Why is the response to Bulldozer so overwhelmingly negative?

[infoiltrator]

It is the Delorean of chips, great promise, much hype, long delayed, and underwhelming at present.
This does not say it is bad, just not what was expected, and so far not justifying cost vs performance.
You can analyze, theorize, research and reexamine. May be a few tweaks from greatness or an also ran.

 

[evident]

Because, like you, people are entitled to their opinion. And there are facts that say otherwise that this is not a trash CPU. You are looking for a reason to justify your rage. Others are trying to find the silver lining. And they have it on the workstation and server spaces despite anything you try to cram down their throat.

what am i cramming down people's throats?
i'm just stating what i see from the charts, and that is more power consumption, less performance, not that cheap, late to the party... what's the silver lining in this? everybody loses.

 

[hooflung]

what am i cramming down people's throats?
i'm just stating what i see from the charts, and that is more power consumption, less performance, not that cheap, late to the party... what's the silver lining in this? everybody loses.

Trash cpu is what subjective word you and your ilk are peddling.

Yet it is as good as anything out for general use and it excels in other areas and is built for the future which is coming despite what anyone says. Hell even Battlefield 3 proves it is coming for gaming.

No one standing up for the CPU is claiming its greatness or its hall of fame reservation. But it is far from Trash. Space heater would be more apt.

 

[Arg Clin]

LOL, AMD fanboys these days. Sandy Bridge is great when overclocking because performance/watt actually increases until the point most people get: 4.5-4.6GHz. The minimum additional voltage and the fact the die is so small means that little additional heat is produced, and performance increases so much while power consumption increases so little that efficiency is HIGHER at those speeds than running at stock.

So this time around you couldn't help yourself and pull the Fanboy card. I did wonder why you edited it out the last time. I'm quite confident to let whoever reads this debate make their own call as to anyone comming off as a fanboy - including comming off as a negative fanboy/hater.

Bulldozer doesn't "win some, lose some". In 80% programs it'll be outright slower, and in others it can match or beat the 2500K by a small margin. That you're able to find one perticular benchmark where it performs well and use it as "evidence" that it has its merits proves how much of a crappy CPU it is. It needs to be shown in a good light by people that tell you to forget about all other twenty benchmarks and focus on one exception instead.

We'll just have to disagree on that. I think it serves as an excellent evidence that the architecture can perform well when the software supports it well. Even if not a lot of such software exists yet. And do remember that this is NOT a buy/do not buy discussion.

Performance/watt is and has always been an extremely important part of computing, especially in servers. That's why Intel will still dominate there as well; they give everyone from small businesses to large enterprises much more performance at the target power they want to achieve, and the configurable TDP in Ivy Bridge is another idea that will cement their lead in that market.

Sure it is - to some people. and to others it means absolutely zip. You cherry picked this particular aspect to support your biased view against Bulldozer. I agree 100% that performance pr watt is better on Intel right now. And in cases where that matters, Bulldozer wouldn't be a good choice. But to claim that this particular aspect is the only success criteria for a CPU all of the sudden is quite one sided.

On desktops it's less relevant, but once you get to a certain point it's definitely something to take into account. It's not just the extra power consumption, but also the heat that it outputs. All your other components will run hotter because of it and it could quite literally become a room heater. Then there's the fact it's so expensive for the performance it gives, not to mention that it's an example of engineering laziness: you have a CPU with over two times the number of transistors and yet it's convincingly beaten, not to mention the other CPU was launched three quarters earlier.

So you'd call the i920 a "room heater"? http://www.pcper.com/reviews/Processors/AMD-FX-Processor-Review-Can-Bulldozer-Unearth-AMD-Victory/Power-Consumption-and-P (table 2)

Bulldozer holding a record for highest overclock means NOTHING. It does not apply to any real-world situation. It was done using extreme methods of cooling and the only reason why it got there is because the multiplier and base clock is completely unlocked unlike unlocked Sandy Bridge CPUs where you can only get a few MHz from BCLK overclocking and the Turbo functionality-enabled higher multipliers only let you go to 57x. The stupid CPU department at AMD thought it would be a good idea to make the CPU with a longer pipeline even though they should've seen from the beginning they weren't gonna hit their target 4.5GHz+ frequencies so BD could be competitive.

