I think its fair to say that we all hoped for something better from the fine folks at AMD. The Bulldozer architecture seems to take the worst parts of Intel's initial P4 offering (longer pipeline, lower IPC, relies on clock speed for performance) and Sun's Niagra (lots of slow integer cores, each relateively wimpy, relies on lots of threads for performance). Unfortunately what we get is a chip that for today's real world usage is often bested by the chip it replaces performance-wise, at a premium price, eating premium power.

I'm not sure where they will go from here. Down-clocked samples of this chip will surely look even worse by comparison, they would have to undercut not only Intel but also their pre-bulldozer lineup!

I was really disappointed too because I like AMD gear. They stick to the same CPU socket for a long time and across ranges (how many do Intel have at the moment, at least 5 or 6) and you get all the features even on the low end models (virtualisation extensions, AES, ECC RAM support etc), and their on-board graphics are the best available. Their mobo chipsets are always reliable too, unlike nVidia's.

If Bulldozer had turned out better I'd have probably got a Phenom for the moment with a view to upgrading later. I really can't see them selling many of these, and I honestly can't figure out why they even released them in this state.

Yeah, it's kind of sad. I really liked my Athlon 64.

That performance per dollar chart is pretty damning.

I tend to root for the underdogs; this is kinda sad. I mean, the only reason to buy this new product appears to be if you want to keep AMD in existence, but if they're not even capable of slugging it out with Intel anymore it makes one wonder if they're worth having around — with performance like that, Intel can (and does) charge pretty much whatever it wants, at least for anything more powerful than BD (or, as it were, Phenom II's).

I guess they still have the GPU stuff. I guess the integration/low-voltage race is where it's at now... and it doesn't look too good for them there, either, as Intel's GPU stuff isn't terrible and will likely improve whereas AMD doesn't seem to be doing much to cut its CPU shortfall.

A well focused article.

Your math is not at issue but the numbers you're using as input. The 2100 is definitely not 8% faster!
The 2100 is faster at many things alright. It's a good desktop chip. But this idea according to which you can put a single number on a chip's performance is a mental affliction.
There's many jobs at which 8-core Bulldozers are about twice as fast (if not stock then certainly overclocked). The typical desktop user or gamer isn't going to see that performance but it's there.

The 8-core Bulldozer should not be compared to the 2100 but the 2500K. That's the comparison which makes Bulldozer a non-starter for most users.
There are some niches where Bulldozer might make sense (maybe some loads do run better on a Bulldozer which can run 8 threads than on a Sandy Bridge which can only run 4 threads? maybe Bulldozer supports ECC? maybe it's got the VM I/O extension which is disabled in Intel's affordable CPUs?) but for most people it doesn't.

Some of us have been discussing the virtues (or rather a lack of) of the Bulldozer for a couple of days, the discussions are enlightening especially when you then take a look at the SPCR review that concentrates on "Power Usage" age tells us how much power was consumed during each test - this is probably the single most damning review I have seen of the "Bulldozer" architecture.

Feel free to have a look - but please consider carefully as to which thread you should post your comments.

viewtopic.php?f=28&t=63230


Andy

As many have pointed out, there may be something wrong with the current Asus BIOS.
Reviewers who used a different board from Asrock gets better results, pretty much only in gaming though.
Too bad AMD gave Asus boards to all reviewers. .

http://www.hardwareheaven.com/reviews/1 ... ution.html

I'd still not buy it, due to the high power consumption.

I think BIOS can be optimized further (like the link implies), Windows 8 beta will make it faster (as already shown by the Developer Preview),
and GF will improve the manufacturing and get the power consumption down (new revision on the quite immature 32 nm process).
So when the 8170 is launched, these three things will make it better, but still not the best.

Where's the (performance per watt) per dollar graph? I'd be very interested to see that.

And some of us were discussing it over two months ago: viewtopic.php?f=28&t=62854

Here's what I wrote: "It depends on what you're doing exactly but, in general, Intel has a clearly superior technology. ... At stock speeds, I suspect Intel's 95W quad would come ahead at most tasks."

"The FX-8150 will also be sold in a bundle with an Asetek water cooler similar to the Corsair Hydro H50."
I think it should be Antec Kühler H20 920 because they look very similar, the H50 have a taller pump, thicker tubes, and a thinner radiator with a single fan.

http://www.silentpcreview.com/antec-kuhler

Considering the poor performance per watt compared to SB, I wasn't even sure if AMD had bothered sending you guys an FX-8150 to review.

Thanks for the review, nice to see SPCR's take on BD.

