First Shanghai benchmark results.

[duploxxx]

Yet they use different amount of instances.

And still so different from anyone else....sure!

And since when is a normal server configuration 16GB? Plus you would most likely use 4x4GB and not 8x2GB.

I put this along the 40% faster and 3Ghz Phenoms.

They didnt learn from it last time it seems.
http://uk.youtube.com/watch?v=G_n3wv...eature=related
http://uk.youtube.com/watch?v=R7EZmY...eature=related
http://uk.youtube.com/watch?v=6ipfxI...eature=related

there are lots of benches on that site with different amount of instances. off course when you think its convenient for your narrowed view you bring it up, well it isn't:

Intel Corp. Super Micro Computer, Inc. Supermicro 6025B-TR+ (Intel Xeon processor E5345) BEA JRockit(R) 1.6.0_02 (build P27.4.0-10-90053-1.6.0_02-20071009-1827-windows-x86_64, compiled mode) 2 419 Text HTML
Intel Corp. Super Micro Computer, Inc. Supermicro 6025B-TR+ (Intel Xeon processor E5345) BEA JRockit(R) 1.6.0_02 (build P27.4.0-10-90053-1.6.0_02-20071009-1827-windows-x86_64, compiled mode) 4 435 Text HTML


to the world, the days of 32bit and less then 4gb ram are over...but perhaps for some it takes some time to evolve.

choosing between 4 times 4 or 8 times 2 is 25% difference in ram price. lots of reasons since only minor amount will ever upgrade memory.

[Shintai]

there are lots of benches on that site with different amount of instances. off course when you think its convenient for your narrowed view you bring it up, well it isn't:

Intel Corp. Super Micro Computer, Inc. Supermicro 6025B-TR+ (Intel Xeon processor E5345) BEA JRockit(R) 1.6.0_02 (build P27.4.0-10-90053-1.6.0_02-20071009-1827-windows-x86_64, compiled mode) 2 419 Text HTML
Intel Corp. Super Micro Computer, Inc. Supermicro 6025B-TR+ (Intel Xeon processor E5345) BEA JRockit(R) 1.6.0_02 (build P27.4.0-10-90053-1.6.0_02-20071009-1827-windows-x86_64, compiled mode) 4 435 Text HTML


to the world, the days of 32bit and less then 4gb ram are over...but perhaps for some it takes some time to evolve.

choosing between 4 times 4 or 8 times 2 is 25% difference in ram price. lots of reasons since only minor amount will ever upgrade memory.

Why are you posting 65nm Xeon numbers? And cant we agree that if you are so honest and clearly. You would bench it all with the same settings. Yet they didnt.

And did you even notice the 2 numbers you posted are on different confs? Both memory amount and speed.

Tell me how many servers that need more than 2 or 4GB. Besides DB servers.

Not even to talk about 16GB vs 8.

You didnt instantly throw 16GB memory in your desktop either did you with x64?

[gosh]

Tell me how many servers that need more than 2 or 4GB. Besides DB servers.

how many servers that need speed are NOT database servers?

[Shintai]

how many servers that need speed are NOT database servers?

Plenty. Considering the poor AV and other things out there like layer 7 firewalls. Loaded AD servers etc.

But ofcourse. The majority doesnt need eitehr of the 2. Performance is in higher need than memory tho.

[Lightman]

I found this info interesting:

That reassured us that our transaction log disk was not a bottleneck. On a "normal" SLES 10 SP2 we achieved 1400 tr/s on a quad core (an anonymous CPU for now ;-). But Anand's article really got us curious and we replaced our mighty Cheetah disk with the Intel x25-M SSD (80 GB). All of a sudden we achieved 1900 tr/s! No less than 35% more transactions, just by replacing the disk that holds the log with the fastest SSD of the moment. That is pretty amazing if you consider that there is no indication whatsoever that we were bottlenecked by our log disk.

