I want to believe because it really isn't performing as it should with 8 "cores"

Single core performance I'm still believing will be rubbish.

 

[Baam]

Pretty sure he is fake. Trolling for page hits, he has been claiming the last couple days he has a registry fix but hasn't posted a download for it.

Bigfoot sighting on his page.

 

[LordJummy]

quinetiam, that name is familiar. Isn't that the name of the user who posted the same fanboyish post on a bunch of forums yesterday about how bulldozer crushes SB etc, etc?

Who is this guy again?

 

[newtekie1]

Do we have another AMD Dual-Core Optimizer fiasco.

And really, if this was true, AMD would have discovered it and released the patch in a nice permanant form when the processors launched, like the they did with the Dual-Core Optimizer. And if they didn't already discover this, then I fear for their ability as a processor designer.

 

[Neuromancer]

Been reading about htis for over 24 hours, and think it was even discussed here.

Check out the guys website, he makes me look sane.



I DO believe a patch could improve performance, I think it would be 10% like windows 8 is purported to improve. (benchmarks show most CPUs (not bulldozer) have 0% improvement in performance.


That said, any of you guys WITH a bulldozer care to experiment? Try enabling Thread ordering service. Please post your results

 

[Senupe]

Yep, AMD purposely didn't write a fix for the big new CPU so that users would experience less them expected performance. It is part of their plan to lose market share and get bought by Intel.
I would hope this was true, but it most likely isn't. And it won't increase performance 40%, testing on Windows 8 shows a whopping 10%, at best. and Windows 8 is aware of the cores/threads for scheduling cascading functions.

Yep, Winwdows 8 is aware of scheduling cascading but is not already finished in that way, the OS is going for Launch next year and the proccesor wasn't avaible when the developers version came out.
I just hope AMD get the right programming for this cause the poteintial is there; In Electronic Workbench you can simulate a circuit with similar config at a bulldozer module (shared resources) and the result it's about ~40% on improvement, i just wanna see the damn potential explioded.

 

[erocker]

No more Bulldozer threads.

Next time:

Use the search function. Find Bulldozer thread.

Post pertinent information in that existing thread.