The Impact of Bulldozer's Pipeline

With a new branch prediction architecture and an unknown, but presumably significantly deeper pipline, I was eager to find out just how much of a burden AMD's quest for frequency had placed on Bulldozer. To do so I turned to the trusty N-Queens solver, now baked into the AIDA64 benchmark suite.

The N-Queens problem is simple. On an N x N chessboard, how do you place N queens so they cannot attack one another? Solving the problem is incredibly branch intensive, and as a result it serves as a great measure of the impact of a deeper pipeline.

The AIDA64 implementation of the N-Queens algorithm is heavily threaded, but I wanted to first get a look at single-core performance so I disabled all but a single integer/fp core on Bulldozer, as well as the competing processors. I also looked at constant frequency as well as turbo enabled speeds:

Single Core Branch Predictor Performance—AIDA64 Queens Benchmark

Unfortunately things don't look good. Even with turbo enabled, the 3.6GHz Bulldozer part needs another 25% higher frequency to equal a 3.6GHz Phenom II X4. Even a 3.3GHz Phenom II X6 does better here. Without being fully aware of the optimizations at work in AIDA64 I wouldn't put too much focus on Sandy Bridge's performance here, but Intel is widely known for focusing on branch prediction performance.

If we let the N-Queens benchmark scale to all available threads, the performance issues are easily masked by throwing more threads at the problem:

SMP Branch Predictor Performance—AIDA64 Queens Benchmark

However it is quite clear that for single or lightly threaded operations that are branch heavy, Bulldozer will be in for a fight.

Power Management and Real Turbo Core Cache and Memory Performance
Comments Locked

430 Comments

View All Comments

  • kiwidude - Wednesday, October 12, 2011 - link

    I think this shows what a great job Intel have been doing more than confirming your insulting comment about AMD engineers.
  • JohanAnandtech - Wednesday, October 12, 2011 - link

    "Brand new and spanking Bulldozer has it roots in ancient K6"

    There is some K7 heritage left, but I can not see in any way how this CPU relates to the K6! The K6 had a very short pipeline, a unpipelined FPU for example.

    As when it comes to the server market: AMD seems to have overclocked and cherry picked the 3.6 GHz FX-8100. For the desktop market, clockspeed rules, so AMD didn't care too much about power consumption.

    For the server market, they can go with lower clocked 95 W TDP parts. These should have a much better performance/watt ratio. Also, the server market runs at 30-80% CPU load, the desktopmarket runs a few cores at 100%. So the powermanagement features will show better results in the server market.

    The gaming software needs fast caches (latency!) as IPC is decent. The server software is more forgiving when it comes to cache latency as IPC is more determined by the number of memory accesses and thread synchronization. That is the reason why that L3 is so handy. I think you should wait to condemn bulldozer until it is has been benchmarked on our server benchmarking suite.

    I am worried about the legacy HPC performance of this chip though.It will take some recompiling before the chip starts to shine in this market.
  • FunBunny2 - Wednesday, October 12, 2011 - link

    Had to get this far in the comment thread for sanity. Clearly, AMD (and one may disagree) has chosen to go for superior integer performance in a threaded architecture. D'oh! So what? It means they don't give a rat's rectum about gamers. They care a whole lot about application and database servers. They also accept the fact that single threaded is dying, so just kill it.
  • Makaveli - Wednesday, October 12, 2011 - link

    I stayed up and read this its 2 in the morning excellent review as always anand.

    But instead of back to the future its back to the P4???

    Why AMD WHY for the love of everything holy!
  • Sind - Wednesday, October 12, 2011 - link

    Disappointing.. I hope they can get it together with the aggressive road map.
  • wolfman3k5 - Wednesday, October 12, 2011 - link

    I know, right. I'm also patiently waiting for the AMD Bendover architecture. Maybe it will be competitive, who knows...
  • kiwidude - Wednesday, October 12, 2011 - link

    Hi, the CPU Specification Comparison chart has incorrect info listed under X6 1100T and X4 980 NB clocks. Great review as always love your work.
  • wolfman3k5 - Wednesday, October 12, 2011 - link

    NewEgg doesn't even have any Bulldozers in stock, at all. Not the AMD FX 8150 or AMD FX 8120. I guess that no one is in a hurry to grab one...
  • enterco - Wednesday, October 12, 2011 - link

    Hell, Amazon UK doesn't have any Bulldozer neither...
  • ckryan - Wednesday, October 12, 2011 - link

    Maybe Newegg filed them under Server CPUs where BullDozer belongs.

Log in

Don't have an account? Sign up now