Goes to prove the point that the architechture has potential for high clockspeeds.

 

[Arg Clin]

This is similar to gtx 480's problem. BD uses WAY too much power for the performance that it gives. If an 8150 was 10-15% faster than 2600k then the power usage would still be an issue, but not a deal killer. For 10-15% SLOWER performance, however, you need to offer something better in another area; ie, lower power consumption.

I support your observation. AMD needs to address the power consumption issue.

But your comparison is interesting by the way. I wonder how many of the people who's primary occupation these days seems to be AMD/Bulldozer bashing would call the GTX480 a crappy product.

 

[Ferzerp]

GTX 480 is/was a huuuuge power hog, but it was also not slower than the competing offerings (usually it was slightly faster), and was by no means slower that the previous generation nvidia cards in any situation.

I don't see how that's a parallel you can try to draw. A more correct parallel would be if it were power hungry, and traded performance wins with the GTX 280, while being dominated by the AMD 5870 except for 1 or 2 niche applications.

 

[LOL_Wut_Axel]

So this time around you couldn't help yourself and pull the Fanboy card. I did wonder why you edited it out the last time. I'm quite confident to let whoever reads this debate make their own call as to anyone comming off as a fanboy - including comming off as a negative fanboy/hater.

We'll just have to disagree on that. I think it serves as an excellent evidence that the architecture can perform well when the software supports it well. Even if not a lot of such software exists yet. And do remember that this is NOT a buy/do not buy discussion.


Sure it is - to some people. and to others it means absolutely zip. You cherry picked this particular aspect to support your biased view against Bulldozer. I agree 100% that performance pr watt is better on Intel right now. And in cases where that matters, Bulldozer wouldn't be a good choice. But to claim that this particular aspect is the only success criteria for a CPU all of the sudden is quite one sided.

So you'd call the i920 a "room heater"? http://www.pcper.com/reviews/Proces...r-Unearth-AMD-Victory/Power-Consumption-and-P (table 2)


Goes to prove the point that the architechture has potential for high clockspeeds.

There's no point in discussing with you, as you keep talking about this mythical "future" that doesn't exist. Even if future software three years from now is completely multi-threaded (hint: it won't be) it'll perform just slightly better than the 2500K while consuming 50% more power, which negates the small advantage to most people. There's no secret performance for AMD to reveal. The facts are here today, and it performs in multi-threaded comparably to the i5-2500K because Sandy Bridge has 50% higher IPC than Bulldozer.

You buy today's CPUs for today's performance, and comparing today's performance Bulldozer is junk in comparison; therefore, at the present, Bulldozer is junk. Even if everything would be multi-threaded it'd merely be a "meh" because of its power consumption and the fact it overclocks no better than the competition, not to mention the price is higher.

The i7-920 at 4GHz could also be a room heater, but guess what: it also had the performance to back it up. In 2008 it simply demolished everything else in the desktop market, whether it was single or multi-threaded workloads. It's still so good, in fact, that both at 4GHz in multi-threaded tasks this THREE-YEAR-OLD CPU should be comparable or faster than an 8-core Bulldozer.

And you being able to get over 8GHz out of a processor with LN2 isn't impressive and proves nothing about it in the real-world. It doesn't prove anything at all about it having good headroom with air cooling, and as we can see from the facts it overclocks no better than Sandy Bridge: both reach 4.5GHz on average. We don't know yet what Sandy Bridge can achieve with LN2 because, again, the multiplier is limited to 57x. With Sandy Bridge-E we'll be able to see.

 

[exar333]

There's no point in discussing with you, as you keep talking about this mythical "future" that doesn't exist. Even if future software three years from now is completely multi-threaded (hint: it won't be) it'll perform just slightly better than the 2500K while consuming 50% more power, which negates the small advantage to most people. There's no secret performance for AMD to reveal. The facts are here today, and it performs in multi-threaded comparably to the i5-2500K because Sandy Bridge has 50% higher IPC than Bulldozer.

You buy today's CPUs for today's performance, and comparing today's performance Bulldozer is junk in comparison; therefore, at the present, Bulldozer is junk. Even if everything would be multi-threaded it'd merely be a "meh" because of its power consumption and the fact it overclocks no better than the competition, not to mention the price is higher.

