Why Bulldozer's spotty performance is good news.

[Altered]

even if this isnt as good as expected, the price of the overall platform is cheaper. i think it will be pretty good in a few years as games and apps get more threaded

Im not sure you added in the added cost of Windows 8 to perform properly.

And this is where the BD issues arises. I dont think buying anything be it a car, boat, TV, or PC hardware to only get partial use of it that day you take it home is what any consumer considers a deal. And I sure dont see anyway its justifiable for any company to ask you too pay, at top dollar, for something that is performing below the other offerings in its price range on the hope something from an entire different company may allow it to run as it should. I can see enthusiast doing this but AMD or Intel do not stay in business on that slim % of buyers.

Hopefully a revision or the retail shows up stronger.

 

[Super XP]

Well Said......
http://www.rage3d.com/reviews/cpu/amd_fx_8150/index.php?p=10

Verdict? It's tempting to be disappointed and doom'n'gloom about FX's performance. The clean kill and win isn't here against Intel's top Sandy Bridge. But, AMD hasn't priced it against the 2600K, it's head to head with the 2500K, and in that context AMD's Scorpius platform is a solid buy. AMD's FX processor is enough to challenge Intel in a lot of areas, but AMD still has work to do on improving performance while they live in an x86 and low-thread count integer based workload world. The first generation Bulldozer's have slimmed down the 'big cores' as part of their ongoing strategy for shifting to APUs, and this has obvious consequences for performance. Despite that, there is a very real and impressive boost in performance for current real world applications vs. the previous generation. Well done, AMD.

 

[Inceptor]

As somebody who has owned both an i7 870 and a X4 975, I have not seen a big difference between the two. But if you were to talk about pure performance; then C2Q, i5, i7, Sandy they all beat the Phenom 2, and changes are most of them will beat BD, whether both chips are OCed or not.

Those two sentences contradict each other.
Anyway, I'm not a rabid fan of either company. Here's what I based my comments on:
http://www.anandtech.com/bench/Product/49?vs=80
Choose a Phenom II x4 965 or higher vs a C2Q Q6600 or higher. Those are all stock speeds, which is what I was referencing. I never said anything about overclock vs overclock. And I never said anything about clock for clock, since stock speeds are different for Intel and AMD.
And I never reference game benchmarks.

 

[Neuromancer]

Bulldozer just needs the right benchmarks to show off it's real performance. That said, AMD needs to release a revision ASAP to ensure it does what it was meant to do, perform like a Bulldozer. Hopefully they can iron out it's issues and get back into the game.

A large, heavy gas guzzling machine, that sucks at digging holes, but can tear down hills no problem.

Bulldozer is a perfect name for this chip.

The power consumption on this chip is absurd, if it contained a 6500 series GPU in it, I could see needing as much power as it does.

That said I see a lot of people saying it eats multithreaded apps up, and that not entirely true either.

Consumer level multithreaded goodness...

I thought that handbrake only uses up to 8 threads. Although I never had a 12 threaded CPU to test that out for myself And the incredible hexacore times from the 990X show that it probably was using all 12.


However it should be noted that AMD announced (subtly) a LONG time ago that bulldozer would have reduced performance per core/per clock compared to deneb/thuban/zosma.

What they actually released was a multithreaded testing scenario that showed the BD chips performing (making up the numbers ) like 80% faster than a four core deneb. They marketted it as a kudo, but the math showed lower performance per core.