Courtesy of Johan De Gelas

Wonder what that anonymous CPU is

Here are results to compare from his previous article:

[justapost]

I think AMD posted that Intel result because the only other linux results where from sgi running on altix cluster nodes. I wonder why they used an redundant 700W psu for the dual socket opteron test(#18) and non redundants for the quad tests (#89,#90).
Their 2.7GHz results (#91,#92) are the only ones using windows server 2008 on the other hand.

[duploxxx]

I found this info interesting:



Courtesy of Johan De Gelas

Wonder what that anonymous CPU is

Here are results to compare from his previous article:

will even be more interesting when first shangai benches arrive on the virtualization layer, the red line which now has some clear distance will be very narrow and dunny will only survive with the 7460 and off course the marketing bull that 24 cores provide more then 16 cores... if you don't count price and power consumption

[Hornet331]

will even be more interesting when first shangai benches arrive on the virtualization layer, the red line which now has some clear distance will be very narrow and dunny will only survive with the 7460 and off course the marketing bull that 24 cores provide more then 16 cores... if you don't count price and power consumption

better watch that comment about power consumption 6 core dunnington L7455 only has a TDP of 65W and clocked at 2,13ghz -> Opteron 8354 TDP 75W 2.2ghz.
Or the E7450 2,4ghz TDP 90W -> Opteron 8358 2,4ghz TDP 105W

[Macadamia]

better watch that comment about power consumption 6 core dunnington L7455 only has a TDP of 65W and clocked at 2,13ghz -> Opteron 8354 TDP 75W 2.2ghz.

That is definitely not Dunnington's max TDP.

And assuming perfect scaling, The 2.13Ghz 6-core gets only 880 transactions. Of course it doesn't scale perfect so it'll be higher, but it isn't even a victory thanks to the crappy bloodsucking FB-DIMMs.

[saaya]

and the shism between the real world and amds markom world grows larger and larger...
2.7ghz and an average score... wow, that was really worth the cheating

[duploxxx]

better watch that comment about power consumption 6 core dunnington L7455 only has a TDP of 65W and clocked at 2,13ghz -> Opteron 8354 TDP 75W 2.2ghz.
Or the E7450 2,4ghz TDP 90W -> Opteron 8358 2,4ghz TDP 105W

those are barcelona's not shanghai and those are AMD ACP ratings, not TDP

That is definitely not Dunnington's max TDP.

And assuming perfect scaling, The 2.13Ghz 6-core gets only 880 transactions. Of course it doesn't scale perfect so it'll be higher, but it isn't even a victory thanks to the crappy bloodsucking FB-DIMMs.

dunnington's max tdp is 170W, that's why you don't see the top bin in all servers.

[Hornet331]

those are barcelona's not shanghai and those are AMD ACP ratings, not TDP

i never said they where shanghai, and if those rateings are ACP its even worse...

dunnington's max tdp is 170W, that's why you don't see the top bin in all servers.

http://processorfinder.intel.com/det...px?sSpec=SLG9P -> TDP 130W and thats the highest 74xx you can get.
I think you mixed it up with the X5492 -> TDP 150W

[Helmore]

i never said they where shanghai, and if those rateings are ACP its even worse...

AMD rates all their server chips with an ACP rating. You can pretty much count an extra 15 Watts on top of it to get you to the TDP rating.

[Shintai]

AMD rates all their server chips with an ACP rating. You can pretty much count an extra 15 Watts on top of it to get you to the TDP rating.

15?

55=79
75=115
105=137

http://www.amd.com/us-en/assets/cont...61C_ACP_WP.pdf

[Helmore]

15?

55=79
75=115
105=137

http://www.amd.com/us-en/assets/cont...61C_ACP_WP.pdf

It was just a rough figure, a 105 Watt ACP would result in a 125 Watt TDP for the same clocked desktop chip. So 20 would probably be more accurate in this case, but I didn't really put much thought into it.