The i7-920 at 4GHz could also be a room heater, but guess what: it also had the performance to back it up. In 2008 it simply demolished everything else in the desktop market, whether it was single or multi-threaded workloads. It's still so good, in fact, that both at 4GHz in multi-threaded tasks this THREE-YEAR-OLD CPU should be comparable or faster than an 8-core Bulldozer.

And you being able to get over 8GHz out of a processor with LN2 isn't impressive and proves nothing about it in the real-world. It doesn't prove anything at all about it having good headroom with air cooling, and as we can see from the facts it overclocks no better than Sandy Bridge: both reach 4.5GHz on average. We don't know yet what Sandy Bridge can achieve with LN2 because, again, the multiplier is limited to 57x. With Sandy Bridge-E we'll be able to see.

Keep in mind that we are all comparing the AMD 8150 to the 2500k/2600k in most of these comparisons. Essentially this means we are comparing Intel's 'mainstream' platform with the best of the best from AMD. SB-E and IB-E will simply destroy BD. Flat-out.

 

[Arg Clin]

There's no point in discussing with you, as you keep talking about this mythical "future" that doesn't exist. Even if future software three years from now is completely multi-threaded (hint: it won't be) it'll perform just slightly better than the 2500K while consuming 50% more power, which negates the small advantage to most people. There's no secret performance for AMD to reveal. The facts are here today, and it performs in multi-threaded comparably to the i5-2500K because Sandy Bridge has 50% higher IPC than Bulldozer.

The future doesn't exist? Sorry but you lost me there..

You buy today's CPUs for today's performance, and comparing today's performance Bulldozer is junk in comparison; therefore, at the present, Bulldozer is junk. Even if everything would be multi-threaded it'd merely be a "meh" because of its power consumption and the fact it overclocks no better than the competition, not to mention the price is higher.

My argument was not a "should buy/should not buy" Bulldozer. "Junk" is going way overboard. Everything that isn't performance king isn't necessarily junk pr definition. When buying a cpu, one should look carefully at how any given chip might perform in real world applications that one mihgt be planning to use. Forget the rest. (e.g I'm 100% "meh" as to how a chip performs in 3dmax, since I'll never in my life use that app, but care a great deal how it performs in games.. just an example) Reviews and bechmarks are useless if you do not compare them to your particular needs and wishes.

The i7-920 at 4GHz could also be a room heater, but guess what: it also had the performance to back it up. In 2008 it simply demolished everything else in the desktop market, whether it was single or multi-threaded workloads. It's still so good, in fact, that both at 4GHz in multi-threaded tasks this THREE-YEAR-OLD CPU should be comparable or faster than an 8-core Bulldozer.

Now I'm confused. So absolute performance DO matter. Wait! that means that Bulldozer does perform well in some situations and isn't crap in every imaginable situation after all..?

And you being able to get over 8GHz out of a processor with LN2 isn't impressive and proves nothing about it in the real-world. It doesn't prove anything at all about it having good headroom with air cooling, and as we can see from the facts it overclocks no better than Sandy Bridge: both reach 4.5GHz on average. We don't know yet what Sandy Bridge can achieve with LN2 because, again, the multiplier is limited to 57x. With Sandy Bridge-E we'll be able to see.

It says something about what can be expected down the road for an architecture. Assuming the future exists that is..

 

[LOL_Wut_Axel]

The future doesn't exist? Sorry but you lost me there..

Learn to read. I said the future YOU'RE envisioning doesn't exist. Stop trying to create a spin; it won't work

My argument was not a "should buy/should not buy" Bulldozer. "Junk" is going way overboard. Everything that isn't performance king isn't necessarily junk pr definition. When buying a cpu, one should look carefully at how any given chip might perform in real world applications that one mihgt be planning to use. Forget the rest. (e.g I'm 100% "meh" as to how a chip performs in 3dmax, since I'll never in my life use that app, but care a great deal how it performs in games.. just an example) Reviews and bechmarks are useless if you do not compare them to your particular needs and wishes.

It's not that it's not the performance king. It's that it gets absolutely trashed in single-threaded workloads and mildly multi-threaded workloads (ones that use two-four cores, like gaming). Even in heavily multi-threaded workloads, on average it's insignificantly faster (<5%) than the i5-2500K. It also costs more, whether it's in upfront price or in electricity costs. And no, you don't compare two processors by looking at one or two particular programs unless you're simply gonna use it for work (and even then, time is money, so a Core i7-3930K would be a better option). You're supposed to look at the overall picture, and that's where Sandy Bridge has a huge advantage. No matter the workload, it performs great.

