Bulldozer now

erek

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
11,310
i know that some people suggest that the Phenom II X6 (Thuban) outperforms the Bulldozer, but is anyone willing to discuss Bulldozer now since some time has gone by? seems like the Bulldozer is performing a lot better with more up to date benchmarks programs as observed here @ http://www.anandtech.com/bench/Product/203?vs=434 ...also i've found more of a justification of the design of the Bulldozer architecture in that it was intended for large virtual machine environments @ http://www.informationweek.com/news/...sors/231903448 ...anyone care to comment? will things start coming to together after all these tweaks and the Windows 7 scheduler is patched up and more modern software/games are released?
 
All those things combined i dont see it passing sandy bridge. I am not saying i dont like bulldozer but it still needs a price cut to get my recommendation.

The fx 4100 needs to be $99
Fx 6100 $139
Fx 8120 $179 if you ask me...
 
Bulldozer is selling well, Bench marks look better in some ways, gamimg compatibilty issues have toned down (complaints anyway) and AMD have anounced end of life for Athlon II and Phenom II (except 960T). We will be using Bulldozer if using AMD.
There is an actualality here that may see many Bulldozer business/general use builds.
There does seem to be some discount pricing showing up.

Power consumption, gaming, and price/performance has a way to go.

Microsoft may/probably will find a way to to improve scheduler performance, a maybe at least time wise.

For now it seems a believer question, or completing prior investment. Bird in the hand stuff.

Basically good luck to us all. Best I can see.

I see the 4100 at $100.
 
with the caveat - on an Asus AM3 board (M4A89TD-PRO USB3) with a beta bios. Anyone get this to work?
 
It isn't worth it until it comes down in price to around the ~$170 level. As it stands, it underperforms against SB alternatives (and even the Thuban) and uses too much power as well as requiring a new platform.

Unless you're running programs that can use its new instruction sets and are heavily threaded then you may as well pass. If you are running said programs, then the 2600k is a better deal, uses less power while providing better single-threaded performance and the chipset offers more.

The windows 7 scheduler won't do much for performance unless it's pinned to spread the threads to modules rather than cores within modules, but from what AMD has stated they prefer the threads being shared within the modules rather than between them. The latter will give you maybe a couple % gains whereas the prior would net you better gains (but probably worse power consumption). That too depends on what the CPU is being tasked to do, because the chip itself is rather picky due to the shared resources design. Win7 scheduler fix is supposed to help with the random nature windows 7 currently assigns threads, namely low number of threads (as you go up in number after 4 it just fills itself in), but the win7 scheduler should help both AMD and Intel as well.

10% performance gains across the board is wishful thinking...

picc-morph.gif


http://techreport.com/articles.x/21865/2

Both Intel and AMD will benefit from an improved scheduler, but because IPC is so lower on Bulldozer chips (lower even than deneb/thuban), and even when a single core has access to the full module's resources, the gains shouldn't be that significant. Once you move past 4 threads you'll likely see diminishing returns on improvement past 4 threads because you'd be taxing each module with 2 threads and forcing the 2 cores to share within the module. If you were to disregard IPC you'd still get more performance past 4 threads because the Intel chips don't have extra cores behind those 4 extra threads, but since the 2600k is so far ahead in the benchmarks that are multi-threaded and single-threaded it really goes to show just how important IPC is regardless of core count.

TL/DR version

If you're trying to justify buying the chip: I'm sorry but you overpaid for an underperformer.
If you're thinking it'll get better with a new scheduler and beat out Intel and show its real ability: I'm sorry but it won't.
 
Last edited:
Bulldozer is not a complete failure. It does not use any more power than a 1366 system would. All and all its a decent CPU. If you dig into the architecture you'd see its strengths and understand its weaknesses better.
 
Poisoner said:
Bulldozer is not a complete failure. It does not use any more power than a 1366 system would. All and all its a decent CPU. If you dig into the architecture you'd see its strengths and understand its weaknesses better.
Click to expand...

that's exactly what i am trying to do. dig deeper into the architecture to understand and see it's strengths and weaknesses better.

care to elaborate on that further?
 
