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
ET - Wednesday, October 12, 2011 - link
I gather you're trying for a job at AMD.wolfman3k5 - Wednesday, October 12, 2011 - link
Even the most hardcore AMD Fanboys are having a hard time defending this bulldozed Bulldozer chip.I have seen people blaming anything from Global Foundries to saying that "It's not all that bad."
If you're an AMD Fanboy in Denial, you should read what's wrong wigt Bulldozer:
1) AMD's marketing is extremely misleading, because Bulldozer is not more of a Quad Core chip than Intel's Core i7 for example. What AMD did was to create a crude implementation of CMT (chip-level multithreading), which is debatable as of now if it's any better than Intel's Hyper Threading, with the mention that Hyper Threading takes less than 5% of DIE Space per Core. Basically each Module (as AMD likes to call its Cores now) has an additional Integer Unit, but the problem is that the additional Integer Unit shares the Fetch and Decode units with the other Integer Unit in the Module. Ergo, the CPU is not a true 8 Core. It is just a cheap AMD marketing gimmick.
2) The reason for the horrendous single threaded performance is the poor design of the Integer units in the modules (you can interpret this as very low IPC count). Couple that with the fact that two Integer Units share the same Fetch and Decode units, and when one additional thread is added that screws things up and the pipeline needs to be flushed, it messes things up for the other Integer Unit as well. At a bare minimum AMD should have added a Fetch and Decode unit for each Integer Unit. But wait, if they would have done that, then they would have needed to add a dedicated FPU for each separated Integer unit. Hell, this is almost worst than Hyper Threading. At least Hyper Threading makes use of unused resources in the CPU. AMD's implementation can leave resources unused.
Bottom line: no matter how you slice and dice it, the Bulldozer chip cannot be called an 8 core chip. It's a 4 Core CPU with CMT.
3) In order to compensate for the design flaws in each Module (what I've described in point no. 2), AMD increased the L2 cache to almost absurd levels for each module, to mask some of the latency created by their poor CMT implementation. Intel's Sandy Bridge does away with 256KB L2 cache per core. Hence, we have ~1.6 Billion Transistors in Bulldozer, which creates leakage, which in turn translates into high power draw. Those of you that defended this failure of a CPU claiming that it was designed for servers should know that Performance / Watt is all that matters for servers, because you run into cooling issues otherwise, which translate into high energy costs both for the servers themselves and for the cooling which is done mostly by Air Conditioning. And we all know that Air Conditioning draws lots of electricity.
Bottom line: it's not Global Foundries fault, it's AMD's fault for not designing a more efficient CPU with less transistors (Sandy Bridge quad core has ~1.12 Billion).
Belard - Wednesday, October 12, 2011 - link
Those are excellent points.Its like AMD decided to DO EVERYTHING wrong that Intel and Nvidia did and put it all onto a single chip!!
The Fusion CPUs provided small size, good performance and excellent GPU at a good price. Bulldog is a mutt of a mess.
Brinip - Wednesday, October 12, 2011 - link
After years of being an avid amd fan. I finally switched this year to Intels sandybridge CPU. Amd could no longer compeat with the pure power of the sandybridge core.I was really hoping the bulldozer would re-write the books and put amd back on an even footing. In some ways they have re- written the books as the architecture of the new CPU is quite a change, however for now at least they don't pose ant threat to Intels superb sandybridge core.
Not only are sandybridge CPUs fast, they are also fairly low power and run very cool. What more can you say really. Fast, cool and efficient.
Perhaps as amd develop it further then improvement may come, however don't expect intel to sit still either.
LordConrad - Wednesday, October 12, 2011 - link
Looks like AMD jumped on the NetBurst bandwagon. My next computer build will be Intel.Artas1984 - Wednesday, October 12, 2011 - link
QUOTE:I was hoping for Bulldozer to address AMD's weakness rather than continue to just focus on its strengths
UNQUOTE
And while the new chip does feature a slight improvement in mutli-threaded apps, it's pound per pound performance is slower than AMD Phenom II X6 1100T
8 core, 32 nm, 8 Mb L2, 3600 MHz CPU SLOWER than 6 core, 45 nm, 3 Mb L2, 3300 MHz!!!
Seriously...
W H A T T H E F U ^ K I S T H I S S H I T???
But it's not a big deal. There are greater problems in this world that cpu debates. At least we have AMD and Intel. Imagine if there was only one company making processors - my god that would be a problem then...
ClagMaster - Wednesday, October 12, 2011 - link
The Bulldozer would be a dynamite server processor provided the OS was optimized for the cores that are available.I think the processor has real potential but needs further optimization with a latter stepping.
As for PC useage, performance of the FX-8150 is not bad but I could do better with an i7-2500. I am a low power person and if I were opting for a PC upgrade tonight at NEWEGG, I would choose the i5 2400S and a X68 Intel motherboard.
jaygib9 - Wednesday, October 12, 2011 - link
I don't know about the rest of you, but personally, if I'm going with an 8 core CPU, I'd go water cooling anyhow. Air cooled pc's should be on their deathbed and water cooling should be the norm. AMD and Intel are leaving more CPU's unlocked for overclocking anyhow. What do you think the performance would look like at 5 GHz/core compared to stock speeds?jaygib9 - Wednesday, October 12, 2011 - link
Personally I wonder if soon they'll require water cooling and quit worrying so much about the TDP(within reason of course), then they can concentrate on the clock speed.Belard - Thursday, October 13, 2011 - link
Geee... a 25% overclock for the fake 8-core CPU will not make up for the 30~50% performance disadvantage over a cheaper $200 intel CPU.THAT is a serious problem.
Water cooling is stupid, in general. Less than 1% of desktop computers have water cooling. They will fail sooner or later.