 

[Inceptor]

If people buy a computer to last 2 years and most of them do then AMD makes a lot of sense MAYBE. The maybe part is the hard one to buy into unless you know how it will work with optimized software for it's design. All cpu's depend on this aspect to do their best.
So do you think the folks at AMD are that stupid?
...
Most computer users don't hack up there computers every 6 months, only the benchmark led fanatics do that an it's a very small percent of computer owners that are in this category.
...
This is really what it boils down to isn't it. What you see when you use the computer and what you THINK you should see because of the black and white world of benchmarks while most any current brand are so fast now it doesn't really matter what cpu your running because you can run any game, or app and at home and you would never know the difference between them until you looked inside.
This is the most funny and insane part of the computer benchmark world where it is more applicable for non standard software use then for home systems. Saying yeah, I bought this cpu because I only run 2 programs on it and it's twice as fast in doing it. That is funny. And not realistic.
It's exciting to take a look at new things except when you can't see all of what the new thing can do and a redesigned cpu can fit in this category.
Like watching a Ferrari being tested on a 100ft. chunk of roadway.

Very good points, and a perspective to which nearly everyone who posts on this forum is completely blind.
Most people hold onto their systems for 4 or 5 years, even some of the people who post on this forum. In some cases, the systems are even older. On that time scale, the 'original' BD would turn out to be acceptable even years down the line, just as the Core 2 Quads and Phenom IIs have been acceptable to many of the people in this forum, years after supposed obsolescence.
Don't think only about gaming and benchmarking, people. Because neither AMD or Intel is thinking about you as their main money maker and customer. They may allocate some marketing dollars towards you, but you don't pay their bills.

 

[thebluebumblebee]

I feel that to expect great things from BD in multi-threaded apps is some kind of hype. Folding@Home is a muli-threaded app and the BD only does about the same as the 2500K! Linky

 

[twilyth]

I feel that to expect great things from BD in multi-threaded apps is some kind of hype. Folding@Home is a muli-threaded app and the BD only does about the same as the 2500K! Linky

Ah, but the catch is interger apps.

 

[Steevo]

But BD isn't a server chip, it's a desktop chip. AMD has had server chips with loads of cores before BD, like the "Magny-Cours" chip. I think this is a moot point for the majority of users, since most software still isn't highly compatible with multithreading yet.

On multi user account servers you don't need the software to be multi threaded, you allocate resources to each terminal user. More cores or threads means more users effectively per server.

This is the only location that a OPTERON, not bulldozer, is a win.

Companies constrained by server space are using two and four way boards, not single socket. Adding two more under-performing cores that suck up power is not the option they want.

 

[repman244]

Adding two more under-performing cores that suck up power is not the option they want.

Now that you mention that, does anyone imagine what the power consumption is of an 8 module Opteron is :shadedshu

The power consumption could really be an issue (and a big one!), they really need to lower it if they plan on putting a GPU in there...

 

[Zubasa]

Now that you mention that, does anyone imagine what the power consumption is of an 8 module Opteron is :shadedshu

The power consumption could really be an issue (and a big one!), they really need to lower it if they plan on putting a GPU in there...

Actually there already is the Interlagos which is an 8-module 16-core Bulldozer.
But no it does not suck up a ton of power because Opterons are made for parallel processing power thus are clocked much slower than the Zambezi.
For example the Opteron 6276 is only clocked at 2.3Ghz.

 

[repman244]

Actually there already is the Interlagos which is an 8-module 16-core Bulldozer.
But no it does not suck up a ton of power because Opterons are made for parallel processing power thus are clocked much slower than the Zambezi.

I know there are.

It's not so much about the clocks, the whole chips is insufficient, you can't take that away. I bet that per core it still consumes more power than Magny.
And AFAIK the BD based Opterons also have the agressive turbo and have higher clocks than Magny.

 

[Zubasa]

I know there are.

It's not so much about the clocks, the whole chips is insufficient, you can't take that away. I bet that per core it still consumes more power than Magny.
And AFAIK the BD based Opterons also have the agressive turbo and have higher clocks than Magny.

Actually I am pretty sure that the Interlagos is much more efficient that you believe it is.
The reason that the Zambezi uses so much power is because it is clocked as high as it could (within TDP) to get some single threaded performance.

Edit: BTW the 45nm Magny Cours (Opteron 6100s) are actually clocked higher than the 32nm Interlagos.

 

[Bo$$]

Im not sure you added in the added cost of Windows 8 to perform properly.

