A few months ago I launched something we quickly titled "Bench". The idea behind AnandTech Bench is that it's a publicly accessible version of the database of benchmarks we've run internally.  You can currently compare 34 AMD CPUs and 36 Intel CPUs in the engine across 18 benchmarks.  I'm working on adding power data as well.

You can access Bench at its own URL: http://www.anandtech.com/bench  

Currently Bench only has CPU data in it but there are plans to expand it to storage and GPUs in the future, the former being far easier than the latter due to constantly changing drivers. The data used in bench is the same data used in our reviews, but it has to be entered in manually after a new CPU launches. If you ever see a chip get reviewed on AT but don't see its data in Bench, drop me a line and I'll make sure it gets in there. 

Today I added in data for the Atom 230 and 330 processors using Intel's D945GCLF and D945GCLF2 motherboards so you can see exactly how both single and dual-core Atom stack up to modern day desktop microprocessors. 

I'm also considering running data on an older CPU. In my recent Zotac Ion review I included performance results from a single-core Northwood Pentium 4 2.66GHz processor, which inspired me to want to run a whole slew of older P4 numbers for inclusion in bench. I don't think it's wise to spend several weeks rerunning every single old CPU out there, but I figured one or two couldn't hurt. 

Any suggestions from the crowd? Is a single-core Pentium 4 good enough or would you like to see some dual-core P4 stuff? What about anything from the Athlon 64 days? Respond in the comments and come to some sort of reasonable agreement and I'll see about getting the data in there :)
Comments Locked

92 Comments

View All Comments

  • 36772 - Saturday, May 30, 2009 - link

    Hi, I own a pair of Athlon MP 1600+ processors. I got them as a gift a year ago. They're "slow" but very responsive. There's only one problem I have so far. The PCI on my MB is only 2.1 and I need 2.3 for my TV card. As one of the first consumer SMP platforms - can the MPs be benchmarked once again? Thank you :)
  • Seramics - Friday, May 29, 2009 - link

    Anand,

    according to AMD official website here,
    http://www.amd.com/us-en/Processors/ProductInforma...">http://www.amd.com/us-en/Processors/Pro...on/0,,30...

    LE-1640 2.7 GHz 512KB 45W
    LE-1620 2.4 GHz 1MB 45W

    so the bench info about LE-1620 having 512kb L2 cache is wrong.
  • Hrel - Friday, May 29, 2009 - link

    I'd really like to see them compared to P4 Hyperthreaded 3.0GHz and the Pentium D 930, as the Pentium D 930 was the first processor from Intel that I feel meets my now heightened performance requirements; and I won't really be interested in anything less than a 3.0GHz Hyperthreaded P4; I don't care how little power it uses. HTPC CPU's are different than mobile CPU's, and I just don't really care about power consumption when I have a wall outlet.
  • DerwenArtos12 - Wednesday, May 27, 2009 - link

    I know the lower is better graphs are a pain and confusion for everyone, and I've had an idea that I hope can help. Instead of doing them by pure time increment, do them as either a percent improvement chart or as a less time chart, essentially, set the graphs so that the largest number is zero, every other processor will then be scored on a negative scale, invert the scale and show seconds of improvement over the baseline.

    For example, the Windows Media Encoder 9 x64 test, the core i7 965 had a low score of 24 seconds, The Atom 230 had a high score of 271 seconds, under the scale that I'm suggesting the Atom 230 would have a score of 0 and the Core i7 965 would have a score of 247, and the chart would be labeled as "seconds of improvement over a baseline, higher is better." The concept would take some explaining in articles at first, but it would give people a defined scoring metric that would remain linear(simply oposed) to the lower is better score.
  • croc - Tuesday, May 26, 2009 - link

    While I am sure that you will get many requests for re-running older CPU's, I am sure that you won't get very many volunteers to help in this endeavour. I doubt if one in one hundred readers understand the time required in setting up a system then running the suite of tests against that CPU, and recording the results. What has been done so far must represent 100's of man-hours to have accomplished, and I for one applaud that effort.

    I am looking forward to the HD benches, as their benchmark suite shouldn't change near as much over time. I hope you include some SCSi / SAS drives for completeness.

    GPU tests? Seems as difficult of a task as trying to measure the climatic changes of one butterfly flapping in the Amazon... But I commend you for at least thinking of making the attempt.
  • NICOXIS - Tuesday, May 26, 2009 - link

    I'd like to see, single core:

    1.6 P4-M
    1.6 Athlon XP-M
    1.6 Atom
    1.6 Turion 64
    1.6 Nano
    1.6 Pentium M
    1.6 Athlon Neo
    1.6 C7
    1.6 Core Solo
    1.6 Phenom (one core enabled)

    I'd love to see performance/watt figures on that.
    Keep up the good work,
    Cheers.
  • Johnmcl7 - Tuesday, May 26, 2009 - link

    Once choice I've seen debated is a new Atom notebook against an older refurbished ULV system, it would be interesting to see how the likes of a Pentium-m 1.1Ghz or Core Solo 1.33Ghz processor compared to the Atoms.
  • 9nails - Tuesday, May 26, 2009 - link

    How close are we to being able to predict CPU and GPU scores combined? At some point all of this data is going to establish a trend. With some math that trend could show predictions for hardware combination's which are synthesized from these results. For those of us who are system builders, it would be nice to be able to pick our CPU's and add in some expected GPU and see where the graphs lie.
  • MrPoletski - Tuesday, May 26, 2009 - link

    how about including some of the other processors, such as arms cortex, vs atom in cross platform benches. Cortex is a competitor of atom after all.
  • jjuhl - Monday, May 25, 2009 - link

    > Any suggestions from the crowd?

    Yeah, I've got a suggestion.
    I own a 'fit-PC slim' (http://www.fit-pc.com/new/fit-pc-slim-specificatio...">http://www.fit-pc.com/new/fit-pc-slim-specificatio... with a 500MHz AMD Geode CPU. I'd love to see that included in the benchmarks to see how it stacks up to the competition.

Log in

Don't have an account? Sign up now