"I'm running a bulldozer." I do heavy video editing w/ AE, Premiere. I'd consider it if there were tangible gains - the fluidity is crucial when making editing decisions - but I haven't heard enough from the community to justify this.
 
Poisoner said:
Bulldozer is not a complete failure. It does not use any more power than a 1366 system would. All and all its a decent CPU. If you dig into the architecture you'd see its strengths and understand its weaknesses better.
Click to expand...
It uses as much power as a 1366 system, whose platform is several years old, and still doesn't come close to its performance. It uses a significant amount more power than current sandy bridge platforms which it was marketed as competing against in price and performance, and still doesn't come close. In many instances it is outclassed by AMD's own previous architecture and in very few desktop performance scenarios does it beat it.
From a desktop perspective, I believe it is a failure. Server side is a whole other story, but from what I've seen its not much better.
 
Last edited:
erek said:
that's exactly what i am trying to do. dig deeper into the architecture to understand and see it's strengths and weaknesses better.

care to elaborate on that further?
Click to expand...

Read these...

http://forums.anandtech.com/showthread.php?t=2210645

http://techreport.com/articles.x/21813/1

http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested
^^ really good reading that. Anand goes into great detail regarding the architecture and explains each benchmark and why/how he sees the results.

http://www.anandtech.com/show/4997/amd-working-on-bulldozer-b3-stepping
Steppings usually come at least 4 months after the initial release, but AMD has generally been very slow in releasing its steppings lately. If they're looking to address cache/clock/power consumption issues in a single go then the stepping will take far longer.
 
what people dont see yet and wont until windows 8 is that its really made for that OS. Teh multi-thread app's and such it will be on par with w\e Intel has out but win 7 doesnt know what to do with it. Argue ALL you want with MEH i dun care :D you'll see it be faster in win 8 then win 7 you heard it hear first :D
 
fightingfi said:
what people dont see yet and wont until windows 8 is that its really made for that OS. Teh multi-thread app's and such it will be on par with w\e Intel has out but win 7 doesnt know what to do with it. Argue ALL you want with MEH i dun care :D you'll see it be faster in win 8 then win 7 you heard it hear first :D
Click to expand...
Yes, it will be faster in Win8. Its doubtful it will be faster than the Intel chips it is competing with in Win8 either. You seem to forget that Intel cpu's will also be gaining performance with Win8 and that Win8 still won't be out for another year. Intel will leapfrog AMD yet again with Ivy Bridge and Ivy Bridge E by then.
 
In my testing, my FX4100 needed at least 500mhz advantage over my AMD 631 to match IPC for SINGLE threaded. In a quad threaded test it was unable to be in the same region as the 631 even with 800mhz advantage. I suspect a 1500 mhz advantage it would of been close at least.

I would of really liked to test it against a phenom II X3, but to be honest it cost $20 more then my 631 at the time of purchase. I always feel products should be compared against other products at their price point.
 
Last edited:
The Athlon 631 has stars-based cores on 32nm and has a 6-7% IPC advantage over the Phenom II (which is stars at 45nm). Throw in that Phenom has a ~5-10%ish IPC advantage over Bulldozer and that a 4 core BD sometimes behaves like a 2 core, that makes perfect sense. Bear in mind the athlon chip lacks L3 as well which, depending on the task, can increase performance by a few more percentage points.

The biggest thing wrong with Bulldozer chips is the price. Currently they're just not priced according to they're performance.

http://www.extremetech.com/computing/100583-analyzing-bulldozers-scaling-single-thread-performance
 
Last edited:
Not all of the motherboard manufacturers haven't fixed the Steam CEG issue. Asus and MSI still have the problem that I know of. Gigabyte has fixed it supposedly. So if you buy BD today you're limited to certain motherboard manufacturers.

And that's not working as intended to me.
 