And this is where the BD issues arises. I dont think buying anything be it a car, boat, TV, or PC hardware to only get partial use of it that day you take it home is what any consumer considers a deal. And I sure dont see anyway its justifiable for any company to ask you too pay, at top dollar, for something that is performing below the other offerings in its price range on the hope something from an entire different company may allow it to run as it should. I can see enthusiast doing this but AMD or Intel do not stay in business on that slim % of buyers.

Hopefully a revision or the retail shows up stronger.

well it is already no slouch in games, but it is keeping up with 2500k which is similarly priced, and out performs it in certain apps

 

[FordGT90Concept]

Bulldozer was intended to run at 4 GHz+ but because of the issues at Global Foundries, they can't get that high. Their low model number (8150) suggests that they anticipate getting the problem fixed so they'll have headroom for higher clocked processors down the road.

It is what it is but more will eventually come. The question is if the problems will be resolved before Ivy Bridge. Ivy Bridge will raise the bar another 10-20% and I'm not so sure fixing the foundry issues will afford that much of a gain. Raising the clockspeeds will only be able to net them about a 25% increase max from where they are now.

 

[Zubasa]

Bulldozer was intended to run at 4 GHz+ but because of the issues at Global Foundries, they can't get that high. Their low model number (8150) suggests that they anticipate getting the problem fixed so they'll have headroom for higher clocked processors down the road.

It is what it is but more will eventually come. The question is if the problems will be resolved before Ivy Bridge. Ivy Bridge will raise the bar another 10-20% and I'm not so sure fixing the foundry issues will afford that much of a gain. Raising the clockspeeds will only be able to net them about a 25% increase max from where they are now.

Here is the problem.
Performance per Watt goes down the shitter as they increase the clock speeds higher and higher.

 

[repman244]

Actually I am pretty sure that the Interlagos is much more efficient that you believe it is.
The reason that the Zambezi uses so much power is because it is clocked as high as it could (within TDP) to get some single threaded performance.

I know that it has to be efficient as much as possible.
But if we can believe the Opteron's "leaked" clockspeeds: http://www.xbitlabs.com/news/cpu/di...ck_Speeds_of_AMD_Opteron_Bulldozer_Chips.html
You can see that they are really pushing it.
Also I wonder how the performance is if you run multiple VM's on BD (let's assume you assign 1 core to each VM), there could be a performance hit due to resource sharing like we see it with Zambezi. I can't find any info about VM performance anywhere...
It's really a hit or miss with performance, we also must not forget that core per core BD based Opterons are slower than Magny Opterons (or if we compare a BD Zambezi core with Thuban's/Deneb's core at the same clock speed).

And I know for a fact that even with servers, nothing is always parallel...

I really hope that AMD can improve BD ASAP since with this kind of performance it's not really looking good, and I want AMD to stay competitive.

 

[Zubasa]

The Interlagos uses Turbo Core 2.0 which is nothing like the one on the Thuban, so I don't really see that thet are "pushing" the clock speed all that much in terms sacrificing efficiency.
The Magny is basically the same chip as the Thuban (2 in MCM), they only reason they didn't use Turbo Core on it is most likely because of how TC 1.0 increases the voltage on the entire chip (even the cores in idle).
Using Turbo Core 1.0 on a server CPU will basically be a disaster interms of power efficiency.

 

[repman244]

The Interlagos uses Turbo Core 2.0 which is nothing like the one on the Thuban, so I don't really see that thet are "pushing" the clock speed all that much in terms sacrificing efficiency.
The Magny is basically the same chip as the Thuban (2 in MCM), they only reason they didn't use Turbo Core on it is most likely because of how TC 1.0 increases the voltage on the entire chip (even the cores in idle).

If you look at the slide you will see a +500MHz turbo across all cores, stock is 2.3GHz. That is quite a high turbo for a server chip...

You cannot deny the fact that the power consumption for the % of the performance increase (and sometimes even a decrease) is actually horrible. I just hope it's the GF having issues with there chips and that the power requirements will be improved with new stepping.

