The Bulldozer Review: AMD FX-8150 Tested
by Anand Lal Shimpi on October 12, 2011 1:27 AM ESTThe Pursuit of Clock Speed
Thus far I have pointed out that a number of resources in Bulldozer have gone down in number compared to their abundance in AMD's Phenom II architecture. Many of these tradeoffs were made in order to keep die size in check while adding new features (e.g. wider front end, larger queues/data structures, new instruction support). Everywhere from the Bulldozer front-end through the execution clusters, AMD's opportunity to increase performance depends on both efficiency and clock speed. Bulldozer has to make better use of its resources than Phenom II as well as run at higher frequencies to outperform its predecessor. As a result, a major target for Bulldozer was to be able to scale to higher clock speeds.
AMD's architects called this pursuit a low gate count per pipeline stage design. By reducing the number of gates per pipeline stage, you reduce the time spent in each stage and can increase the overall frequency of the processor. If this sounds familiar, it's because Intel used similar logic in the creation of the Pentium 4.
Where Bulldozer is different is AMD insists the design didn't aggressively pursue frequency like the P4, but rather aggressively pursued gate count reduction per stage. According to AMD, the former results in power problems while the latter is more manageable.
AMD's target for Bulldozer was a 30% higher frequency than the previous generation architecture. Unfortunately that's a fairly vague statement and I couldn't get AMD to commit to anything more pronounced, but if we look at the top-end Phenom II X6 at 3.3GHz a 30% increase in frequency would put Bulldozer at 4.3GHz.
Unfortunately 4.3GHz isn't what the top-end AMD FX CPU ships at. The best we'll get at launch is 3.6GHz, a meager 9% increase over the outgoing architecture. Turbo Core does get AMD close to those initial frequency targets, however the turbo frequencies are only typically seen for very short periods of time.
As you may remember from the Pentium 4 days, a significantly deeper pipeline can bring with it significant penalties. We have two prior examples of architectures that increased pipeline length over their predecessors: Willamette and Prescott.
Willamette doubled the pipeline length of the P6 and it was due to make up for it by the corresponding increase in clock frequency. If you do less per clock cycle, you need to throw more clock cycles at the problem to have a neutral impact on performance. Although Willamette ran at higher clock speeds than the outgoing P6 architecture, the increase in frequency was gated by process technology. It wasn't until Northwood arrived that Intel could hit the clock speeds required to truly put distance between its newest and older architectures.
Prescott lengthened the pipeline once more, this time quite significantly. Much to our surprise however, thanks to a lot of clever work on the architecture side Intel was able to keep average instructions executed per clock constant while increasing the length of the pipe. This enabled Prescott to hit higher frequencies and deliver more performance at the same time, without starting at an inherent disadvantage. Where Prescott did fall short however was in the power consumption department. Running at extremely high frequencies required very high voltages and as a result, power consumption skyrocketed.
AMD's goal with Bulldozer was to have IPC remain constant compared to its predecessor, while increasing frequency, similar to Prescott. If IPC can remain constant, any frequency increases will translate into performance advantages. AMD attempted to do this through a wider front end, larger data structures within the chip and a wider execution path through each core. In many senses it succeeded, however single threaded performance still took a hit compared to Phenom II:
At the same clock speed, Phenom II is almost 7% faster per core than Bulldozer according to our Cinebench results. This takes into account all of the aforementioned IPC improvements. Despite AMD's efforts, IPC went down.
A slight reduction in IPC however is easily made up for by an increase in operating frequency. Unfortunately, it doesn't appear that AMD was able to hit the clock targets it needed for Bulldozer this time around.
We've recently reported on Global Foundries' issues with 32nm yields. I can't help but wonder if the same type of issues that are impacting Llano today are also holding Bulldozer back.
Facebook
Twitter
RSS
430 Comments
View All Comments
Pipperox - Saturday, October 15, 2011 - link
With a 40-50% gain Bulldozer would be even ahead of Ivy Bridge.. and what comes next.Or are we still talking about SuperPI?
Or games run at 640x480 lowest quality settings?
The fact is, almost all single threaded applications are old and they run already super fast on ANY cpu and the difference can be seen only in benchmarks.
All recent performance demanding applications are properly multithreaded, and Bulldozer there is competitive with i5 2500 and occasionally with i7 2600 (and with a 10% boost Bulldozer would be competitive with i7 2600).
And this will become more and more the standard one year from now.
Sure Bulldozer has not met the enthusiasts' expectations, it doesn't perform as people would expect an "octacore" (but it's not, it's just a quad with a different form of hyperthreading and "clever" marketing) and it doesn't deserve the FX moniker.
But still it's the most competitive CPU AMD has launched in years, perhaps with the exception of Zacate.
nirmv - Saturday, October 15, 2011 - link
Not all applications are heaviliy multi-threaded, there is still need to improve single thread performance.And even for those few loads that are competitive in performance, they do it with twice the power draw.
See here from xbitlabs review :
http://www.xbitlabs.com/images/cpu/amd-fx-8150/pow...
Pipperox - Sunday, October 16, 2011 - link
But increasing single threaded performance has a cost, on die space and circuit complexity.Bulldozer has a huge die just because it has enormous caches (8MB L2 vs 1Mb on SandyBridge) which probably will turn useful on server workloads (but that's just a guess).