ASUS has fixed the steam problem with the Crosshair V board with the beta bios 9921, there is a final on the FTP site that also has the fix's in it. It should be on the DL for the MB on ASUS's main site by Mon or Tues. Still no joy for the Sabertooth.
 
Well I have an unofficial bios altered by a 3rd party and a program that I have to run to get my Steam games to work. In Saint's Row the Third if I minimize the application to say check an instant message, when I bring SR3 back up it's stuttering at 1 fps. I have to minimize the application again and then launch a 2nd copy of the game. Then the first copy of the game works again.

Not a solution that I would present as a fix for a person that not a hardcore PC user. Until all of the motherboard manufacturers fix the issue then I wouldn't recommend Bulldozer to others. It's not that I hate AMD as I ordered this chip on launch night based on prior ownership of their products. It's just that Bulldozer isn't ready for the masses yet.
 
pelo said:
The Athlon 631 has stars-based cores on 32nm and has a 6-7% IPC advantage over the Phenom II (which is stars at 45nm).
Click to expand...

Not quite. Llano is based off the Athlon II x4 architecture, which is Stars without the L3 cache. See the %7 improvement here:

http://www.anandtech.com/show/4448/amd-llano-desktop-performance-preview/2

The Phenom II enjoys up to a %25-30 performance advantage over the Athlon II x4 in data-heavy operations (compare 920 to 630):

http://www.legitreviews.com/article/1073/6/

In games, the performance difference is typically %10-15:

http://www.techpowerup.com/reviews/AMD/Phenom_II_X4_840/9.html

Do not confuse Phenom with Athlon and Llano.

Phenom > Llano > Athlon
 
Last edited:
I went from a 1090t to a 8120 and couldn't be any happier. It overclocks like a dream. Boot and shutdown times have decreased. My unrar times have nearly been cut in half. Everything loads quicker. The system as a whole is more responsive. Overall I felt it was a good upgrade.
 
This processor multitask extremely well yet no real tests to confirm this has been done as far as I can see on the reviews. It is hard to recommend this processor when application failures (games on steam) is still occuring, no real talk from AMD on resolution, relying on unknown folks with modified bioses and CEG programs to correct the issue. I think much more performance will be seen as time goes on with Bulldozer but it would be nice to know when some of these issues will be solved or expected dates.
 
The athlon 631 is a Llano chip with a dead GPU-side that sits in an FM1 socket. What you're comparing is the old 45nm Athlon to the 45nm Phenom II, not a 32nm Athlon (Llano w/o the graphics) to a Phenom II.

In terms of IPC, oddly enough, it's Llano (or 631 Athlon), > Phenom II > old Athlon. Which makes sense because the Phenom II and older athlons are the same as the athlon 631 but the 631 is a die-shrink without the L3 cache.

The Athlon 631 chips are essentially Llano chips without the graphical portion. You're confusing the Athlon 630/640 with the Athlon 631. They're different chips manufactured at 45nm and 32nm respectively and fit into different socket motherboards entirely (am3/am3+ and FM1) but share the same core design (stars, which is the same as the Thubans/Deneb).
 
noko said:
This processor multitask extremely well yet no real tests to confirm this has been done as far as I can see on the reviews. It is hard to recommend this processor when application failures (games on steam) is still occuring, no real talk from AMD on resolution, relying on unknown folks with modified bioses and CEG programs to correct the issue. I think much more performance will be seen as time goes on with Bulldozer but it would be nice to know when some of these issues will be solved or expected dates.
Click to expand...

There is a thread dedicated to this topic, folks in that thread with gigabyte boards claim it has been completely fixed with bios f6e. Should only be a matter of time before bios update for the other motherboards.

edit: just noticed this info was already posted on first page, sorry for repost.
 
Its only a matter of time before BD is cheap. If Phenom II's are no longer being produced, AMD can't sell overpriced CPUs and still be the "budget" choice.
 
CaptNumbNutz said:
It uses as much power as a 1366 system, whose platform is several years old, and still doesn't come close to its performance. It uses a significant amount more power than current sandy bridge platforms which it was marketed as competing against in price and performance, and still doesn't come close. In many instances it is outclassed by AMD's own previous architecture and in very few desktop performance scenarios does it beat it.
From a desktop perspective, I believe it is a failure. Server side is a whole other story, but from what I've seen its not much better.
Click to expand...