Now I'm confused. So absolute performance DO matter. Wait! that means that Bulldozer does perform well in some situations and isn't crap in every imaginable situation after all..?

No, you're trying to spin everything and playing the "I don't understand" card even though you know what I meant right from the beginning. I made it clear right from when we started this "discussion". Absolute performance=overall performance. Using your own logic, that means Bulldozer is crap because it only does well in 1/10 or 2/10 scenarios, whereas Sandy Bridge does well in all 10. The same held true for Nehalem, and to this day it's impressive to see what the i7-920 could and still can do. It's pathetic to think that the FX-8150, currently a $280 CPU, isn't even faster at the same clocks in multi-threaded workloads as a $300 CPU that was released three years ago.

It says something about what can be expected down the road for an architecture. Assuming the future exists that is..

No, it doesn't. The Pentium 4 was able to hit 8GHz+ with LN2 as well. Does that mean it was a great architecture to build upon in the future? No.

Also, do you actually think that Intel is gonna just go to rest and keep putting out quad-core CPUs for <$400 in the future? Intel can make CPUs with moar cores&#8482; as well, and two years from now we'll probably be seeing six-core Intel CPUs under $400 that also improve IPC from Ivy Bridge. What's AMD gonna do then? Make 12 and 16-core CPUs that can only match them in multi-threaded workloads while consuming 2x more power?

Read what's in bold.

 

[AtenRa]

CPUs are not only in the $200-300 range, you people forget it all the time

 

[Arg Clin]

@ LOL Wut Axel

Learn to read. I said the future YOU'RE envisioning doesn't exist. Stop trying to create a spin; it won't work

Sorry I forgot you have a magic crystal ball. Your speculation that multithreading isn't the future is just as hopeless as those who thought we'd be on single cores forever back in the days. Or the nay-sayers even further back, that said SIMD was a cheap excuse for not making a powerful FPU.

It's not that it's not the performance king. It's that it gets absolutely trashed in single-threaded workloads and mildly multi-threaded workloads (ones that use two-four cores, like gaming). Even in heavily multi-threaded workloads, on average it's insignificantly faster (<5&#37 than the i5-2500K. It also costs more, whether it's in upfront price or in electricity costs. And no, you don't compare two processors by looking at one or two particular programs unless you're simply gonna use it for work (and even then, time is money, so a Core i7-3930K would be a better option). You're supposed to look at the overall picture, and that's where Sandy Bridge has a huge advantage. No matter the workload, it performs great.

No, you're trying to spin everything and playing the "I don't understand" card even though you know what I meant right from the beginning. I made it clear right from when we started this "discussion". Absolute performance=overall performance. Using your own logic, that means Bulldozer is crap because it only does well in 1/10 or 2/10 scenarios, whereas Sandy Bridge does well in all 10. The same held true for Nehalem, and to this day it's impressive to see what the i7-920 could and still can do. It's pathetic to think that the FX-8150, currently a $280 CPU, isn't even faster at the same clocks in multi-threaded workloads as a $300 CPU that was released three years ago.

It's still not a "should buy Bulldozer/should not buy Bulldozer" argument I'm making.

No, it doesn't. The Pentium 4 was able to hit 8GHz+ with LN2 as well. Does that mean it was a great architecture to build upon in the future? No.

I disagree. The Netburst architecture actually did quite well later even though Willamette wasn't a very good chip in its time. Like Bulldozer it was ahead of its time. Was it a crappy product? Not by a long shot.

Also, do you actually think that Intel is gonna just go to rest and keep putting out quad-core CPUs for <$400 in the future? Intel can make CPUs with moar cores&#8482; as well, and two years from now we'll probably be seeing six-core Intel CPUs under $400 that also improve IPC from Ivy Bridge. What's AMD gonna do then? Make 12 and 16-core CPUs that can only match them in multi-threaded workloads while consuming 2x more power?

