So what went wrong with BD?

How long do you guys think till there is a price drop? A lower price can fix some of the unsatisfied reviewers, maybe?
 
If they can do anything about the cache latency and thrashing things might look up. AMD's next thing on Win 8, might do better too.

I'm not going to buy one, and some are buying cheap ones just to OC them.
 
If they can do anything about the cache latency and thrashing things might look up. AMD's next thing on Win 8, might do better too.

I'm not going to buy one, and some are buying cheap ones just to OC them.

bulldozer has been tested on windows 8 and its not much better.
win%208%20solidworks.png

win%208%20wow%201680.png
win%208%20wow%202560.png
 
One of your charts shows a 10% difference in favour of win.8. That's actually quite a BIG difference. Reference different charts next time.
 
It's actually... Not.

40 vs 44fps
60 vs 66fps
100 vs 110fps

That is often the difference between AMD BD and the often compared Intel counter parts. Ergo, your logic is Intel has a small lead on AMD.
 
And often times the difference is a lot more than that... With an i5, that has half the cores, and costs less. Here you're seeing a max difference of ~10% in a single graph

Ergo, your analysis regarding my logic is flawed.
 
It's actually... Not.

40 vs 44fps
60 vs 66fps
100 vs 110fps

world of Warcraft: Cataclysm 2560x1600 no AA

Win. 8 - 80fps
Win. 7 - 71.50fps

My maths makes that more than 10% difference.

Not sticking up for Bulldozer, not saying anything other than 10% is a rather a big difference to be called 'not much better'. MrBigshot should have picked some different charts.
 
There are quite a few benchmarks around right now for the FX processors. But even among those more variation than I would have expected. Not defending FX now because it's not what we wanted not even near it...but that is interesting anyway.

I would expect prices to fall pretty quickly I can't see why you'd pay £200+ for the 8 core FX when the i5 2500k is cheaper than that. If they pull prices down in line with that and drop the 6 and 4 cores down cash wise it might escape a thumping on many sites.

Yes performance is a problem, but the biggest problem is price v performance AMD are dreaming with these prices.
 
There are quite a few benchmarks around right now for the FX processors. But even among those more variation than I would have expected. Not defending FX now because it's not what we wanted not even near it...but that is interesting anyway.

I would expect prices to fall pretty quickly I can't see why you'd pay £200+ for the 8 core FX when the i5 2500k is cheaper than that. If they pull prices down in line with that and drop the 6 and 4 cores down cash wise it might escape a thumping on many sites.

Yes performance is a problem, but the biggest problem is price v performance AMD are dreaming with these prices.

RE: Variation of results, the BIOS situation is in alot of flux at the moment. New releases have been flying around even since launch. Another sore point of the BD launch then, BIOSs not really ready.
 
bulldozer has been tested on windows 8 and its not much better.
win%208%20solidworks.png

win%208%20wow%201680.png
win%208%20wow%202560.png

Here is what I am seeing in these charts.

BD improves under the Win 8 alpha.

SB gets worse under the Win 8 alpha.

Now, we know its an alpha build, so its not going to be mature, and performance is not eeked out of it yet. This is why we see a decline with SB. once the final version comes out I expect SB to do at least as well as it does under Win7 if not better.

What does this mean for BD? Well, seeing that SB isnt optimized, neither - probably - is BD. This means that BD will likely benefit even more than it does in these charts in the final version.

I'm guessing by as much as 15% in WOW, and probably by as much as 6% in solidworks.

From a OS standpoint, this difference is nothing to shrug at. It' won't suddenly make BD faster in single threaded operations, but it will narrow the gap a little bit.
 
Zarathustra[H];1037899586 said:
What does this mean for BD? Well, seeing that SB isnt optimized, neither - probably - is BD. This means that BD will likely benefit even more than it does in these charts in the final version.

I'm guessing by as much as 15% in WOW, and probably by as much as 6% in solidworks.

From a OS standpoint, this difference is nothing to shrug at. It' won't suddenly make BD faster in single threaded operations, but it will narrow the gap a little bit.

I was surprised that the BD launch didn't have an accompanying windows update to tweak the scheduler. Thought Microsoft would have played along tbh, obviously not.
 
world of Warcraft: Cataclysm 2560x1600 no AA

Win. 8 - 80fps
Win. 7 - 71.50fps

My maths makes that more than 10% difference.

Not sticking up for Bulldozer, not saying anything other than 10% is a rather a big difference to be called 'not much better'. MrBigshot should have picked some different charts.

The 10% claim was yours. I didn't do the math, I was just going by what you said in your previous post. I don't think 10% is a big difference at all personally.
 