Turbo core 1.0 wasn't really that good, like you said it had the problem of not being able to control individual cores voltage (+ it really depended on the board and the BIOS, it was all over the place). And you don't want to implement not so polished features in a server chip.

I guess we will need to wait for an Interlagos benchmark run. But IMO it won't shine all that much.

 

[Zubasa]

If you look at the slide you will see a +500MHz turbo across all cores, stock is 2.3GHz. That is quite a high turbo for a server chip...

You cannot deny the fact that the power consumption for the % of the performance increase (and sometimes even a decrease) is actually horrible. I just hope it's the GF having issues with there chips and that the power requirements will be improved with new stepping.

Turbo core 1.0 wasn't really that good, like you said it had the problem of not being able to control individual cores voltage (+ it really depended on the board and the BIOS, it was all over the place). And you don't want to implement not so polished features in a server chip.

I guess we will need to wait for an Interlagos benchmark run. But IMO it won't shine all that much.

Indeed that will depend on how well Intelagos performs in practice.
On the other hand, Bulldozer do have some aggressive power saving features to help with its power consumption.

If this site is to be trusted:
http://www.cpu-world.com/CPUs/Bulldozer/AMD-Opteron 6276.html
Interlagos will have a significantly lower TDP than Magny Cours.
(115W TDP/85W ACP vs 140W TDP/105W ACP.)

Also the fact that AMD stated the Interlagos will share the same platform with Mangy Cours means that they should be within a similar power envelope.

 

[repman244]

Indeed that will depend on how well Intelagos performs in practice.
On the other hand, Bulldozer do have some aggressive power saving features to help with its power consumption.

If this site is to be trusted:
http://www.cpu-world.com/CPUs/Bulldozer/AMD-Opteron 6276.html
Interlagos will have a significantly lower TDP than Magny Cours.
(115W TDP/85W ACP vs 140W TDP/115W ACP.)

Also the fact that AMD stated the Interlagos will share the same platform with Mangy Cours means that they should be within a similar power envelope.

Yes that is correct the G34 socket does have it's limits when it comes to power delivery.

Don't get me wrong though I do not say that BD is a total failure on all fronts. It is actually a really good platform for future and something AMD can build upon.
It's just that it's not consistently good (it's a beast in some cases but then in some others it falls behind) and AMD needs a consistently good chip, it doesn't have to be the fastest CPU out there. I hope they polish out the problems with Piledriver and better process from GF.

 

[FordGT90Concept]

Here is the problem.
Performance per Watt goes down the shitter as they increase the clock speeds higher and higher.

They might be running at higher volts than intended because of the fab issues. Higher voltage increases stability.

 

[Steevo]

They might be running at higher volts than intended because of the fab issues. Higher voltage increases stability.

I can't even pretend to care about their issues for a three year wait to release this.

I know they have had a hard time trying to come up with the next big thing, but they failed, and not just at one part. Everything but a option that is only used 10% of the tie was missed by this processor.

Power consumption? Fail
IPC? Fail
Clocks/Overclocks? Fail
Cost? Fail


The only market this is good for is servers, and not much else considering a dual socket with two four core prior generation Opterons offers more performance for less price.

 

[kid41212003]

AMD has put some head scratching cpu's out when there was not any present need for what they did. It started with the the 64 bit cpu when 64bit home OS software that wasn't out yet but is now. They just kept the 32bit part working well till it started to come together. Started adding more core's to their designs.

Kind of like making a product to influence the future. Intel said what and why bother and then copied AMD while still saying why bother until their duel sort of cores came out. Funny.

AMD's timing really sucks..

Agree.

 

[Champ]

a bit off topic guys, but is the Thuban, I believe it called, a true 6 core chip?

 

[twilyth]

a bit off topic guys, but is the Thuban, I believe it called, a true 6 core chip?

Yup. All on one die. No MCM.