Unlike you ( ?) I'm excited about all new computer technology, whether that be from Intel or AMD. I certainly wouldn't hope for Intel to stagnate their development. If Intel does in fact continue to come out with great cpus at affordable prices, I'd say great! - competition still exists. Maybe you don't remember when you had to fork out $500+ to get a half decent CPU. Some of us does - but that's really besides the point though.

 

[exar333]

@ LOL Wut Axel


Sorry I forgot you have a magic crystal ball. Your speculation that multithreading isn't the future is just as hopeless as those who thought we'd be on single cores forever back in the days. Or the nay-sayers even further back, that said SIMD was a cheap excuse for not making a powerful FPU.



It's still not a "should buy Bulldozer/should not buy Bulldozer" argument I'm making.

I disagree. The Netburst architecture actually did quite well later even though Willamette wasn't a very good chip in its time. Like Bulldozer it was ahead of its time. Was it a crappy product? Not by a long shot.

Unlike you ( ?) I'm excited about all new computer technology, whether that be from Intel or AMD. I certainly wouldn't hope for Intel to stagnate their development. If Intel does in fact continue to come out with great cpus at affordable prices, I'd say great! - competition still exists. Maybe you don't remember when you had to fork out $500+ to get a half decent CPU. Some of us does - but that's really besides the point though.

With that said, your credibility is now gone...

 

[Arg Clin]

With that said, your credibility is now gone...

Northwood holding its own against AthlonXP: http://www.anandtech.com/show/866
Prescott doing well: http://www.hardcoreware.net/reviews/review-195-1.htm

 

[frostedflakes]

With that said, your credibility is now gone...

P4C was actually a pretty good chip back in the day. Things started to go downhill after Northwood, though.

 

[LOL_Wut_Axel]

@ LOL Wut Axel


Sorry I forgot you have a magic crystal ball. Your speculation that multithreading isn't the future is just as hopeless as those who thought we'd be on single cores forever back in the days. Or the nay-sayers even further back, that said SIMD was a cheap excuse for not making a powerful FPU.

I never said that. You really need work on reading comprehension. I said not everything will be multi-threaded in the future, as some workloads are very difficult to parallel. If you actually think 100&#37; of desktop workloads in the future will be multi-threaded I have a bridge to sell you for the low price of $1,000. Even if it is, which it won't be, the advantage will be academic at most (<5%) while the 50% higher power consumption is something most people will take into account. Not only that, but it [FX-8150] also costs more upfront, and if you go for the comparatively priced FX-8120 has shown to be clocked lower because it's not as well binned. It can reach the same overclocks, but it needs even more voltage, bringing power consumption even higher. The higher power consumption not only means higher electricity costs but also higher heat output and more heat on your other components. It also means you need a better heatsink for cooling, and a motherboard with a better power delivery system. You also need a higher wattage power supply, so it's not JUST the CPU itself that you need to take into account.

So even if every desktop workload where to be magically multi-threaded in the future, the advantage wouldn't turn it into a better alternative either. And if you want to talk about the future, what about Ivy Bridge? That will bring 10% higher performance than Sandy Bridge and over 25% better power consumption, certainly higher than what AMD is aiming for with Piledriver.

It's still not a "should buy Bulldozer/should not buy Bulldozer" argument I'm making.

And why should I care? That's what the consumer cares about. Consumers also care about future upgrade-ability to better components. Again, you buy today's CPUs for today's performance and tomorrow's CPUs for tomorrow's performance.

I disagree. The Netburst architecture actually did quite well later even though Willamette wasn't a very good chip in its time. Like Bulldozer it was ahead of its time. Was it a crappy product? Not by a long shot.

Netbust was a piece of crap, but Bulldozer is just as bad or even worse. At least Netburst had the ability of scaling to much higher frequencies than K7/early K8, and that allowed it to make up somewhat for its huge IPC deficit. There were also some good budget models released like the Pentium D 805 which offered comparatively better performance than the Athlon 64 3000+ at a competitive price, but at the cost of much higher power consumption.

With Bulldozer AMD wasn't even able to get better frequencies than Intel, plus there's no model being released that's comparatively better than what Intel has. For example, the FX-4100, due to its architecture and ~10% IPC deficit in comparison to K10.5, needs to be clocked at around 4.4GHz to match a 4GHz Phenom II X4. That's certainly not an upgrade, as by then both are close to or already at their overclocking limits. The Core i3-2120 is already good competition for the Phenom II X4 965. The only quad-core from AMD that really makes sense is the budget Athlon II X4 631 at just $90.

