Intel Avoton Benchmarks Live – Intel C2750 with Supermicro A1SAi-2750F Platform

14
Supermicro A1SAi-2750F Test Platform
Supermicro A1SAi-2750F Test Platform

For those not following in the forums we are currently testing an 8 core Intel Avoton C2750 system. The SoC is central to a Supermicro A1SAi-2750F platform that integrates the SoC, 4x 1 gigabit Ethernet ports, IPMI 2.0 with a dedicated LAN port and many other features (full review to follow). Since we have been getting questions on the forums and directly on this platform over the past week, we decided to start posting benchmarks earlier than normal… and as they are in-progress!

Now Benchmarking Avoton

Today we completed initial testing with the Windows 7 platform and 2x 1333MHz DDR3 1.5v Hynix SODIMMs. The Intel C2750 is able to utilize 1600MHz DDR3 at 1600MHz so this is certainly not a maximum configuration. The positive side is that this did provide perhaps the easiest install method and we were able to test early. We used some of the tests in our older Windows 7 benchmark suite. Here is a decent point of reference for archived results.

Just recently we received 4x 8GB 1.35v SODIMMs from Kingston and are running our Linux Benchmark suite against the second Supermicro A1SAi-2750F right now. Performance is certainly a mixed bag but a VAST improvement over Centerton. Whereas our Intel Atom S1260 benchmarks showed a low power SoC-like design, that could compete with low end cloud instances, the Intel C2750 is a completely different performance class.

What we have seen is a higher power envelope moving from the Intel Atom S1260 to the Intel C2750. On the other hand we are now able to use four times the memory, have an integrated 4x 1 gigabit Ethernet solution. Along with that, this is a revolutionary performance boost rather than an evolutionary performance boost as we have seen with the previous Atom generations and we see on the Intel Xeon E3 side. Here is a quick sample using our c-ray benchmark results:

  • Using our simple scene we saw an Atom S1260 render time on the S1260 of 86 seconds and the complex scene took 1297 seconds
  • In contrast, the Avoton based C2750 simple render time was 5 seconds and the complex scene took 86 seconds

The command we use (in the event that you wanted to replicate in Ubuntu) is:

wget http://www.futuretech.blinkenlights.nl/depot/c-ray-1.1.tar.gz && tar -zxvf c-ray-1.1.tar.gz  && cd c-ray-1.1 && make && cat scene | ./c-ray-mt -t 32 -s 7500×3500 > foo.ppm | tee c-ray1.txt && cat sphfract | ./c-ray-mt -t 32 -s 1920×1200 -r 8 > foo.ppm | tee c-ray2.txt && cd ..

Certainly these are sample benchmarks but Intel may have underestimated their 8x performance increase claims with the new Avoton architecture.

For those that want to see a bit more about the SoC features, here is the HWiNFO64 buses output from the Supermicro test platform:

Supermicro A1SAi-2750F HWiNFO64 Bus
Supermicro A1SAi-2750F HWiNFO64 Bus

Check out the forums for benchmarks as they come in here: Intel Avoton C2750 Benchmarks – Supermicro A1SAi-2750F

Notes

A special thanks to Supermicro for providing the Avoton test platforms and Kingston for providing the 1.35v SODIMMs (model number: Kingston KVR16LSE11/8 ) on short notice.

For those seeking Rangeley platform benchmarks we should have them next week.

14 COMMENTS

  1. Intel was sandbagging? WTF? First time in history we have seen a manufacturer say a lower performance increase than actual?

    I subscribed to that thread.

  2. Interesting.
    And what about performance in comparaison vs Xeon E3 1200L v3 Series (Xeon Haswell low power like E3 1230L v3 : TDP 25w or E3 1220L v3 : TDP 13w) ?

  3. Bobii – getting the low power chips has been very difficult but something I certainly want to take a look at. I think we would also need a quad gigabit mITX motherboard to be able to make a similar comparison.

  4. Peter,

    Link didn’t work but was able to search for it. 26 week lead time but the pricing/ qty is as follows:

    1-$478.1978
    2-$466.7053
    5-$458.7595
    10-$454.1302
    50+-$442.2000

    So it looks like prices are going down!

  5. Un-freaking-believable.

    I want one badly. I saw your redis benchmarks at the end of the thread. I guess too much to ask that it is faster than an E3 V3. I wonder why it is slower than that chip though? Then again, the E3-1220 V3 has higher clock speed.

  6. I am planning to get an asrock itx mobo with this 8 core processor.it have 4 dimm slots,a nice feature.However,the Intel ARK says it is 36Bits,neither 32 or 64.Does this mean I cannot run windows server 64 bit on this setup?Or must I run a fully custom OS?Also,what is the maximum gb of ram i can put in there?
    Thanks for any help given.

  7. These are full 64-bit processors. If you look at his forum post linked in the article, he was using 64-bit Linux and Windows.

  8. Hi
    I want to use this as my pfsense 2.1 Do you know if Marvel Alaska quadport is supported on this ? Doesnt show on freebsd 8.3 HCL though.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.