Posted Wednesday 16th January 2008 22:20 GMT
erm, so, according to El Reg, this motherboard can't exist?
http://www.tyan.com/product_board_detail.aspx?pid=566
dual xeon processors with DDR2 memory.
Posted Wednesday 16th January 2008 22:20 GMT
I work for Intel - for another opinion - look here: http://blogs.zdnet.com/Ou/?p=956
+ If you want industry standard power efficiency benchmark results, look here: http://www.spec.org/power_ssj2008/results/res2007q4/
Posted Wednesday 16th January 2008 22:20 GMT
He compares Intel 2006 parts to AMD Q2 2008 parts.
How is this news worthy? The AMD parts he 'tests' can't even be purchased yet.
For more info (and even replies from Neal):
http://blogs.zdnet.com/Ou/?p=956
Posted Wednesday 16th January 2008 22:50 GMT
I can't comment on the AMD Opterons due to not having worked with them or the exact power consumption figures of the Xeons, but the new Dell Quad-core Xeon based Oracle 11 server we are running here is a heat generating beast with so many fans it sounds like a jet engine. Given that power consumption is proportional to heat output it must use a lot of wattage.
Posted Wednesday 16th January 2008 22:55 GMT
@Paul — I'm not Neil Nelson. His quote. Good spot though! You win 5 Register points, redeemable as 1 (one) issuance of respect knucks from any Reg Hack you should see sidling down the street.
@Ben and Nick — Oh yes, I do recommend a peek at that story. If only for the shouting match between George and Neal in the comments.
Intel v. AMD stories always bring out the best in people.
Posted Wednesday 16th January 2008 23:48 GMT
Intel quad core DP Xeons are offered with chipsets supporting the high bandwidth FBDIMMs and ALSO with chipsets supporting the cheaper DDR2 DIMMS (with the 5100 chipset).
Her are some links to a couple of SuperMicro's 5100 chipset DP Xeon boards:
http://www.supermicro.com/products/motherboard/Xeon1333/5100/X7DCL-3.cfm
http://www.supermicro.com/products/motherboard/Xeon1333/5100/X7DCL-i.cfm
This "study" was clearly an AMD funded hack job done to spead FUB about the superior Intel products.
Posted Thursday 17th January 2008 03:56 GMT
How much difference would it have made if they'd used 2 x 8 Gb DIMMs like any sane person?
Posted Thursday 17th January 2008 06:20 GMT
"[The test] showed Intel Xeon-based servers reaching 14 per cent higher throughput performance over similarly configured AMD Opteron-based servers, which consumed 41 per cent less power."
That could also read "Intel servers reached 14% higher throughput whilst consuming 69% more power".
Which is impressive (for AMD)...if my maths is working this time of the morning.
John
(flame icon for the data centre, obviously)
Posted Thursday 17th January 2008 08:45 GMT
It's a huge generalisation to say that databases are idle 80% of the time. Perhaps he should have said the average site's back-end DB is idle 80% of the time, even then........
Posted Thursday 17th January 2008 12:53 GMT
The problem for Neal was that he wasn't allowed to have the newer Intel chips so could not test them.
If you want this bias removed, ask your employer to loan some kit to them, as you do with other hardware review sites.
As to the George Ou link, although George does have a point that the test is using new vs old chips, he refuses to recognise that being unable to get the newest Intel chip means testing it is impossible, he just continually tells Neal to test the new stuff and must pay for it because Intel owe him nothing (and forgetting that Neal owes Intel nothing too).
And for the "Industry standard tests", well we remember what NVidia and ATI did with 3DMark 2003 etc, don't we? They wrote their drivers to perform better on the benchmark than on real games (and in NVidia's case, even with named games, since if you changed the name of the exe, it ran slower). If Intel produce a core with enough cache to hold all the data in the test and AMD don't have enough cache, then Intel will do better than AMD even if in real life the difference doesn't show. A little ironic coming from people who point out (however correctly) that Neal is using old Intel chips which isn't fair.
Heck, Neal offered to show George how to run the tests if George can get the newest CPU's but George dodged it by telling Neal HE doesn't owe Neal any effort (neglecting that Neal is really suggesting that George test the CPUs himself if he thinks Neal got it wrong).
Where George doesn't want to know, he doesn't listen.
Posted Thursday 17th January 2008 13:26 GMT
The benchmark was done using the Xeon E5335/E5345, Clovertown, 65nm parts.
The Quad core E5472 is rated at the same power (80W) as the E5345, but runs at 3.0GHz (instead of 2.33), has 12MB cache (instead of 8MB) and runs a FSB of 1600MHz (instead of 1333) - so effectively will run much faster than the benchmarked unit at the same power rating.
Posted Thursday 17th January 2008 13:42 GMT
[]quote]"[The test] showed Intel Xeon-based servers reaching 14 per cent higher throughput performance over similarly configured AMD Opteron-based servers, which consumed 41 per cent less power."
That could also read "Intel servers reached 14% higher throughput whilst consuming 69% more power".
Which is impressive (for AMD)...if my maths is working this time of the morning.
John
(flame icon for the data centre, obviously)
[/quote]
Nope it would read Intel Servers reached 14% higher throughput whilst comsuming 41% more power.
or
AMD servers were operating at 86% performance of the intel servers whilst consuming 41% less power.
Posted Thursday 17th January 2008 14:08 GMT
"41% less power" means (to me) that the AMD servers consumed 59% of the power of the Intels (100-41).
Therefore, the Intels consumed 41/59 = 69% more power.
No?
John
(Paris icon for me if I get this basic maths wrong)
Back to the article