Unlike you ( ?) I'm excited about all new computer technology, whether that be from Intel or AMD. I certainly wouldn't hope for Intel to stagnate their development. If Intel does in fact continue to come out with great cpus at affordable prices, I'd say great! - competition still exists. Maybe you don't remember when you had to fork out $500+ to get a half decent CPU. Some of us does - but that's really besides the point though.

Oh, of course. Because telling it like it is and being objective means "I'm not excited about computer technology". If Bulldozer were good, I'd be the first one here recommending it, just like I frequently recommend AMD GPUs over their NVIDIA competition. AMD makes overall better GPUs than NVIDIA and has for some years, which is why I recommend the Radeon HD 6870 over the GTX 560, the Radeon HD 6950 over the GTX 560 Ti, the Radeon HD 6770 over the the GTS 450, and the Radeon HD 5670 DDR5 over the GT 440 DDR5. But hey, I clearly hate AMD.

If they make a crappy product, they need to be called out for it.


If anything, it's you that hates technological progress because you're saying a CPU that should've never seen the light of day is "okay". You're the one that's not criticizing them for making a crappy product and pushing them to try harder.

Same thing again.

 

[Vesku]

"a CPU that should've never seen the light of day", a bit much there Lol_Wut_Axel. I mean ideally they could have held off for GF 32nm improvements but they were heading towards another fiscal year with the Phenom II in their top tier (entry tier for Sandybridge) cpu slot. At least we know the CMT hit was right on target, but I have concerns that AMD will continue to have problems until and unless GF does a great job with their next node shrink. When was the last time GF (AMD) had a smooth node transition?

 

[notty22]

AMD needed to get a 32nm server cpu to the market. At 2ghz all those cores will accomplish some of the design goals and be a viable option for some customers.
Without a die shrink server cpu, the server market share would just continue to dwindle. Now it might increase.

 

[LOL_Wut_Axel]

"a CPU that should've never seen the light of day", a bit much there Lol_Wut_Axel. I mean ideally they could have held off for GF 32nm improvements but they were heading towards another fiscal year with the Phenom II in their top tier (entry tier for Sandybridge) cpu slot. At least we know the CMT hit was right on target, but I have concerns that AMD will continue to have problems until and unless GF does a great job with their next node shrink. When was the last time GF (AMD) had a smooth node transition?

Releasing a die-shrinked Phenom II X8 using Llano's revised K10.5 improvements and implementing core gating to it would've been a better option. It would've meant a smaller die, lower power consumption, ~13-14&#37; better IPC, and possibly higher clock speeds (the Phenom II X4 980 is already at 3.7GHz base).

As it stands now, the Athlon II X4 631 is a better option than the FX-4100 and at a lower cost.

Also, the Phenom II and consequently AMD's transition to 45nm went smoothly.

 

[Vesku]

C2 Phenom IIs had the Phenom I mediocrity to thank for getting a nod of semi-approval from enthusiasts. AMD should hope they get as much of a pass with BD improvements.

Judging from the X4 631 the 32nm process as is would hold back a Llano x8 as well. If they release a Black Edition this year we can get an apples to apples comparison. But those recent posts about the MSI board overclocking shows in the 4GHz range before voltage and power start spiking. A Llano x8 would probably be ~20-30% better than FX 8 series as is in MT tasks but it seems they would hit a wall in 6-9 months as the Phenom II core's age would make finding additional improvements difficult.

 

[Maximilian]

Why is there still a debate over this?

Bulldozer, the CPU released a couple of weeks ago absolutely fails across the board. There is nothing good about this CPU at all.

- Worse IPC than phenom II
- Power consumption is average, terrible when overclocked
- Meh at multitasking despite having "8 cores"
- Dosent even beat the old nehalem stuff, trades blows with C2Q and PII, gets beaten by PII x6 in quite a few benchmarks...
- Costs a lot considering its performance/the competitions performance
- Costs a lot for AMD to manufacture, dosent affect us directly but its still a negative for this CPU

Maybe it is a necessary stepping stone to something better, maybe in the future the architecture will mature into something competitive and desirable on some level but for now the product we can buy today is garbage and those CPU's will not get better with time as multithreading becomes more common as some seem to predict. This will end up just like the phenom 1, completely eclipsed by its predecessor and forgotten about with very few people owning one.

