Would AMD be better off if they had not developed Bulldozer?

[Martimus]

Would it have been better if Intel had not Produced P4 , think how it could have effected current Intel architecture if P4 was not developed ?

The development of the P4 architecture really helped Intel in a lot of areas. If they didn't develop the P4 architecture and try to make it work, they likely never would have the phenominal Cache design they have now. They would likely not have nearly as good of a fabrication advantage either, since both of those things were needed to even try to sell the P4. They developed a great deal of workarounds based on the limitations of that microarchitecture that they likely never would have spent the time and money to develop otherwise.

That said, AMD doesn't have to luxury of being able to afford to do that. They are likely to just cut their losses at the high end and abandon it completely, in my opinion.

 

[blckgrffn]

Depends who you ask. If you ask reviewers obsessed with canned benchmarks theoretical numbers over user experience, and others with an interest in intel doing well (or AMD failing), of course the results will be biased. But if you ask random individuals without the ability to bias, then the results would seem to fly in the face of those results. Take [H] blind reality check for example: http://www.hardocp.com/news/2012/01/23/amd_reality_check_challenge. Maybe reviewers should take a little time to actually focus on what matters to, you know, consumers.
That blind test reminds me of when AMD compared to the Zacate APU to intel's much more expensive solutions at IDF, and the APU wiped the floor with intel's attempt. intel had Anand run right over to 'correct' what obviously had to be a mistake! But it just wasn't to be and intel ended up looking worse. lol

lol at the comments at [H] - of course, many of the same ones would be made here if we had a thread about this

 

[sm625]

I think it speaks to the failings of the review sites and enthusiast community in general. Clearly there is something we're missing that is more than subjective. Maybe its minimum framerate. Maybe it is minimum framerate combined with some other easily overlooked variable. But I'm pretty sure the experiment is repeatable and it should be repeated and expanded.

Or maybe amd hacked their catalyst driver so that it will regularly drop frames if the cpu is an intel and the user inputs some sort of simple but obscure hotkey combination.

 

[LOL_Wut_Axel]

I'm interested in future AMD CPUs, I think the overall design vision has a lot of promise, but the actual execution so far has been somewhat of a let-down for Bulldozer.

The biggest problem that I have with Bulldozer, is not the lack of IPC, or anything related to performance, per se, but rather, the high power-consumption. For 32nm, it's bizarre and inexcusable. This is almost like AMD's 65nm, that consumed as much power as their prior 90nm CPUs. It took a long time for them to improve the process as much as was needed.

I came to the stark realization of the power-consumption of AMD CPUs, when I tried to shoehorn my 95W 1045T X6 CPU into my HTPC, with only two 40mm exhaust fans (and the PSU).

I set the BIOS warning temp to 65C, and shutdown to 70C. Well, after running DC apps on all six cores, CoreTemp was telling me that the core temp was 52C, but then the BIOS started beeping, and then a few minutes later it shut down.

I disabled the warning and shutdown temp settings, and booted Win7 64-bit again. Running CoreTemp, the temps seemed to start at 50C, and slowly increase to 65C.

At those temps, the exhaust air coming out the back felt warm enough to pop popcorn with. So I shut down the DC apps, and I'm going to swap the 45W dual-core back in.

Remember what I told you about the power consumption and how it was important for DC? Anyway, just run it without a case for now and it should be fine.

 

[Chiropteran]

I don't understand how higher peak power consumption (and actual lower consumption at idle) as seen as HUGE negative for bulldozer, but when comparing AMD & Nvidia graphic cards the higher power consumption of the nvidia cards is basically ignored, even when it's over a hundred watts more.

FX-8150 + 7970 vs i5 2500k + 580GTX, which consumes more power? Which performs better? My bet is that the AMD system uses less power AND performs better in the majority of games, but you don't see any review sites testing these sorts of things.

 

[blckgrffn]

I don't understand how higher peak power consumption (and actual lower consumption at idle) as seen as HUGE negative for bulldozer, but when comparing AMD & Nvidia graphic cards the higher power consumption of the nvidia cards is basically ignored, even when it's over a hundred watts more.

FX-8150 + 7970 vs i5 2500k + 580GTX, which consumes more power? Which performs better? My bet is that the AMD system uses less power AND performs better in the majority of games, but you don't see any review sites testing these sorts of things.