But it does compete with 1366 systems, just like Deneb and Thuban do.
 
Poisoner said:
But it does compete with 1366 systems, just like Deneb and Thuban do.
Click to expand...

I don't think competing with an old dead platform from a couple years ago is something to gloat about. You don't compare a processor to previous processors or future processors, you compare them to what you can buy today with roughly the same amount of money. We're not time travelling here
 
pelo said:
I don't think competing with an old dead platform from a couple years ago is something to gloat about. You don't compare a processor to previous processors or future processors, you compare them to what you can buy today with roughly the same amount of money. We're not time travelling here
Click to expand...

Its not anymore dead than LGA 775 which still has many enthusiast users. As an owner of both 1156 and 1366 platforms, Sandy Bridge is nothing to write home about. It may overclock real high and real easy but that has already been done before.
 
pelo said:
I don't think competing with an old dead platform from a couple years ago is something to gloat about. You don't compare a processor to previous processors or future processors, you compare them to what you can buy today with roughly the same amount of money. We're not time travelling here
Click to expand...

I have no problem finding new 1366 CPUs and motherboards from major retailers. The Gulftown processors are a little harder to find, but they were always limited in stock.
 
Nside said:
I have no problem finding new 1366 CPUs and motherboards from major retailers. The Gulftown processors are a little harder to find, but they were always limited in stock.
Click to expand...

And I can buy a brand new 60" tube TV.

You have to compare it to the current crop, because you assume the current crop is better than the older stuff. BD at best is a sidegrade from a Thuban and more often than not a downgrade. Unless the price drops to under $200, and more specifically the ~$170-180 mark for the 8150, they're not worth the $$

By dead I mean you can get newer/better and there are no new mobos/chips being produced for that platform. Generally the fact that new chips aren't being released means it's a "dead platform." At least that's what the popular term is. Dead platform has nothing to do with supply and stock. Like I said, big tube TVs are a dead platform
 
pelo said:
And I can buy a brand new 60" tube TV.

You have to compare it to the current crop, because you assume the current crop is better than the older stuff. BD at best is a sidegrade from a Thuban and more often than not a downgrade. Unless the price drops to under $200, and more specifically the ~$170-180 mark for the 8150, they're not worth the $$

By dead I mean you can get newer/better and there are no new mobos/chips being produced for that platform. Generally the fact that new chips aren't being released means it's a "dead platform." At least that's what the popular term is. Dead platform has nothing to do with supply and stock. Like I said, big tube TVs are a dead platform
Click to expand...

Big Tube TV's have their benefits over LCD televisions.
 
Windows 8 is already out in beta form somewhere. Try it first, don't speculate.
 
pelo said:
In terms of IPC, oddly enough, it's Llano (or 631 Athlon), > Phenom II > old Athlon. Which makes sense because the Phenom II and older athlons are the same as the athlon 631 but the 631 is a die-shrink without the L3 cache.
Click to expand...

No, Phenom II is still faster in data-heavy multithreaded tasks. If you checked the links I provided, Phenom II gets %30 faster than Athlon II x4 in data-heavy tests and %10-15 faster in games. And the Llano at the SAME CLOCK SPEED is about %7 faster than an Athlon II x4.

Anyone who can do basic math knows: when one chip is %10 to %30 faster and another chip is only %7 faster THAN THE SAME ARCHITECTURE, it's pretty clear which one is the better chip in terms of real IPC. Thus, Phenom II > Llano > Athlon II clock-per-clock.

If Llano was %7 faster than a Phenom II LIKE YOU CLAIM, Anand would have compared those two instead; he's not an idiot.

Anyway, most of the Llano IPC improvements come from doubling the cache (1MB per-core versus 512K per-core). In tasks that are not data-starved, the speed of the 3850 and 635 are nearly the same, indicating it's just a cache upgrade producing the higher Cinebench and 7-zip scores. But it's still going to fall-short of a large shared 6MB L3 that the Phenom II boasts.
 
