AMD 'Bulldozer' gets an Update from Microsoft

[Chiropteran]

Pentium Pro was intended for the workstation market. AMD's FX line is intended for the rest of us, who actually run Windows 7 (well not me anyway). If Bulldozer were Opteron-only and then AMD came out with an optimized version of Bulldozer for the consumer market, then your objection would be valid. However, AMD didn't do that. It's the same architecture for Opteron and FX.

Why are you changing the question? I was responding to THIS:

So AMD makes a processor that doesnt perform at 100% on the most popular OS out there? And that seems ok to you?

So Intel makes a processor that doesnt (sic) perform at 100% on the most popular OS out there? And that seems ok to you?


Yes, it seems okay.

Everything else you bring up is irrelevant to the question. Also, factually false.

"The Pentium Pro is a sixth-generation x86 microprocessor developed and manufactured by Intel introduced in November 1, 1995 [1]. It introduced the P6 microarchitecture (sometime referred as i686) and was originally intended to replace the original Pentium in a full range of applications"

 

[IntelUser2000]

Umm, Intel did need "special microsoft updates" when they released their original HT CPUs.

It seems like such an obvious issue now, but AMD's talk that it was an 8 core CPU might have caused MS to not worry about such things. Now that its released and the ramifications of the decisions are revealed, they made a patch.

You ever heard of the Pentium Pro? It's performance in windows 98 was terrible for it's cost, but it did very well on NT and later windows 2000.

But that chip was expensive and intended for workstations in the beginning.

 

[Chiropteran]

But that chip was expensive and intended for workstations in the beginning.

I'm not sure what your point is, could you be more specific?

Edrick said "AMD makes a processor that doesnt perform at 100% on the most popular OS" "that seems ok to you?"

So I brought up another situation where intel did the same thing, which led to the Pentium 2, a highly successful CPU by any measure.

So yes, it does seem okay to me. The intent or price of the Pentium Pro really have no bearing at all on the discussion, as the point was that the CPU was optimized for an OS other than the the most popular available.

 

[Olikan]

http://www.overclock.net/t/1185039/the-1st-windows-7-scheduler-patch-results-are-in-have-fun

some nice gains in games...( almost 10% in SS3 at minimun fps )

 

[skipsneeky2]

Umm, Intel did need "special microsoft updates" when they released their original HT CPUs.

The only issue here is why AMD didn't work with Microsoft to get this released prior BD's launch. The scheduler is one of the fundamental functions of a modern OS, and while the changes needed for BD should have been relatively minor, AMD should have known that MS wasn't going to ship a hotfix to the scheduler without serious validation.

That was then this is now.

Intel still makes processors with ht that need no update....but a 8 core cpu needs a special update in 2011/2012 i would think we would be pass this the i7 980x has 12 threads and it did not need a update.

The update won't fix the power consumption that makes the processor a horrible choice for most.

Many people fixate a purchase based on power consumption thats why many people jumped on the sandy bridge train from their nehalems excellent ocing and cooler temps.

If they fix performance fine thats dandy but you still got a hot potatoe anyone with self dignity still won't buy it after all the harsh reviews.

The update would be like patching a leaking upper radiator hose on a overheating car with duct tape.

 

[Chiropteran]

i would think we would be pass this the i7 980x has 12 threads and it did not need a update.

Wrong. 980x required a BIOS update or it wouldn't even post on some boards.

http://www.tomshardware.com/forum/272938-30-core-980x-woes





I don't know why you think this is a bad thing. AMD FX CPUs work fine without the update, it isn't required, but it does give a nice free boost to the performance.

 

[Cerb]

The damage has already been done it still needs a second revision you don't see our intel processors and phenom 2 processors requiring special microsoft updates.

That has happened. In fact, it's not uncommon at all, even occurring well after launch. The difference with BD is that people are paying close attention, in vain hopes that it will somehow fix the CPU's performance ailments.

A CPU that provides good performance with suboptimal OS support will get a pass. HT P4s, Athlon64 X2s w/ C'n'Q enabled, Core 2 Duos, and Thubans I know have all gotten updates for performance and reliability issues specifically related to multithreading, and probably Phenom II X3/X4s, though I can't quickly find any right this second.

One I have personal experience with was Windows 7 freezing for a few seconds at a time while multitasking on a Core 2 Duo. C2Ds were far from new. MS released an update that fixed it, and life went on.

