By: Michael S (already5chosen.delete@this.yahoo.com), November 9, 2008 6:08 pm
Room: Moderated Discussions
EduardoS (no@spam.com) on 11/9/08 wrote:
---------------------------
>Joe Chang (jchang6@Xyahoo.com) on 11/9/08 wrote:
>---------------------------
>>Inquirer points out the SPEC Power results posted by AMD for Shanghai and Penryn (Xeon L5420) both at 2.5GHz.
>>AMD results for Shanghai is 731 versus 561 for Xeon.
>>SPEC Power results posted by others for the L5420 were higher, the SuperMicro results
>>being 990, but not as close a system match as the 2 AMD results (Windows instead of Linux, and BEA vs Oracle).
>>Now SPEC Power is not my field, as I have never run it. Inquirer points out the
>>different memory config. However in reading through the reports, AMD has the Xeon
>>L5420 system power at 179, 197 and 279W for idle, 10% and 100% respectively. The
>>SuperMicro result has 97, 108 and 174W respectively. I do not think 2 DIMMs will
>>make up 100W, The difference in the SuperMicro result includes settings:
>>Adjacent Sector Prefetcher disabled in BIOS.
>>C1 Enhanced Mode enabled in BIOS.
>>3pin Server Mode enabled.
>
>Note that for the best results for Xeon the chipset used was the 5100, not the
>traditional X5000, and configured as 2x4GB instead of 8x2GB, also remember that
>5100 use registered DDR-2 and the X5000 FB-DIMMs, other submissions with the X5000
>and same amount of chips chipset have almost the same power consumption as AMD's submission.
>
>Regarding to real life systems and configurations, DL380 is the best seller of
>HP, and use the X5000 chipset, not 5100, the last one is very limited so ofen it's
>avoided,
Avoided by whom? By HP, may be, that just recently started to sell it in DL1xx (180) line and still has no 5100-based machines in DL3xx line, but not by customers and not by other server vendors.
IMHO, HP should replace DL320 with 1S 5100-based gear ASAP.
replacing the 8x2GB for 4x4GB modules costs $202, for the LP versions it's
>$442 in HP web site, both are reasonable prices in my opinion but I would go for
>8x4GB instead of saving a few Watts with DIMMs.
>
>On the other side the Opteron submission use a standard power version, not an (good
>mostly for power efficiency benchmarks) HE version.
>
>From a power effiency perspective these submission just proves (again) that FB-DIMMs
>sucks, doens't trust on this benchmark for reducing your power bill.
>
>But they give a preview of Shangai's performance on SpecJBB 2005, 338,577 isn't bad at all.
>
Is the performance test used in SpecPower identical to regular Jbb2005? 338,577 sounds a bit too good to believe. The best official Barcelona score on BEA JVM=219,269.
338,557 is even a bit higher than 2x3.33 QC Penryn on IBM JVM.
---------------------------
>Joe Chang (jchang6@Xyahoo.com) on 11/9/08 wrote:
>---------------------------
>>Inquirer points out the SPEC Power results posted by AMD for Shanghai and Penryn (Xeon L5420) both at 2.5GHz.
>>AMD results for Shanghai is 731 versus 561 for Xeon.
>>SPEC Power results posted by others for the L5420 were higher, the SuperMicro results
>>being 990, but not as close a system match as the 2 AMD results (Windows instead of Linux, and BEA vs Oracle).
>>Now SPEC Power is not my field, as I have never run it. Inquirer points out the
>>different memory config. However in reading through the reports, AMD has the Xeon
>>L5420 system power at 179, 197 and 279W for idle, 10% and 100% respectively. The
>>SuperMicro result has 97, 108 and 174W respectively. I do not think 2 DIMMs will
>>make up 100W, The difference in the SuperMicro result includes settings:
>>Adjacent Sector Prefetcher disabled in BIOS.
>>C1 Enhanced Mode enabled in BIOS.
>>3pin Server Mode enabled.
>
>Note that for the best results for Xeon the chipset used was the 5100, not the
>traditional X5000, and configured as 2x4GB instead of 8x2GB, also remember that
>5100 use registered DDR-2 and the X5000 FB-DIMMs, other submissions with the X5000
>and same amount of chips chipset have almost the same power consumption as AMD's submission.
>
>Regarding to real life systems and configurations, DL380 is the best seller of
>HP, and use the X5000 chipset, not 5100, the last one is very limited so ofen it's
>avoided,
Avoided by whom? By HP, may be, that just recently started to sell it in DL1xx (180) line and still has no 5100-based machines in DL3xx line, but not by customers and not by other server vendors.
IMHO, HP should replace DL320 with 1S 5100-based gear ASAP.
replacing the 8x2GB for 4x4GB modules costs $202, for the LP versions it's
>$442 in HP web site, both are reasonable prices in my opinion but I would go for
>8x4GB instead of saving a few Watts with DIMMs.
>
>On the other side the Opteron submission use a standard power version, not an (good
>mostly for power efficiency benchmarks) HE version.
>
>From a power effiency perspective these submission just proves (again) that FB-DIMMs
>sucks, doens't trust on this benchmark for reducing your power bill.
>
>But they give a preview of Shangai's performance on SpecJBB 2005, 338,577 isn't bad at all.
>
Is the performance test used in SpecPower identical to regular Jbb2005? 338,577 sounds a bit too good to believe. The best official Barcelona score on BEA JVM=219,269.
338,557 is even a bit higher than 2x3.33 QC Penryn on IBM JVM.