 

[LOL_Wut_Axel]

C2 Phenom IIs had the Phenom I mediocrity to thank for getting a nod of semi-approval from enthusiasts. AMD should hope they get as much of a pass with BD improvements.

Judging from the X4 631 the 32nm process as is would hold back a Llano x8 as well. If they release a Black Edition this year we can get an apples to apples comparison. But those recent posts about the MSI board overclocking shows in the 4GHz range before voltage and power start spiking. A Llano x8 would probably be ~20-30&#37; better than FX 8 series as is in MT tasks but it seems they would hit a wall in 6-9 months as the Phenom II core's age would make finding additional improvements difficult.

It was pretty easy for AMD to get that ~4% IPC boost from K10.5 in Llano. They could've just have made a K11 that continued to tweak the architecture and made improvements in more areas, or perhaps gotten off their lazy asses and put essential transistors in the Bulldozer die by hand instead of using an automated process even if it saved them some time and money. To this day I'm still baffled by how many transistors it has. ~Two Billion for that kind of performance, are you kidding me? It has tons of slow, unneeded cache for desktop workloads.

Taking into account AMD is predicting a 10-15% improvement each year in performance/watt for Bulldozer, it really doesn't look good at all. Ivy Bridge will get a much bigger boost than that.

 

[bononos]

Why is there still a debate over this?

Bulldozer, the CPU released a couple of weeks ago absolutely fails across the board. There is nothing good about this CPU at all.
......

Maybe it is a necessary stepping stone to something better, maybe in the future the architecture will mature into something competitive and desirable on some level but for now the product we can buy today is garbage and those CPU's will not get better with time as multithreading becomes more common as some seem to predict. This will end up just like the phenom 1, completely eclipsed by its predecessor and forgotten about with very few people owning one.

The debate has narrowed to sniping over "nothing good"/"complete trash" vs "slower/hotter/more expensive compared to Sandy Bridge" except for fanboy postings.

BD's poor/mediocre multithreaded performance puts some doubt into estimated perf of future respins.

AMD doesn't seeem to have the clout or wherewithal to get vendors onside esp wrt Microsoft. Win7 thread scheduling is suboptimal for BDs, Win8 is supposed to fix this. SSE5 is another headache since it is competitor to Intel's SSE4 and forces developers to have 3 codepaths if they want to cover all bases. Intel's anti competitive nonsense in SSE makes it difficult for others(AMD) in their circuit design which is one reason why AMD chose to develop its own SSE.

 

[soccerballtux]

Why is there still a debate over this?

Bulldozer, the CPU released a couple of weeks ago absolutely fails across the board. There is nothing good about this CPU at all.

- Worse IPC than phenom II
- Power consumption is average, terrible when overclocked
- Meh at multitasking despite having "8 cores"
- Dosent even beat the old nehalem stuff, trades blows with C2Q and PII, gets beaten by PII x6 in quite a few benchmarks...
- Costs a lot considering its performance/the competitions performance
- Costs a lot for AMD to manufacture, dosent affect us directly but its still a negative for this CPU

Maybe it is a necessary stepping stone to something better, maybe in the future the architecture will mature into something competitive and desirable on some level but for now the product we can buy today is garbage and those CPU's will not get better with time as multithreading becomes more common as some seem to predict. This will end up just like the phenom 1, completely eclipsed by its predecessor and forgotten about with very few people owning one.

anybody see that new benchmark earlier today about the 0x55 masked thread scheduling in win7 for the 8150, made all the <=4 thread benchmarks 15-20% faster.

 

[Accord99]

anybody see that new benchmark earlier today about the 0x55 masked thread scheduling in win7 for the 8150, made all the <=4 thread benchmarks 15-20&#37; faster.

A Korean site had done a pretty thorough investigation of this behavior on release day.

http://udteam.tistory.com/442 [comparison of 2M/4C versus 4M/4C)

http://udteam.tistory.com/440 (scores from other CPUs of the benches used above)
http://udteam.tistory.com/441

In short:

Pro: Measurable performance increase

Con: At <= 4 threads, still loses to the lower clocked Phenom II 980 in everything except the WinRar benchmark.