By looking at the die shot, you'd get a 40% die area reduction with "normal" caches.
So AMD engineers decided to drop single threaded performance improvements in favor of higher multithreaded scalability and higher clock speed scalability.
We'll see if in the long run this will pay off.
I agree power consumption doesn't look good in comparison with Intel, but it does look good in comparison to Thuban.
This is the first released silicon of Bulldozer.. i expect power consumption to improve with newer steppings and silicon process tuning.
That being said, Intel has the best silicon process in the whole industry.
AMD can't compete with that.
But i'd guess that at lower clock speeds (like in server), AMD's power consumption will improve a lot.
Looks like with the FX AMD tried to push their current silicon to the maximum which they could (within the 125W TDP which is sort of an industry standard).
LiveandEnjoyLife - Friday, October 14, 2011 - link
Some people are missing the point. At this stage in the game, processor speed is a moot point beyond benchmarks. AMD and Intel make very fast CPUs in relation to what gamers and every day users use them for. Intel CPUs are blazing fast and AMD CPUs are fast. The average Joe does need more than a dual-core CPU. If you were going to actually do something that would require heavy multi-threading, then it comes down to the efficiency the app to make use of the cores and the ability to use hyperthreading. If you wanted the most performance for a mult-threaded application, you would pick more physical cores over virtual cores. So for most of use it comes down to bang for buck.8 cores is better than 2, 4, or 6 for true multi-threaded capable applications.
For speed tests Intel wins hands down.
If you were sitting next to someone playing a game and all things were the same except CPU, you would not be able to tell which machine is running what CPU. However you would notice if one costs significantly more than the other.
That is my 4 cents.
7Enigma - Friday, October 14, 2011 - link
Hi Anand,Great review but there is a text error when referring to pass one vs. pass two of the benchmark mentioned in the Subject line. You said:
"The standings don't change too much in the second pass, the frame rates are simply higher across the board. The FX-8150 is an x86 transcoding beast though, roughly equalling Intel's Core i7 2600K. Although not depicted here, the performance using the AMD XOP codepath was virtually identical to the AVX results."
But the graph clearly shows a complete flip-flop from first pass to second pass. When I look closely it appears you ordered the text and graphs differently and were referring to if you had the non-AVX and AVX-enabled graphs next to each other instead of in separate sections. Basically the text and graphs don't match up.
HTH
Iketh - Friday, October 14, 2011 - link
You're an utter retard. The reason they're sold out is newegg advertised these nicely all over their site, including the front page, with "World's first 8-core desktop processor."There are plenty of reasons to purchase these processors aside from their performance and that's ok. But the majority bought them thinking they're gonna "rock", and those are the ones "showing intelligence." Same goes to you for thinking the majority is well-informed/intelligent.
What's even worse, the 8-core version for sale is the 3.1ghz, not the 3.6 tested in this review. I'm seriously LOL'ing...
How many did Newegg have in stock anyhow? Wouldn't that figure matter regarding your ignorant comment?
rcrossw - Friday, October 14, 2011 - link
I have used AMD products for years. I use Intel at work. So to me there is no real difference between the two for what Business and the Average Computer user want or expect.Does it run, does it do the work I require of it, and do my programs and Network Access
work well and are reliable?
Intel indeed has incredible Processors, fast and reliable, and in the high end - expensive.
AMD is Low and Mid range - with processors that the average person can afford. Who is the most innovative - both. Today Intel has been , now I think the user needs to give this New X86-64 Architecture a chance.
I have a Asus M5A99x EVO with an FX6100 installed. The only problem I have had is having to upgrade the BIOS to accept the new Processor. So far I have had the Processor to 4.2 Ghz. Though AIDA 64 caused a BOD on one test. At 3.8 Ghz runs like a champ. Stil
working back to as close as I can get to 4.2 on Air.
After three years I have retired my old Phenon II Tri Core 720 for this, and it works for me.
I am not an extreme gamer, etc. But test it your self, before being too overly critical.
Does it work for me.
As an aside, next a SSD for faster response.
For those interested:
Asus M5A99x MB BIOS 0810 ( Newest)
AMD FX 6100 at 3.8 ghz
Corsair Vengeance 1600 - 16 gigs
HIS Radeon HD 6850
Windows 7 Ulimate 64
HPLP2475W Monitor at 1920x1200 DP
WD 500 SATA
WD 1001 SATA
LG H20L BD-R
Plextor DVDR
Enermax 620 Liberty PS - I know old but works.
Thanks
davbran - Saturday, October 15, 2011 - link
I have been having a hard time writing a comment on this topic without drawing fire from trolls.This review is hogwash without more information.
If the hardware is the same on all test machines, apart from the CPUs, then there is no wonder the performance was so bad. 6 Cores are going to utilize, and I am just pulling a number out of my ... hat, 4gb of RAM more efficiently than an 8 core using simple kitchen math. No need to break out the slide rules. It's a known fact, to most, that the big bottleneck in the multicore/multiprocessor world is memory. Mind you that's if we are factoring in that all the code that was used for testing purposes was written with multi-threading in mind.
You just can't compare apples to bananas in this manner.
silverblue - Saturday, October 15, 2011 - link
Each to their own. I thought it was a pretty good review, and Anand certainly held back from slating AMD to hell.Iketh - Sunday, October 16, 2011 - link
LOLLLLL