I have hope that AMD can still salvage BD. Future steppings (the earliest is due out in Q1 2011 from what I've read) will hopefully reign in on power consumption a bit and allow them to hit higher clocks, which BD desperately needs to be competitive. Programs compiled with AVX, FMA4, and XOP can see some pretty huge gains with BD. Although Sandy Bridge sees similarly huge gains from AVX, so relative to it BD doesn't necessarily gain much, but it does give BD a big leg up over old Phenom II chips that don't support these instruction sets. Win 8 is supposed to have scheduler optimizations that improve single/lightly threaded performance up to 10-12% from what I've seen. etc. Currently, though, Bulldozer is pretty underwhelming, if you want an affordable CPU that performs very well *right now*, 2500K is the obvious choice. But in 6-12 months when hardware and software support has improved I think Bulldozer will be a much better option. Of course by then Intel will have Ivy Bridge, which is supposed to offer some pretty significant gains in performance per watt. Unfortunately AMD always seems to be a generation or two behind these days.

Any shot of getting some heavy multi-threading benches in there? time to complete 4/6/8/10/12 parallel Photoshop benches for example. maybe dual TPMGEnc benches as well. Or at least list the benchmarks thread count along with the results. Also any chance of some of these run under linux as well(like handbrake) just to see if windows is affecting the results? I have been hearing rumors that windows may be un intentionally scheduling threads on cores on bulldozer in a way that harms performance.

A few observations:

• NOD32 looks single threaded to me
• iTunes looks single or dual threaded
• TPMGEnc is probably only 4 or 6 threads.
• Handbrake x264 will use all 6 of my cores at 98% each, and i'd bet would scale up to 8 just fine.

I agree with the performance per watt comparisons, I just would like to see if AMDs bet on heavy multi-threading did pay of or not, and testing with one and two threads just won't show that off at all. judging by the x264 results, i'd say the AMD chips do better on high thread counts.

Thanks for the in-depth look as always.

2012 maybe? q1 2011 has come and gone

D'oh! Yeah meant 2012.

Future stepping is called Enhanced Bulldozer/Piledriver. FX-8150 and other currently released CPU's will stay as they are.

@Cynyr: high multithreading performance is useless when it fails at single thread performance. Single thread usage is much more common than the other one.

I delayed on Llano as I was hoping Trinity would have much lower power consumption since it includes the new bulldozer cores. But seeing these results from all the reviews with poorer performance per watt makes me really hesitant if worse CPU performance with improved GPU performance is worth it.
Think I'll wait until ivy bridge is out and compare those two.

AMD has been hoping for software to alleviate the shortcomings of the bulldozer architecture, but is GPGPU and similar really ready to deliver that promise in the near future?

The only positive thing there is that it will be the "Piledriver" core, it will be fully integrated with the GPU, and it has PCI-E built into the core with a next-gen GPU and hopefully cache that doesnt suck donkey-balls, all in a new platform called "FM2".

Waiting is the only answer to find out, AND I don't have high hopes at all because the "Bulldozer" was so disappointing, and what AMD have said the "Piledriver" performance increases over the "Bulldozer" are going to be doesn't impress me at all - not least because I now don't believe what AMD say, and they might not change the core at all, and all of the performance increases might come courtesy of "Windows 8".


Andy

I may probably be wrong, but I don't think W8 may help so much, Bulldozer performances are disappointing even under Linux (look at Phoronix review).

This gives a fair description of how windows 8 will actually improve performance with Bulldozer CPU's (and others as well perhaps.!?), I don't know whether the next Linux kernel has this in the pipeline, it certainly makes a lot of sense to do this. Basically W8 is putting all of the load (on lightly threaded apps) onto as few cores as possible, and it then shuts down pairs of cores at a time where possible. the basis of the theory is pretty simple (and sensible).

An example. Please note this is not defending the Bulldozer architecture as it stands - which is lame, but how we expected it to be when AMD was telling lies to us 12-months ago.

Windows 8 would make the CPU "more able" to hit the top automatic overclock due to having all of the threads (4 for example) all running on 2-modules, this also allows for 4 cores (2-modules) to fully shut down reducing the power draw substantially. This as it seems according to the sneak peek W8 performance tests looks good for the Bulldozer architecture - Windows 8 plus the next gen "Piledriver" might allow AMD to gain a bit on "Ivy Bridge" compared to how bad the current "Bulldozer" looks compared to "Sandy Bridge". It still wont be enough IMO, but at least it is something. My next system looks like it will be an "Ivy Bridge" system because I cant see AMD catching up anywhere near enough to make an impact,


Andy

While this is a good idea in theory, I think for at least some workloads it would be detrimental, possibly greatly so. BD does not actually have two full cores per module. It is more like hyper-threading on steroids. Some workloads may be fine with this, but others would be competing for the shared parts of the module. This would cause them to run slower than if they were scheduled on a different module, even with the lower overclock.