What didnt go wrong should be the topic. overclocks meh, needs a nuclear plant, sux compared to last gen. It didnt get anything right.
 
Honestly, I see BD as a the stepping stone to what will come. I do think APU's are the future, although I think Llano did better then Bulldozer, I think bulldozer was basically an experimental product that really shouldn't have seen the light of day. But it was so hyped, for so long, that SOMETHING needed to be released.

I'm betting 2ish generations from now Intel will realize they backed the wrong horse with integrated graphics versus APU.

I think right now, Sandy Bridge (and Ivy Bridge) spank BD. But I'm betting the concept behind BD really shines once it's been revised some.

To answer the question originally posed:
Q: "What went wrong with BD?"
A: It's a complicated creature, that at it's base level is rather radically different then CPU+IG, and it had/is still having some teething problems.
 
I am really not sure that BD was a failure, per se. Sure, new builds may not go that route, but for alot of people that already have 890/990 boards, at least BD brings them within spitting distance of the Intel chips. For games, its really going to depend more on your gpu. For encoding, BD is comparable to the higher SB chips.

As an AMD fan, I would have obviously liked to see AMD come out with a knock out chip. Even without it dominating, BD is in the realm of 2500k/2600k at least, so that's nice to see. These chips are sold out on newegg, so I guess they aren't doing too poorly. I would like to get one just to see what I could get it to on an H50.
 
Well fx 8 core is close to the i5-2500k in some cases better but it's too mixed to really be conclusive.
Trouble is it costs more than the Intel chip so where's the deal here?

Forget about overclocking the performance is rubbish on the FX range you'll suck a load more power for not that much gain all the benchmarks show the same thing. The 2500 overclocks great.

This has the price of a top end processor without the performance. The design of the FX CPU is simply crap, it's flawed to hell. It's a P4 gone wrong ooops lets ramp up the clock speed to help. It didn't convince back then, it sure doesn't right now.

The 6 core is slower than the 6 core PhII the 8 core isn't even all that in some cases. It's a lousy upgrade for 4 core AMD users.
 
Well fx 8 core is close to the i5-2500k in some cases better but it's too mixed to really be conclusive.
Trouble is it costs more than the Intel chip so where's the deal here?

Forget about overclocking the performance is rubbish on the FX range you'll suck a load more power for not that much gain all the benchmarks show the same thing. The 2500 overclocks great.

This has the price of a top end processor without the performance. The design of the FX CPU is simply crap, it's flawed to hell. It's a P4 gone wrong ooops lets ramp up the clock speed to help. It didn't convince back then, it sure doesn't right now.

The 6 core is slower than the 6 core PhII the 8 core isn't even all that in some cases. It's a lousy upgrade for 4 core AMD users.


I definitely agree that the price premium is not warranted. keep in mind though, that it is still selling out- there is no reason for them to lower prices quite yet.

As for vs P4- thats not really an apt comparison. The p4 was absolutely destroyed back then. At least you can put the BD on the same charts at the 25/6k chips right now.

I personally would not upgrade a 955 to a BD, but for those who have the ability, BD takes them out of the i3-2100 range and into the 2500k range. I'd say that is a decent improvement, especially if they use their computer for any 'work' (encoding, editing, zipping, etc etc).

Back when I had a 1055t, I upgraded to the 1090T so I could get to a solid 4ghz for encoding. A few minutes here and there adds up if you have a lot to do.
 
A few benching results with cpu on subzero cooling

BD is a lot of fun, it will take you as high (Mhz) as your cpu cooling can (hold the heat load):)

screenshot018zd.jpg


screenshot007ik.jpg
 
Last edited:
A few benching results with cpu on subzero cooling

BD is a lot of fun, it will take you as high (Mhz) as your cpu cooling can (hold the heat load):)

And circuit breaker will allow. Wow, I wonder how much juice that's sucking up.
 
Last edited:
10% not a big difference? from a scheduler tweak?? that's a huge difference!

That's right... 10% is not a big difference. You can keep asking the same question and I'll keep giving you the same answer. It doesn't matter how you try and sugar coat it. 10% regardless of where they got it "from" is still 10%

It would be one thing if it got the gains it did in WoW across the board but it didn't so yes, again, not a big difference. When it can do this across multiple different tasks and not a one off example, then get excited.
 
Well, to high-end enthusiasts 10% seems like a small increment, but companies will spend billions of dollars to accomplish an increase of 10% in... anything!
 
Too many things went wrong. I think BD could be the next big thing, but it would require enough early adopters to show developers that coding for a seemingly infinite number of cores would be worth the trouble.

