Cache and Memory Performance

I mentioned earlier that cache latencies are higher in order to accommodate the larger caches (8MB L2 + 8MB L3) as well as the high frequency design. We turned to our old friend cachemem to measure these latencies in clocks:

Cache/Memory Latency Comparison
  L1 L2 L3 Main Memory
AMD FX-8150 (3.6GHz) 4 21 65 195
AMD Phenom II X4 975 BE (3.6GHz) 3 15 59 182
AMD Phenom II X6 1100T (3.3GHz) 3 14 55 157
Intel Core i5 2500K (3.3GHz) 4 11 25 148

Cache latencies are up significantly across the board, which is to be expected given the increase in pipeline depth as well as cache size. But is Bulldozer able to overcome the increase through higher clocks? To find out we have to convert latency in clocks to latency in nanoseconds:

Memory Latency

We disable turbo in order to get predictable clock speeds, which lets us accurately calculate memory latency in ns. The FX-8150 at 3.6GHz has a longer trip down memory lane than its predecessor, also at 3.6GHz. The higher latency caches play a role in this as they are necessary to help drive AMD's frequency up. What happens if we turn turbo on and peg the FX-8150 at 3.9GHz? Memory latency goes down. Bulldozer still isn't able to get to main memory as quickly as Sandy Bridge, but thanks to Turbo Core it's able to do so better than the outgoing Phenom II.

L3 Cache Latency

L3 access latency is effectively a wash compared to the Phenom II thanks to the higher clock speeds enabled by Turbo Core. Latencies haven't really improved though, and Bulldozer has a long way to go before it reaches Sandy Bridge access latencies.

The Impact of Bulldozer's Pipeline Windows 7 Application Performance
POST A COMMENT

428 Comments

View All Comments

  • JumpingJack - Sunday, November 06, 2011 - link

    This is a good point. Reply
  • mianmian - Wednesday, October 12, 2011 - link

    How disappointed I am. I can't believe what AMD will claim later on. Reply
  • Marburg U - Wednesday, October 12, 2011 - link

    Cannot see a reason to wait for Piledriver. Am3+ won't survive that chip, and +15%, even in single thread, won't be enough (for Sandy, I'm not even talking about Ivy).

    If BD had not been so bad i would have hoped in a price drop of the Thuban, and would have gone for it. But now, i fear price spikes of the old Phenom II X6 as it approaches it's EOL.
    Reply
  • Ethaniel - Wednesday, October 12, 2011 - link

    ... using a chainsaw. Newegg sells a 2500k for USD 220. I'm thinking something like 170-180 for the FX-8150. I was expecting a lot from the FX line. And I think that was my mistake, probably. Too bad. Reply
  • Leyawiin - Wednesday, October 12, 2011 - link

    I guess we can take comfort in that some things never change - naming AMD processors are always behind the curve (since before Intel's C2 Duo). Guess I'll hang onto my X4 955 @ 3.6 Ghz for a while longer. It'll be the last AMD processor I'll bother with (and I'm tired of being faithful and waiting on them). Reply
  • richard77aus - Wednesday, October 12, 2011 - link

    ""At the same clock speed, Phenom II is almost 7% faster per core than Bulldozer according to our Cinebench results.""

    I am far from being an expert in CPUs but isn't the main advantage intel has had since core2- sandybridge the per core performance? not closk speed and not multi core.

    I've seen some benchmarks showing real world usage of the SB i3 dual core where it out performs a faster clocked quad core phenom 2.
    Reply
  • richard77aus - Wednesday, October 12, 2011 - link

    Meaning AMD giving first priority to clockspeed and core count was the wrong thing to aim for even if they had achieved a 4ghz+ stock 8 core speed processor, but to actually go backwards compared to such an old arch. is a disaster. (my first post here, is there a way to edit posts?) Reply
  • Kristian Vättö - Wednesday, October 12, 2011 - link

    The thing is that Phenom II, which is AMD's arch, is FASTER clock for clock than their new Bulldozer arch. Intel is far ahead of both CPUs, but it's a bit laughable that AMD's older CPUs actually outperform their new ones. Reply
  • Saxie81 - Wednesday, October 12, 2011 - link

    Hey Anand, did you happen to get the power consumption numbers when you hit 4.7ghz?

    This is... disappointing. I knew the Single thread benchmarks were going to be bad, but you need to be running something thats needing the 8 cores, if not its of no use. Kinda like using a Magny Cours to run Crysis.
    Reply
  • Anand Lal Shimpi - Wednesday, October 12, 2011 - link

    I'm going to be doing some more overclocking tomorrow, but I broke 300W at 4.7GHz :-/ Reply

Log in

Don't have an account? Sign up now