The 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:

 

Cinebench 11.5 - Single Threaded

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.

The Architecture Power Management and Real Turbo Core
POST A COMMENT

428 Comments

View All Comments

  • B3an - Wednesday, October 12, 2011 - link

    Yep this really is extremely disappointing. I'm actually going to call this AMD's Pentium 4. Thats how bad this is.

    2 billion transistors - thats a massive increase over the Phenom II X6 and what do we get? Nothing. The Phenom II is atleast as good with WAY less transistors and lower power consumption under load. I'm pretty shocked at how bad Bulldozer is. I wasn't expecting performance clock for clock to be as good as Nehalem, let alone Sandy Bridge, but this is just... appalling. When Ivy Bridge is out the performance difference is going to be MASSIVE.

    Intel are surely going to implement more restrictions and hold there clocks speeds back even further. Theres just no competition anymore. Sad day for consumers.
    Reply
  • bennyg - Wednesday, October 12, 2011 - link

    AMD's Prescott to be exact... ironically that's one thing they seemed to shoot for in deepening pipeline and hoping that process would be better... hopefully this is just immature and soon there will be a GF110-style refresh that does it properly...

    Otherwise the whole next gen of AMD CPUs will continue to fight for scraps at the bottom of the heap... and their laptop CPUs will not even succeed there.
    Reply
  • TekDemon - Wednesday, October 12, 2011 - link

    I don't even know if it's just the process since those power consumption figures seem to suggest that they're being limited by the sheer amount of power it's using and the heat being generated from that. Intel had planned to take the P4 to 10Ghz but the fact that it was a power hog prevented that from realistically happening and it seems like you have the same issue here. The clockspeed potential is clearly there since it can hit 7Ghz under liquid nitrogen but for a normal air heatsink setup this is a recipe for failure. It's just way too power hungry and not fast enough to justify it. Why would anybody choose to use an extra 100 watts for largely the same or worse performance vs an i5 2500K? Reply
  • Thermalzeal - Wednesday, October 12, 2011 - link

    I agree, 2 billion transistors are doing what exactly?

    The worst thing is that the water cooler isn't included with the FX-8150. At the performance levels they are providing, they should have just upped the price 30-50 bucks and provided the cooler gratis. Who's gonna need an AMD branded cooler if their not going to buy bulldozer?

    The other point of these review is that there is no availability of any of the parts. So what a wonderful paper launch we have here. Seems like AMD isn't betting on anyone being interested enough to buy one of these things.

    Blasphemy.
    Reply
  • eanazag - Wednesday, October 12, 2011 - link

    You can find them on Newegg today. The price is jacked up though. Newegg must not read AT. Reply
  • jleach1 - Friday, October 21, 2011 - link

    Sigh...it's quite sad. There must be actual people buying these...either that or the supply is terrible. Because there's no way in hell i'd pay those prices for an AMD processor. Reply
  • defacer - Wednesday, October 12, 2011 - link

    Like most people here, I 'm disappointed with BD performance -- even though I have never owned an AMD CPU after my 386DX/40 myself, competition in the performance segment would be nice for a change.

    I won't argue against "it's not 8 core", but calling it a 4-core is IMHO just as inappropriate (if not more).
    Reply
  • yankeeDDL - Wednesday, October 12, 2011 - link

    Ok, how about 4 modules, with 8 integer EU, 4 fetch, 4 decode, 4 L2 caches ...
    Point being, they are 4 modules, not 8 cores, and from many aspects, they are more similar to a 4-core CPU than to an 8-core CPU, being neither one (somewhere in between).

    The fact of the matter remains: the IPC is bad. In multi-threaded, Integer-intensive tasks, BD should crunch the PhenomII X6 (2 more cores, higher clock speed), but it seems you can hardly see the difference. (ref: Excel 2007 SP! MonteCarlo sims).

    AMD now is left with Llano as the only compelling reason to buy AMD over Intel (for netbooks and small notebooks, where Atom is the contender).
    Against Core, either the FX-8150 goes down to $200 or less, or the i5-2500 is just a better buy for the money.
    The advantage is I don't need a new MoBo (huge advantage for me, but not very compelling, in general).
    Reply
  • yankeeDDL - Wednesday, October 12, 2011 - link

    Forgot to mention, regarding the integer-intensive test: the core-i5 is slower by about 9% slower with 9% slower clock, but only 4 execution units (8 logical, with hyperthreading, but hyperthreading should be nearly irrelevant in this test).
    What a blow.
    Reply
  • Ratman6161 - Wednesday, October 12, 2011 - link

    We can argue about weather its really a 4 core or an 8 core, and the argument is interesting from a technical standpoint. But the proof is in the real world benchmarks. From a practical standpoint, if the benchmarks are not there (and they aren't) then the rest really doesn't matter.

    I looked on Microcenter where you can get a 2600K for $279 and a 2500K for $179. An i5-2400 is only $149. So AMD is going to be right back to having to cut prices and have its top end CPU go up against $149 - $179 Intel parts. Worse yet, it will, at least initially, be competing against its own previous generation parts.

    There is one point of interest though and that is the fact that all the FX's are unlocked (according to the story). So it's pretty likely that an FX 8100 will probably overclock about as high as an 8150 once the process is mature. But there again, among overclockers, AMD could find its highest end 8150 competing against its own lower priced 8100.

    Back in the day, I loved my Athlon 64's and 64 x2's and even though I have switched to an Intel Q6600 and then a 2600K, I still really want AMD to succeed...but its not looking good.
    Reply

Log in

Don't have an account? Sign up now