I honestly do blame the software developers to some extent. With a program where performance really matters, why would it be single-threaded in this day and age? How many generations of multi-core CPUs does it take for these developers to start using them properly? Windows scheduler wasn't built with BD in mind, but how difficult would it be to make an update for it? Right now, BD's tech with current software is like running a SSD on an ATA100 cable...

The power issue is another thing. Hopefully they work out the power consumption issues. I thought smaller tech would need less power. I looked at llano and figured (hoped) if they could run 4 32nm cores and a GPU with 100w, they might be able to run 8 32nm cores with 100w... Obviously the clock speed, cache, etc increased the power consumption, but I'm hoping it is actually a production quality issue, and perhaps later iterations (piledriver) will fair better....

I don't think BD is a total failure. While it doesn't seem to improve on the Phenom II, it does introduce a new architecture that could (<--key word) keep them going for several generations.

I will be skipping this first BD generation, perhaps even the entire AM3+ "chipset" and hold out for AM4... I'm not willing to drop several hundred dollars on something to tinker with before the more refined version comes out. I do, however, hope that this does open some eyes in the software community and gets developers interested in multiple threads... I just look in my programs directories and wonder why there are so many programs that install to the "x86" folder...
 
That's right... 10% is not a big difference. You can keep asking the same question and I'll keep giving you the same answer. It doesn't matter how you try and sugar coat it. 10% regardless of where they got it "from" is still 10%

It would be one thing if it got the gains it did in WoW across the board but it didn't so yes, again, not a big difference. When it can do this across multiple different tasks and not a one off example, then get excited.

You say 10% isn't a big deal but less then 10% is the difference between BD and i5/i7 at stock clocks in most games with a few really cpu intensive games be the exemption like sc2 and WoW. So yes "IF" we saw a 10% increase in performance across the board in games that would be a pretty big increase. Of course there is only 3 benches there and those 3 benches could easilly be nothing like how the rest would turn out but yes 10% could be pretty nice.
 
You say 10% isn't a big deal but less then 10% is the difference between BD and i5/i7 at stock clocks in most games with a few really cpu intensive games be the exemption like sc2 and WoW. So yes "IF" we saw a 10% increase in performance across the board in games that would be a pretty big increase. Of course there is only 3 benches there and those 3 benches could easilly be nothing like how the rest would turn out but yes 10% could be pretty nice.

Except that I actually went through all the benchmarks at anadtech and did the calculations. i5 won 18 tests, BD won 9. i5's average advantage on the tests it did better than was over 10% (not a whole lot more, but more) the i7 was quite a bit higher than that. I didn't actually go through the calculations for i7 2600+ but just by looking at how much better it did than the i5 was proof enough.

Not to mention i5 is outperforming it with 50% fewer cores and a lower price point. The simple truth of the matter is that in the useful lifespan of these two processors, applications aren't going to be threaded to hell and suddenly make BD equal or better than i5.

Windows 8 does not exist and windows 7 does. Highly threaded apps that it does well in are few. This "10% increase" a few people are so happy about is not only small, but meaningless. By the time it's a reality with Windows 8 both Sandy Bridge and BD will be old and irrelevant.

Bottom line, it's a shit cpu. Sorry but 10% on a OS that doesn't exist yet isn't going to save this processor especially considering that that will still only make it ALMOST as good as a cheaper CPU with half the cores and lower power consumption...
 
Windows 8 does not exist and windows 7 does. Highly threaded apps that it does well in are few. This "10% increase" a few people are so happy about is not only small, but meaningless. By the time it's a reality with Windows 8 both Sandy Bridge and BD will be old and irrelevant.

Windows 8 actually does exist. It's likely to be released sometime in 2012, hardly talking about years away. Not to mention SB and BD will hardly be old and irrelevant. You still have a q6600 which is what 5 years old. If BD will be old and irrelevant in a year how are you holding onto such ancient technology.

This "10% increase" a few people are so happy about is not only small, but meaningless.
Noticed you have your CPU overclocked, if 10% is meaningless mind telling me why you are trying to gain a few percent performance gains when it's "meaningless"?
 
Last edited:
Windows 8 actually does exist. It's likely to be released sometime in 2012, hardly talking about years away. Not to mention SB and BD will hardly be old and irrelevant.


Actually it's not the truth, it's your opinion. Hardly see how you can predict the next few years and what is going to be multithreaded. Hell games coming out now are getting better at using multiple threads.

You're right, I shouldn't have said truth, reality is much more accurate.

So 2012 is when BD is going to own eh? Ok, we'll see about that.

I guess I'm just not as impressed as you are by a processor that is so awesome and advanced that it performs like shit.