Last edited:
I love AMD but right now I cannot recommend BD to anyone. I recently bought an 8120 to replace a PII 555 since I already had an AM3+ board. This was despite people here trying to warn me not too buy one, but I didn't listen; $250 lesson learned.
I could have dealt with it if it was only an issue of not being as fast as an Intel platform, but the damn thing is so glitchy. I think I had most of the problems people were complaining about on this forum, plus I found some new ones :D

Last night I finally got my new 2600K up an running and I couldn't be happier. I've noticed a massive FPS increase in BF3 over the 8120, and I haven't even tried overclocking the 2600K yet. Meanwhile my 8120 and Sabertooth board are sitting on a table collecting dust. I don't even think I'll use it to build a HTPC; it's not worth the trouble.
 
Poisoner said:
Its not anymore dead than LGA 775 which still has many enthusiast users. As an owner of both 1156 and 1366 platforms, Sandy Bridge is nothing to write home about. It may overclock real high and real easy but that has already been done before.
Click to expand...

Funny thing, AMD has been on a drum-beating campaign before BD release, saying overclocking BD is gonna be even more of a incentive for "PC enthusiasts" to purchase it.... In the end, "PC enthusiast" are the once calling it a "turd" even though BD has a "World Record" in OC-ing.... Too bad it's got no performance to back it up, unlike SB....:rolleyes:
 
CaptNumbNutz said:
It uses as much power as a 1366 system, whose platform is several years old, and still doesn't come close to its performance. It uses a significant amount more power than current sandy bridge platforms which it was marketed as competing against in price and performance, and still doesn't come close. In many instances it is outclassed by AMD's own previous architecture and in very few desktop performance scenarios does it beat it.
From a desktop perspective, I believe it is a failure. Server side is a whole other story, but from what I've seen its not much better.
Click to expand...

Erm, have you seen the recent benchmarks? The Phenom II X6 goes toe to toe against the i7 965 with several benchmarks, specially on multi threaded scenarios. Considering that Bulldozer tends to outperform often the Phenom II X6, it isn't a bad CPU by any means. Even Battlefield 3 runs better on Bulldozer than i7 2600K, the issue here is that simple threads will run slower, Bulldozer needs tweaks to increase IPC, but it isn't a bad CPU for any means. I see that moving from Phenom II X6 to Bulldozer as bad as moving from i7 965 to i7 2600K. Not enough noticeable gains to justify the expenditure unless if you are a heavy multithreading whore.
 
CaptNumbNutz said:
Yes, it will be faster in Win8. Its doubtful it will be faster than the Intel chips it is competing with in Win8 either. You seem to forget that Intel cpu's will also be gaining performance with Win8 and that Win8 still won't be out for another year. Intel will leapfrog AMD yet again with Ivy Bridge and Ivy Bridge E by then.
Click to expand...

Actually on Anandtech's review, i7 incurred in a small performance hit under Windows 8, kinda odd.

defaultluser said:
Not quite. Llano is based off the Athlon II x4 architecture, which is Stars without the L3 cache. See the %7 improvement here:

http://www.anandtech.com/show/4448/amd-llano-desktop-performance-preview/2

The Phenom II enjoys up to a %25-30 performance advantage over the Athlon II x4 in data-heavy operations (compare 920 to 630):

http://www.legitreviews.com/article/1073/6/

In games, the performance difference is typically %10-15:

http://www.techpowerup.com/reviews/AMD/Phenom_II_X4_840/9.html

Do not confuse Phenom with Athlon and Llano.
Click to expand...

You are right. According to AMD and tests, Llano has several tweaks that improve the IPC performance over Stars by up to 7%. But because of its lack of L3 cache, it might not outperform Deneb when is cache starved, but it will outperform its similarly clocked Athlon II X4 in all situations.
 
Last edited:
You must log in or register to reply here.
Back
Top