For the DC folks, we are stymied by the power consumption being higher than the 2600k by a non-insignificant margin yet the performance is worse. If the 2600k gets cranked up and consumes a bit more power, it is still performing really well and the power consumption scales better than BD.

Because of this power deficit, in non-avx optimized apps, you are literally getting the performance of a Thuban. Consider the current PrimeGrid race, I dug through a bunch of users and an 8120 takes 2300 seconds per WU. My 1090t @ a modest 3.4 takes 1700 seconds. That 600 per WU difference, times six, means that they are scoring the same over a given time period. I think my math is right there... (in that 3600 seconds difference, the Thuban would theoretically finish two extra WU's)...

http://www.xbitlabs.com/articles/cpu/display/amd-fx-8120-6100-4100_8.html

What becomes even more distressing is that an AVX enabled 2500k @ stock is eating the same WU's at ~900 seconds and OC'd ones are dispatching them in ~600 seconds. Let's hope that AVX brings the same magic to BD...

For everyone else, yeah, BD is going to use significantly less power than Phenom II most of the time. Once AVX becomes a big thing, then it will a tolerable trade off between Thuban and the performance of BD (probably?)

 

[TheWay64]

I believe the design is and was ahead of its time, tuned for a highly multithreaded world, the future of Bulldozer lies on APUs, heterogeneous cores, since its the only way to make it out of Intels grasp.

If bulldozer is tuned for highly multithreaded world, then why it doesn't beat quad cores like i7 2600 and i7 3820 ? Why six core i7 3960X is nearly twice as fast FX-8150 in cinebench (multi-thread) ?
http://www.anandtech.com/show/5448/the-bulldozer-scheduling-patch-tested/2

If Bulldozer was tuned for multithreaded tasks, then it should have a real and obvious advantage over its competitor in that area. But even against non-HT quad cores i5 2500 the difference is negligible

 

[AtenRa]

If bulldozer is tuned for highly multithreaded world, then why it doesn't beat quad cores like i7 2600 and i7 3820 ? Why six core i7 3960X is nearly twice as fast FX-8150 in cinebench (multi-thread) ?
http://www.anandtech.com/show/5448/the-bulldozer-scheduling-patch-tested/2

If Bulldozer was tuned for multithreaded tasks, then it should have a real and obvious advantage over its competitor in that area. But even against non-HT quad cores i5 2500 the difference is negligible

Cinebench is not the only Multithreaded application, there are other apps that 8 core BD are next to or faster than Core i7 2600K.

 

[exar333]

Cinebench is not the only Multithreaded application, there are other apps that 8 core BD are next to or faster than Core i7 2600K.

Significantly faster though? No. In the end an 8-core is barely competitive with a 4-core. It is dominiated much more than it dominates. Crappy chip.

 

[AtenRa]

Significantly faster though? No. In the end an 8-core is barely competitive with a 4-core. It is dominiated much more than it dominates. Crappy chip.

If it was faster it would cost more, it cost less and it can have the same performance in some multithreaded apps as Core i7 2600/K. Not to mention that it produces almost the same fps at 1920p and above.

Does it matter that it has 8 cores vs 4 cores ??? It is the same argument i was hearing with Pentium 4 vs Athlon, P4 was competitive working at higher frequencies than Athlon, so what ??

You have to learn to compere performance per price not Cores vs Cores or GHz vs GHz

 

[frostedflakes]

Significantly faster though? No. In the end an 8-core is barely competitive with a 4-core. It is dominiated much more than it dominates. Crappy chip.

That's kind of a silly way to look at it IMO. It's like saying the HD 6970 is a crappy chip because it needs 1536 shaders to do what GTX 570 does with only 480. Who cares how many cores it has for highly scalable threaded workloads, what matters is how it actually performs and how much it costs.

For single/lightly threaded workloads, though, it's obviously a completely different story and the much higher per core performance of something like Sandy Bridge makes it a much better choice than BD (or Phenom II or any other AMD chip really, they haven't been competitive with Intel's per clock and per core performance in years unfortunately).

