Why is the response to Bulldozer so overwhelmingly negative?

[Kevmanw430]

Concillian, if they had done that.... Even though there would be alot of flack from enthusiasts for not making a whole new arch, it wouldv'e most likely performed better on the whole, and since Llano is fairly good on power consumption, been better for servers too. (Not to mention having 8 FPU's if it was an 8 core)

 

[IlllI]

what I have basically figured out is.. Amd targeted the wrong market with this cpu.

 

[frostedflakes]

Yeah, I wouldn't have hesitated to buy a tweaked and die shrunk K10.5. An affordable Phenom III X8 with some IPC improvements like Stars got and slightly higher OCing capabilities would have been a fairly solid upgrade over an X4 955 IMO.

 

[RussianSensation]

If the FX-8150 were priced the same as the i5-2500K, it would be a better value unless you were primarily a single-threaded application user.

People have to realize that if you run 1-4 threads on a 4 core CPU, it's all about single-core performance. Multi-threaded performance only matters if you are running MORE than 4 threads. When people discuss single threaded performance, they aren't necessarily discussing a program that only uses 1 thread.

For gaming, it would be a moot point since anyone using these chips is gaming at 1080P and is GPU-bound, not CPU-bound.

Mostly true for GPU limited games such as BF3, Crysis 2. What about CPU limited games? A lot of people play role-playing and strategy games which are extremely CPU limited.

More here.

As the prices align now, the FX-8150 is not a particularly compelling value, unless you primarily use heavily-multithreaded applications.

Exactly. But if you primarily use heavily multi-threaded apps, $169 X6 1090T looks better vs. a $220 FX8150 or a $245-279.99 FX-8150. So even against Phenom X6, BD is a failure.

Finally, where are the reviews of the $165 FX-6100 vs. the $180 i5-2300? ...What about the $115 FX-4100 vs. the $125 i3-2100?

Those other CPUs are almost irrelevant. Those CPUs stand no chance against an X4 960T for $125 which can be unlocked into an X6, or a $170 X6 1090T. Also for office, general and gaming tasks i3-2300/2400 series beats all of these.

First, you can actually overclock the AMD chips. You can't overclock the i5-2300 nor the i3-2100.

Most people don't know how to overclock. Those who do can get used i5 750/760/i7 860 / i7 920 from previous generation, or get the X4 / X6 processors I mentioned. Also, if I am going to spend $165 for an FX-6100, I might as well spend $225 for a 2500k. Considering a 4.7ghz 2500k is often 40-50% faster than an overclocked 1100T, with nearly half the power consumption, the extra $60 is worth it imho. And well, the FX-6100 is worse than the 1090T....

The FX-8150 either rivals or bests the i5-2500K in Cinebench.

They are very close in Cinebench but BD costs more $$ doing it and consumes 2x the power at stock speeds.

Cinebench R10

Cinebench R11.5

At 4.4ghz, 2500k breaks 7 points in Cinebench. Now when you overclock a 2500k, FX-8150 probably needs to be at 4.7-4.8ghz to match its Cinebench score. Of course by that point, BD is using 200-250W of extra power.

Now, keep in mind that X6 1090T @ 4.0ghz is even better than 2500k in this benchmark and at current market prices that CPU costs $100 less than FX-8150 and $50 less than the FX-8120.

FX-8150 beats 2500K in Monte Carlo simulations, which are enormously important in a number of research fields from economics to evolution.

Conclusion:

Stock speeds
- BD is too slow in 1-4 threaded apps at stock speeds
- BD consumes 2x the power of SB at stock speeds
- BD is more expensive than X6 1090T and yet is barely faster in multi-threaded performance, and often is slower in single threaded apps against Phenom II
- BD isn't even that great for multi-threaded apps against 1st generation i7 CPUs such as 860/870, etc.
- BD is too slow for CPU limited games: strategy, role-playing and massively online multiplayer

Overclocked (not going to repeat some of the points above but most still apply)
- Overclocked BD gains less in performance than overclocked SB chips since it already has higher 3.9/4.2ghz Turbo.

- Power consumption difference becomes even worse for BD

Basically, for a budget all-rounder, X4 955/960 > FX-4100. For multi-threaded apps, X6 1090T is cheaper and is just as good as FX-8120/8150. As an all-rounded processor with good overclocking and power consumption, 2500k/2600k > FX-8120/8150. And if you want a basic lower power consumption chip for games, i3 series takes the cake. BD stands nowhere at current prices. AMD needs a $30-40 price cut on both.

That's not even getting into $149.99-$179.99 Fry's and Microcenter deals on the 2500k........



Also, 990FX boards have inferior SATA3 performance to P67/Z68 for those who are using SATA3 SSDs
http://techreport.com/articles.x/21019/8
and
http://www.hardwareheaven.com/revie...core-i7-2600k-review-platform-benchmarks.html

 

[IndyColtsFan]

And niches are just what AMD needs. They don't have MFG volume to produce head to head samples in a price war. They want a niche that's low-ish volume and high profit margin so they can get their books back in order while giving Glo Flo breathing room while transitioning to making their GPUs on a 28nm process.