With BD, you've got a slower processor and OS performance issues. Once the OS issues are fixed, typical performance will still only be about as good as it is in Linux, where the single-threaded performance is still about the same as in Windows. If the typical single-threaded performance, compared to a Phenom II, had improved significantly, the patch would still be needed, and still would be made, but you wouldn't care. It would be gravy to enhance good mashed potatoes, instead of gravy to flavor KFC "mashed potatoes".

That was then this is now.

And, if Intel had introduced HT in their current generation of processors, instead of having done it a full decade ago, those new CPUs would need a performance update, today.

It seems like such an obvious issue now, but AMD's talk that it was an 8 core CPU might have caused MS to not worry about such things.

8-core or not, each group of two cores share cache. Regardless of what you call the *MT in use, sharing near cache between threads will make it have similar SMP scheduler issues to Hyperthreading. Even if MS had been working on such a patch prior to launch, they would still need to test it extensively on production CPUs, and they couldn't have gotten those much sooner than end users. Also, like HT, the fix should only be needed once, and should apply just as well to any future CPUs that are made the same way.

 

[IntelUser2000]

I'm not sure what your point is, could you be more specific?

Sure, Pentium Pro worked well in its intended segment, that's servers. Pentium II just added on to make it so its suitable for PCs too.

Interlagos performs similarly to Magny Cours for most non-bandwidth intensive applications, and Zambezi performs less while using more power.

 

[skipsneeky2]

Wrong. 980x required a BIOS update or it wouldn't even post on some boards.

http://www.tomshardware.com/forum/272938-30-core-980x-woes





I don't know why you think this is a bad thing. AMD FX CPUs work fine without the update, it isn't required, but it does give a nice free boost to the performance.

Bios update is different from a software update...i don't think the update for the bulldozer is a bad thing but its power consumption still is a big turn off.

But now if your in the market for a pc on a budget you got two choices...intel i3 or the over abundent fx cpus stores can't even give away going to frys 2 days ago when i asked about their fx4100 supply i got the words"we got a over abundent supply of them "

Now anyone with $200 or more for a cpu will obviously buy a i5 or even a i7...unless you own a dual core on the am3+ socket i doubt you would ever buy a bulldozer..the smart ones with phenom 2s aren't even giving them a second thought.

 

[Spjut]

They can take back Windows's fix for the TLB issue as well now when they are working on this

 

[Chiropteran]

But now if your in the market for a pc on a budget you got two choices...intel i3 or the over abundent fx cpus stores can't even give away going to frys 2 days ago when i asked about their fx4100 supply i got the words"we got a over abundent supply of them "

Newegg and amazon are both sold out of FX-8150s. I agree that the 4100 is kinda of a pointless CPU, given the existence of cheap phenom x6, but maybe this will change.

 

[jacktesterson]

10% performance boost is nothing to laugh about from a Patch.

 

[skipsneeky2]

Newegg and amazon are both sold out of FX-8150s. I agree that the 4100 is kinda of a pointless CPU, given the existence of cheap phenom x6, but maybe this will change.

The only people who prob bought the 8150 were the ones maybe on athlon quads and dual and tripple cores.

Or the 8150 is in low supply its very easy to sell out.

8150 is only worth it for those with no common sense or someone looking to replace a athlon or dual core...

 

[RavenSEAL]

So AMD makes a processor that is better than what current operating systems can handle? And that seems ok to you?

Fixed, so, yeah.

 

[jhu]

I'm not sure what your point is, could you be more specific?

Edrick said "AMD makes a processor that doesnt perform at 100% on the most popular OS" "that seems ok to you?"

So I brought up another situation where intel did the same thing, which led to the Pentium 2, a highly successful CPU by any measure.

So yes, it does seem okay to me. The intent or price of the Pentium Pro really have no bearing at all on the discussion, as the point was that the CPU was optimized for an OS other than the the most popular available.

Pentium Pro was intended for the workstation market. AMD's FX line is intended for the rest of us, who actually run Windows 7 (well not me anyway). If Bulldozer were Opteron-only and then AMD came out with an optimized version of Bulldozer for the consumer market, then your objection would be valid. However, AMD didn't do that. It's the same architecture for Opteron and FX.

Regardless of all this, a 4 module Bulldozer running 8 threads is still slower than a 4 core SB with HT running 8 threads.

 

[skipsneeky2]

Raven is the most hardcore amd user here i bet and i am sure he isn't even tempted for the 8150.

Hardcore user on a 1090t i am sure he is waiting for haswell the only choice that makes sense.