 

[CHADBOGA]

Depends who you ask. If you ask reviewers obsessed with canned benchmarks theoretical numbers over user experience, and others with an interest in intel doing well (or AMD failing), of course the results will be biased. But if you ask random individuals without the ability to bias, then the results would seem to fly in the face of those results. Take [H] blind reality check for example: http://www.hardocp.com/news/2012/01/23/amd_reality_check_challenge. Maybe reviewers should take a little time to actually focus on what matters to, you know, consumers.
That blind test reminds me of when AMD compared to the Zacate APU to intel's much more expensive solutions at IDF, and the APU wiped the floor with intel's attempt. intel had Anand run right over to 'correct' what obviously had to be a mistake! But it just wasn't to be and intel ended up looking worse. lol

You have a Herculean task ahead of you for God knows how many years, defending AMD's subpar CPU architecture.

 

[Abwx]

The herculean task is rather to clean out all the dirt that the Intel
afficionados are spreading around AMD.

 

[CHADBOGA]

The herculean task is rather to clean out all the dirt that the Intel
afficionados are spreading around AMD.

There is no need to spread dirt about AMD, the Truth is more than damning enough.

 

[grkM3]

I don't understand how higher peak power consumption (and actual lower consumption at idle) as seen as HUGE negative for bulldozer, but when comparing AMD & Nvidia graphic cards the higher power consumption of the nvidia cards is basically ignored, even when it's over a hundred watts more.

FX-8150 + 7970 vs i5 2500k + 580GTX, which consumes more power? Which performs better? My bet is that the AMD system uses less power AND performs better in the majority of games, but you don't see any review sites testing these sorts of things.

Why do you hook intel up with a 580 and amd with the 7970?

Ill gladly take a 2500k with a 7970 and have faster everything and less power over the AMD setup thank you very much

 

[Lepton87]

K9 would have been a much more fitting name for AMD's new microarchitecture. It really performs like a dog. I've always wondered why AMD skipped K9 and went straight to K10 as it turns out they didn't they just saved it for later.

 

[Abwx]

There is no need to spread dirt about AMD, the Truth is more than damning enough.

The only thing that is damn right in your posts is your continual AMD bashing.

For those with a brain , they know that BD is quite an achievement
even if they didnt master the chip comsumption.

 

[BD231]

Now, remove your client (Desktop) glasses and put on your server glasses. Suddenly, Bulldozer isn't an illogical move at all. They sure missed the implementation on many fronts, but the idea is damn good. I think that in the long run it will pay off once the 32nm process matures and more PD tweaks come in. If AMD can rein in the power consumption issues and boost IPC a bit, BD/PD will be a decent CPU.

People who complain left and right about Bulldozer on desktop should realize that this is not a desktop CPU, it's a server CPU with a desktop after thought: Hey, this could run on desktop too!

PR speak. Basically coming in way overpriced for the market it was intended for, barely faster or in many cases slower than the competition, and to boot even less energy efficient makes the server market jump for joy :hmm: ... Sir the product you're speaking of doesn't exist yet but being that you have the utmost faith and no evidence I'm compelled to believe you

 

[Abwx]

PR speak. Basically coming in way overpriced for the market it was intended for,


I'm compelled to believe you

BS speak...
Overpriced...??..

No one will be compelled to believe you...

 

[Dark Shroud]

With the current prices of BD I have to say they're more than worth it.

 

[BD231]

Intergalagos is overpriced, ask around.

 

[Sweepr]

You have a Herculean task ahead of you for God knows how many years, defending AMD's subpar CPU architecture.

Try to understand him. Years hyping an unrealeased product and when it finally reaches the market you get something slower than your old K10/10.5 buddy while drawing more power (in a new fab process). It's just sad.

Real AMD fans should admit BD is a fail, stick to Phenom II (which still does good) and see what comes next. Downplaying the competition isnt the best idea.

 

[AtenRa]

Try to understand him. Years hyping an unrealeased product and when it finally reaches the market you get something slower than your old K10/10.5 buddy while drawing more power (in a new fab process). It's just sad.

I believe you should re-check Anands FX8150 review,

http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/7

FX8150 only losses in a single benchmark (Compile Chromium Test) while it has lower idle power consumption than Thuban. FX8150 consume more power in x264 but it is faster than Thuban.

Next time do your homework before you speak

 

[Obsoleet]

I believe you should re-check Anands FX8150 review,

http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/7

FX8150 only losses in a single benchmark (Compile Chromium Test) while it has lower idle power consumption than Thuban. FX8150 consume more power in x264 but it is faster than Thuban.

Next time do your homework before you speak

Owned.
Good link.

 

[Phynaz]

Why do you hook intel up with a 580 and amd with the 7970?

To have the test case match the desired outcome of course.