Errr, do you seriously think their margins on BD are high?

Many people are approaching this from the standpoint of a typical enthusiast usage profile. Who said that AMD was even considering the typical user with an FX chip?

Their advertising looks geared to enthusiasts to me. Let's take a few quotes off their site:

"Immerse yourself in the most advanced 3D games."
"Chat, watch movies, play games and more with maximum multitasking."
"Be Epic. Be Brutal. Get AMD FX in Your System."

That doesn't sound like they're targeting NASA to me.

Seems they're happy to leave that to Intel while seeking out specific (and profitable) markets to increase their marketshare and, most importantly, profits. Many of their GPU and CPU advances of late seem to specifically target the scientific modeling market segment. This is a segment with gigantic profit margins that AMD wants a piece of. They know they need to get some stable high margin revenue to bring their share prices up. The average price fickle consumer is much less of a concern, I bet.

I'm not sure what your point is -- I seriously doubt BD satisfies the bolded above. High margin chips are the EE edition and the FX series in the Athlon64 days when they were charging $1K for a processor that performed 10% better than a $300 X2.

 

[RobDickinson]

2 billion transistors vs 995 million for SNB which includes IGP.

No wonder it uses twice the power...

 

[jacktesterson]

delete

 

[BudaBomb]

I wonder how well the six core does its got less power draw and is cheaper so it might be a good idea.

 

[Concillian]

I'm not sure what your point is -- I seriously doubt BD satisfies the bolded above. High margin chips are the EE edition and the FX series in the Athlon64 days when they were charging $1K for a processor that performed 10% better than a $300 X2.

Thuban is bigger and cheaper. FX already increases margin, but as I said countless times, FX is just pulling in what extra business they can. Sure they'll aggressively market FX, that's what the marketing department gets paid for. IMO though, the Opterons are the real target and are significantly higher margin.

 

[bononos]

I think that's a very good point, and obviously, Bulldozer looks terrible compared to Sandy Bridge. How much of that is due to the superiority of Intel's fabrication technology compared to GloFo's? This isn't a leading question, I honestly don't know.

....

I don't know if you can make that argument, its like a car maker slapping a huge engine into one of its models and complaining that its level of automotive research does not allow it to compete effectively in terms of fuel consumption. 2 bazillion transistors and BD still can't get the job done....

 

[cebalrai]

very interesting... supposedly every review out there was done on a Asus test board sent to all the review sites by AMD, but this site used a different board and the results were much better. PRetty interesting, beats the i7 2600k in a lot of cases


http://www.hardwareheaven.com/revie...i7-2600k-review-deus-ex-human-revolution.html

Wow this is a much better review. Especially in the gaming category.

 

[Aikouka]

People have to realize that if you run 1-4 threads on a 4 core CPU, it's all about single-core performance. Multi-threaded performance only matters if you are running MORE than 4 threads. When people discuss single threaded performance, they aren't necessarily discussing a program that only uses 1 thread.

I disagree. Testing multi-threaded performance is about seeing how the architecture is able to handle the multiple tasks assigned by the originating process. There is a direct relation to single core performance in these tests, because if a single core processes 1 + 1 in 1ns, a thread from a process will process it in the same amount of time.

It's also very important to keep in mind that testing the performance is directly related to the process being executed and what the sub-tasks consist of. This directly relates to your statement: "if you run 1-4 threads on a 4 core CPU, it's all about single-core performance" and Bulldozer shows exactly why the statement is incorrect. Bulldozer features a single FPU per module, which means that if it receives 8 threads that are attempting to process floating point operations, it will not perform in a single-threaded manner since one module is considered two cores. That is a scenario where the architecture choices directly reflect how well the processor will perform in a multi-threaded situation.

I think what you're trying to say is that multi-threaded performance usually matters if the process is capable of using resources provided by one architecture that the other does not have. An example would be a process using all 8 of Bulldozer's core's integer processing.

 

[OCGuy]

BD is an enormous flop.

 

[rizla1]

waste of time my next upgrade from amd 955 is goin to be..... intel 2500k.. cant wait lol

 

[chihlidog]

For quite a few reasons.

1) The previous gen CPU beats it in certain tasks

2) It sucks power like my Jeep sucks gas

3) It sucks for games, which is what many enthusiasts do with our PCs

4) We were deceived by AMD on several levels

That's quite enough, IMO, for the backlash we're seeing. That, and many of us wanted it to compete with intel and we're butthurt that it doesnt.. It can beat a 2600k in certain things, but not enough things to justify all the other tasks it gets utterly SMOKED by intel in.

 

[soccerballtux]

We were all expecting them to close the single thread performance delta between the core i7s and AMD's lineup, and they completely failed at that.

 

[Kristijonas]

Hey I like unzipping files as much as the next guy, but I don't buy a CPU based on that.