 

[jhu]

Newegg and amazon are both sold out of FX-8150s. I agree that the 4100 is kinda of a pointless CPU, given the existence of cheap phenom x6, but maybe this will change.

Sounds more like yield issues and they're turning off 2 modules.

 

[BD231]

Looks like the patch makes BD show up as 4C/8T in Windows now instead of 8C/8T. The marketing guys at AMD just cringed.

More like a collective sigh of relief for having avoided any serious lawsuit's over false advertising.

 

[jhu]

Raven is the most hardcore amd user here i bet and i am sure he isn't even tempted for the 8150.

Hardcore user on a 1090t i am sure he is waiting for haswell the only choice that makes sense.

For multi-threaded apps, the 8150 isn't bad and does beat the 1090t (8150 even has lower idle power usage than 1090t). The only problem is performance/price favors 1090t.

I bought my 1090t for $170. Newegg shows a price of $270 for fx-8150. Looking at povray benchmarks:

1090t: 3759 pps
8150: 4512 pps

However pps/$:

1090t: 22.1 pps/$
8150: 16.7 pps/$

I get 5 more pixels/second/$ with the old Thuban.

Just for comparison:
i7 2600k: ~4800pps
i7 2600k: ~15 pps/$ based on current Newegg prices

 

[Vesku]

Sure, Pentium Pro worked well in its intended segment, that's servers. Pentium II just added on to make it so its suitable for PCs too.

Interlagos performs similarly to Magny Cours for most non-bandwidth intensive applications, and Zambezi performs less while using more power.

Apparently Interlagos is showing promise for some server segments. Well enough that Intel felt they needed to start talking about their next family of server CPUs.

 

[pelov]

Single-threaded performance shouldn't change (1 per module is always 1 per module, regardless of which one it is), it's the 2+ threaded performance that should see some changes. If they have opted to go with 1 thread per module under 4, then the chips will see a decent performance increase.

The chip was never an 8 core to begin with. The cores share far too much within a single module for it to be considered separate. A glimpse at the within-module benchmarks in comparison to the thuban cores shows that

 

[RavenSEAL]

Raven is the most hardcore amd user here i bet and i am sure he isn't even tempted for the 8150.

Hardcore user on a 1090t i am sure he is waiting for haswell the only choice that makes sense.

I'm holding out for Piledriver and hoping it's a massive improvement over Bulldozer (fingers crossed).

If that doesn't work out, then yeah, Haswell will close the deal.

 

[skipsneeky2]

I'm holding out for Piledriver and hoping it's a massive improvement over Bulldozer (fingers crossed).

If that doesn't work out, then yeah, Haswell will close the deal.

Considering most people would consider the bulldozer as almost a downgrade over phenom 2 piledriver would have to be huge.

One can hope it delivers for reasoning of competition but perhaps a second revision of bulldozer will come along in between.

 

[Cerb]

Apparently Interlagos is showing promise for some server segments. Well enough that Intel felt they needed to start talking about their next family of server CPUs.

BD didn't fail absolutely everywhere. It just failed often enough to not be a good upgrade for the mass market, which needs high volume, which drives down costs, improving margins at whatever selling prices AMD is stuck with.

 

[frostedflakes]

For multi-threaded apps, the 8150 isn't bad and does beat the 1090t (8150 even has lower idle power usage than 1090t). The only problem is performance/price favors 1090t.

I bought my 1090t for $170. Newegg shows a price of $270 for fx-8150. Looking at povray benchmarks:

1090t: 3759 pps
8150: 4512 pps

However pps/$:

1090t: 22.1 pps/$
8150: 16.7 pps/$

I get 5 more pixels/second/$ with the old Thuban.

Just for comparison:
i7 2600k: ~4800pps
i7 2600k: ~15 pps/$ based on current Newegg prices

Part of that problem is low availability, though. MSRP for the 8150 is $245, but they're scarce so retailers like NewEgg up the price.

Also in newer versions of POV-Ray than the 3.7 beta 23 used in Anand's benchmark suite, the 8150 seems to perform on par with/slightly better than the 2600K. I know people scoff at this, but BD is a completely new architecture and software optimizations will improve performance. It's not wishful thinking, it's just reality. The Win 7 scheduler hotfix is a good start, and optimizations in software like POV-Ray, Cinebench, etc. should provide some gains as well.

http://www.overclock.net/t/1141562/practical-bulldozer-apps

Even taking all of this into account, though, BD still seems to offer slightly lower bang for the buck in that application than the 1090T.