My CPU is over-clocked by 33% its' also free performance, not competing with anything else besides it's own stock clocks. BD is competing with another processor that performs better, costs less and uses less power. That was the most retarded argument I've ever heard anyone try and make. Comparing 33% to 10% in a scenario which isn't even in the same ballpark. Try again. You were much better off leaving the last edit out.
 
Last edited:
I guess I'm just not as impressed as you are by a processor that is so awesome and advanced that it performs like shit.

It doesn't perform like shit, what it does is that it doesn't live up to expected performance.

They need to focus on L1 cache speed
 
It doesn't perform like shit, what it does is that it doesn't live up to expected performance.

They need to focus on L1 cache speed

L1 copy and write speeds are truly abysmal you're right. I'd think this is where turning BD from a snazzy diagram into a physical device met with alot of problems. :(
 
NSide -

I'm a software developer for a leading manufacturer of Computer Automated Engineering software. Think Finite Element Analysis, Computational Fluid Dynamics, Thermodynamics, Aerospace, etc.

Not every problem can be written to run on multiple threads. It's not like we can just go in, flip a few compiler switches, and, "Blamo!", our code is multi threaded and scales to N number of cores. Some problems scale, some do not. Some problems are limited by memory bandwidth or disk I/O more than by the ability of the CPU to crunch the numbers. Some are GPU limited.

I suspect the same applies to the games you want to play and see scale perfectly on your BD, i7-980, or whatever.





Too many things went wrong. I think BD could be the next big thing, but it would require enough early adopters to show developers that coding for a seemingly infinite number of cores would be worth the trouble.

I honestly do blame the software developers to some extent. With a program where performance really matters, why would it be single-threaded in this day and age? How many generations of multi-core CPUs does it take for these developers to start using them properly? Windows scheduler wasn't built with BD in mind, but how difficult would it be to make an update for it? Right now, BD's tech with current software is like running a SSD on an ATA100 cable...

...
 
NSide -

I'm a software developer for a leading manufacturer of Computer Automated Engineering software. Think Finite Element Analysis, Computational Fluid Dynamics, Thermodynamics, Aerospace, etc.

Not every problem can be written to run on multiple threads. It's not like we can just go in, flip a few compiler switches, and, "Blamo!", our code is multi threaded and scales to N number of cores. Some problems scale, some do not. Some problems are limited by memory bandwidth or disk I/O more than by the ability of the CPU to crunch the numbers. Some are GPU limited.

I suspect the same applies to the games you want to play and see scale perfectly on your BD, i7-980, or whatever.


Fully agreed as a software developer who has written multithreaded code since the 1990s.
 
You're right, I shouldn't have said truth, reality is much more accurate.

So 2012 is when BD is going to own eh? Ok, we'll see about that.

I guess I'm just not as impressed as you are by a processor that is so awesome and advanced that it performs like shit.

I agree with both of you.

As multithreaded code gets better Bulldozer will seem like a better and better design. (though it is still puzzling that it performs so poorly in Civ5, which is a highly multithreaded game).

That being said, the move towards multithreaded code has been so slow, that IMHO, if you buy a new bulldozer today, you'll be ready for your next upgrade before good multithreaded code is the norm.

In other words, it will be a waste of a processor generation.

I'm still flirting with the idea of getting a BD. The price is steep for a marginal upgrade in most senses.

At 4Ghz my Thuban X6 performs similarly in single threaded apps to bulldozer at 4.6Ghz. So in most games, it will be more of a side-grade than an upgrade.

For heavily threaded apps, the extra two cores should be nice though.
 
NSide -

I'm a software developer for a leading manufacturer of Computer Automated Engineering software. Think Finite Element Analysis, Computational Fluid Dynamics, Thermodynamics, Aerospace, etc.

Not every problem can be written to run on multiple threads. It's not like we can just go in, flip a few compiler switches, and, "Blamo!", our code is multi threaded and scales to N number of cores. Some problems scale, some do not. Some problems are limited by memory bandwidth or disk I/O more than by the ability of the CPU to crunch the numbers. Some are GPU limited.

I suspect the same applies to the games you want to play and see scale perfectly on your BD, i7-980, or whatever.
\

Couldnt agree more. I work for a major financial and news outlet with a huge assortment of technologies and data centers around the world. 98% of our code is single threaded, and its due to very good realistic reasons. If AMD built a core that needs Multithreading to survive than AMD built a crappy core, thats the end of it. In 10 years all code will not be multi-threaded, and single thread performance will still matter to a very large deal of the market.
 
Maybe they will have better luck in the next revision.I think more cache would help.
 
Back
Top