The idea is to allocate threads to cores in the order 0, 2, 4, 6, 1, 3, 5, 7. That way when you're running four threads or less, each thread has exclusive access to all the resources in a module (such as the front-end and L2) instead of having to share them. Turbo frequency is something the scheduler needs to consider as well. The scheduler in Win 8 is module aware and looks like it will offer some pretty solid improvements in single/lightly threaded tasks by more optimally distributing threads to Bulldozer modules than Win 7 does. Tom's Hardware saw between 8-12% performance improvement in WoW, depending on resolution (link). AMD also had a slide that showed between 2-10% improvement in gaming performance with Win 8 (link). The most CPU limited of the games AMD tested, Left 4 Dead 2, saw a 10% improvement in FPS. The more GPU limited games saw a modest improvement in performance, between 2-4%. But this is very promising, because single/lightly threaded performance was one of the areas BD really fell short of my expectations (that and power consumption). That's why I won't bother with Bulldozer now, but if new steppings (not talking about Piledriver, but B3 revision silicon) can reign in on power consumption a bit and software optimizations can improve performance a fair amount, I think it will be a much better option.

Also worth noting is that HyperThreading has the exact same limitations. For example, with a HyperThreaded CPU, for optimal performance you would not want to send one thread to a physical core and then another thread to the logical core for that same physical core. This would lead to a situation where the two threads are fighting for resources within the core. In theory the performance hit would be greater than Bulldozer, which duplicates a lot of resources within a module, while very little is duplicated with HT. The optimal scheduling is to send each thread to a physical core whenever possible (in the case of a quad core w/HT, when you're running four or less threads). All SMT schemes have this pitfall, whether it's HyperThreading or the module approach. That's why the scheduler needs to be aware of logical cores in HyperThreading just like it needs to be aware of modules for optimal performance. The difference is that Windows has been HyperThreading aware since XP I believe, but the module approach hasn't been around for nearly a decade and schedulers and other software haven't been optimized for it. But once they are it will help.

Thanks for the review.

SPCR and other review sites have all struggled to find some task that AMD FX is good at. It is a noble effort to save a dying breed, and while it seems the FX might have some potential in multi threaded apps this is often very tilted. In SPCR's review HandBreak is as mentioned tailored for new instructions. It is nice to see, but it would be like comparing it to apps using Intel's Quicksync or CUDA - uneven playing field. According to Anandtech.com Cinebench 11.5 multi threaded test is where FX matches i5 2500K's performance per clock, despite having 4 more cores and drawing significantly more power.

Other things that shouldn't be ignored - The 2500K can OC to ~4.4 without effort (just type '44') using common $25-40 air coolers (or 4.5-4.8 + with some voltage and other tweaking). 2500K costs LESS than the FX. Intels current 2600K and not to mention the 6 core Gulftown CPUs, wile being more expensive, are years ahead of the FX's performance.

AFAIAC AMD's new offering is a colossal failure. They made us wait for months, delaying release, only to show up with what seems to be a slightly modified x6 with 8 cores and overall similar performance. I guess AMD forgot about Moor's law. Whats truly sad is that we will soon see socket 2011 8+core chips from Intel wipe the floor with even the 2600K on multi threaded apps. What will AMD do then? Will a 10-20% increase in performance in early 2012 help AMD when they are so far behind now?

Just to comment on Win8 vs Win7, its unclear what the advantage will be. It seems to be task specific, with some tasks favoring Intel and some AMD and others like Adobe Premiere - a highly threaded app - had not affect whatsoever. Some tasks get a boost due to hitting the higher turbo freq, while others get hit by having to share cache and possibly other resources. I don't see how Win 8 helps AMD in a significant way other than (hopefully) better power saving.

Truth be told I would not pay more than $80 for this CPU and even then I would need some strong convincing skills of a very dedicated sales person to even consider......... nah....... I will never buy it, poor power consumption alone is enough to make me look elsewhere.

The performance / watt is really bad so not good for silent PC
But I have expected some undervolt tests and maybe even undervolt underclock see if that will improve drastically the performance / watt.

Linux review: http://www.phoronix.com/scan.php?page=a ... ozer&num=1