This. Sorry for the quote, but that's the only thing BD noteworthy excels at. (while using more power, producing more heat and costing more that i5 2500k) I'm sure AMD will get back on foot though. BD is a lost case.

 

[Darknite39]

very interesting... supposedly every review out there was done on a Asus test board sent to all the review sites by AMD, but this site used a different board and the results were much better. PRetty interesting, beats the i7 2600k in a lot of cases


http://www.hardwareheaven.com/revie...i7-2600k-review-deus-ex-human-revolution.html

VERY interesting find!

 

[RussianSensation]

This directly relates to your statement: "if you run 1-4 threads on a 4 core CPU, it's all about single-core performance" and Bulldozer shows exactly why the statement is incorrect.

What? Bulldozer's module 8-core design isn't able to process a program with 1, 2, 3 or 4 threads faster than a fast 4-core CPU. Basically, the design sacrifices per core performance in favour of situations where you may use 6-8 threads (rare with today's programs). My point is with most of today's programs, you want a fast 4-core CPU, not a slow 8-core CPU, which is why single-threaded ==> i.e., per core performance throughput is more important when discussing quad vs. hexa vs. octa core CPUs. The main exception is if you run programs that actually use 6-8 threads well, in which case the quad may be much slower (but in the case of BD, a single SB core is at least 45-50% faster, so this is also pretty much negated).

I think what you're trying to say is that multi-threaded performance usually matters if the process is capable of using resources provided by one architecture that the other does not have. An example would be a process using all 8 of Bulldozer's core's integer processing.

Ya, that's what I mean. BD could have had 800 cores and it wouldn't make a difference for 90% of people who don't encode, unzip 10,000 files a day or have render farms.

IF BD had similar IPC to Phenom II, same clocks of 3.9/4.2ghz and 8 cores, then it would clearly be better than Phenom II X6, but it isn't. BD's only hope is 5.5+ ghz clocks and that programs start to use 8 threads like tomorrow.

 

[Will Robinson]

We will just have to wait for the revamped Butthurt architecture models due out later this year.

 

[frostedflakes]

VERY interesting find!

It's really not that interesting, if you cherry pick benchmarks you can show BD in a good light. 8150 does well in some heavily threaded benchmarks, beating the 2500K more often than not and sometimes even being very competitive with the 2600K. It also does fine in most games when you crank up the settings so that you're GPU limited, but of course an Athlon II probably does just as well in these situations, not like you need a super fast CPU when you're GPU limited. It's not representative of real-world performance, but most CPU reviewers turn down the details to remove the GPU bottleneck, that way you're actually benchmarking the CPU, not the GPU. That's the whole point of a CPU review after all, to benchmark the CPU.

In single and lightly threaded workloads, though, it falls far behind Sandy Bridge and even manages to lose to AMD's previous architecture in most tasks. And the power consumption is horrific, especially when overclocked.

 

[MrX8503]

Its deemed overwhelmingly negative because:

-It was heavily delayed and still didn't deliver
-Worse than Phenom II in single thread
-Consumes way more power than SNB

Its only positives are that its cheaper and its competitive in heavily threaded apps. IMO, I think the negatives far outweigh the positives.

 

[AMDrulZ]

I don't know why AMD decided they needed to reinvent the wheel in the consumer space with the bulldozer fx chips the old k10.5 design had plenty of potential. Bulldozer in my mind is a server cpu and server cpu only its simply not dynamic enough to be a good desktop cpu.
like i said in another thread they need to shrink the phenom II to 32nm do some tweaks to it like faster cache and better branch prediction clock it to 4.0ghz and it will be cheaper to produce and better performing than bulldozer on the desktop which means better profits for AMD. I mean they already have k10.5 on 32nm in llano so its not like they will have to start from scratch. it would also be better for global foundries it would help them mature the 32nm node so AMD can actually get the clock speeds they want from bulldozer and fix bulldozers bugs so when they launch pile driver it will be performing even better than the 15% they are aiming for now.

 

[cebalrai]

It also does fine in most games when you crank up the settings so that you're GPU limited, but of course an Athlon II probably does just as well in these situations, not like you need a super fast CPU when you're GPU limited

There's no reason to buy anything other than an Athlon II then if you're a serious gamer. I think that is the take home message you just gave us.

 

[gramboh]

People have to realize that if you run 1-4 threads on a 4 core CPU, it's all about single-core performance. Multi-threaded performance only matters if you are running MORE than 4 threads. When people discuss single threaded performance, they aren't necessarily discussing a program that only uses 1 thread.

...(didn't quote the rest)...

Epic freaking post Russian... EPIC.

This post is all anyone needs to read about Bulldozer from a desktop perspective. I was fully expecting BD to be a let down, and felt bad for the people inexplicably buying 990FX boards in anticipation of the release. The shocker for me is the power usage, especially when overclocked, WTF? Hopefully future steppings at higher speeds are better, because I really do not understand how they are going to make money/get a good ROI on this, and I want AMD to do well financially.