A 105 W. ACP server chips is usually a 125 W. (TDP) desktop chip
A 75 W. ACP server chip is usually a 89 W. (TDP) desktop chip
And a 55 W. ACP server chip is usually a 65 W. (TDP) desktop chip

All pretty rough, but in the end it's simply better to just measure at the wall outlet. From a average consumer perspective that is.

[informal]

Remember K10 includes IMC and HT controller.

[Shintai]

Remember K10 includes IMC and HT controller.

And i7 uses less power than Core 2.

So whats the new excuse?

[informal]

Less power in what situation?Full load or idle?
You want to say that in power draw test ,a 2.66Ghz Core i7 vs 2.66Ghz Core2Q is a win for i7 in full load across 4 cores?

Also you missed one reply in other thread.

[Shintai]

Less power in what situation?Full load or idle?
You want to say that in power draw test ,a 2.66Ghz Core i7 vs 2.66Ghz Core2Q is a win for i7 in full load across 4 cores?

Also you missed one reply in other thread.

http://www.anandtech.com/cpuchipsets...spx?i=3453&p=3



Its always been a retarded comment with the MCH using ohh so much. And like HT uses alot vs FSB?

Maybe thats why we got huge heatsinks and sometimes fans on PCIe switches alone.

[Macadamia]

What a joke. Q9770 has a lot more voltage than most other Core 2 Quads.

Here's an idea- why don't we just overvolt the i7 by some and run it on load, eh? 80 degrees on 100% load under watercooling for 3.8Ghz, not bad eh. Not even the 65nm Agena at 3.5Ghz produces that kind of heat.

[Shintai]

What a joke. Q9770 has a lot more voltage than most other Core 2 Quads.

Here's an idea- why don't we just overvolt the i7 by some and run it on load, eh? 80 degrees on 100% load under watercooling for 3.8Ghz, not bad eh. Not even the 65nm Agena at 3.5Ghz produces that kind of heat.

http://www.abload.de/img/4ghzb9cy.png



65nm K10 at 3.5=chernobyl

[Helmore]

What a joke. Q9770 has a lot more voltage than most other Core 2 Quads.

Here's an idea- why don't we just overvolt the i7 by some and run it on load, eh? 80 degrees on 100% load under watercooling for 3.8Ghz, not bad eh. Not even the 65nm Agena at 3.5Ghz produces that kind of heat.

You can't really rely on the on-board temperature sensor of a CPU, if this CPU says its running at 50 degrees Celcius and another completely different says its running at 70 degrees, then which one is running at a higher temperature? You probably still don't have a clue, unless you take a properly calibrated temperature sensor to do the measurement of course.

[informal]

Wow did i ask about C2Q Extreme comparison or a normal 2.66Ghz part??Can you read my post at all?
You pick the worst case for C2Q,an EE that has had "issues" with very high power draw at 3+Ghz clocks.Even the QX9650 overclocked (clocks touched,not volts) manages to have a lot lower power draw than a QX9770...

A graph for you.It's from LS:


Watch QX9650 @3Ghz VS Core i7 920 .Core2Q wins in power draw test by a large margin while running at 11% higher clock.

[Bellisimo]

http://www.anandtech.com/cpuchipsets...spx?i=3453&p=3



Its always been a retarded comment with the MCH using ohh so much. And like HT uses alot vs FSB?

Maybe thats why we got huge heatsinks and sometimes fans on PCIe switches alone.

http://www.anandtech.com/cpuchipsets...spx?i=3448&p=9

look at the power consumption at 2.66ghz....

better start pm'ing mods for removing this post again....

btw, its almost logical nehalem would consume more power at same freq/voltage as a C2q, it has more transistors
although intel used static cmos

[Shintai]

http://www.anandtech.com/cpuchipsets...spx?i=3448&p=9

look at the power consumption at 2.66ghz....

better start pm'ing mods for removing this post again....

*cough* performance difference *cough*