Intel Launches Six-Core Microprocessors for High-End Servers.

[@@@@]

http://xbitlabs.com/news/cpu/display/20080915164829_Intel_Launches_Six_Core_Microproces sors_for_High_End_Servers.html

http://www.tgdaily.com/content/view/39324/135/

http://www.infoworld.com/article/08/09/15/Intels_sixcore_Dunnington_chip_hits_the_market-IDGNS_1.html

maybe someone already reviewed them? looking at Movieman :D

[Sly Fox]

Daddy like!! :D:shocked:

[twilyth]

$87,000 for a 32 socket system? I don't know. I was hoping to keep it under $70k.

WTF? $2700 for a 6-core chip? And 1/10th that for a 4-core? What am I missing?

[WeaponX]

Your wallet if you're not careful!

[bowman]

$87,000 for a 32 socket system? I don't know. I was hoping to keep it under $70k.

WTF? $2700 for a 6-core chip? And 1/10th that for a 4-core? What am I missing?

It's definitely not 1/10th of the Tigerton's price.

[twilyth]

Your wallet if you're not careful!

tru dat. But somebody out there is buying all those "male enhancement" products, so I guess this is sort of the same thing. :shrug:

[[XC] riptide]

$87,000 for a 32 socket system? I don't know. I was hoping to keep it under $70k.

WTF? $2700 for a 6-core chip? And 1/10th that for a 4-core? What am I missing?

ya... but just think. wprime destruction at stock clocks.

[twilyth]

riptide;3293752']ya... but just think. wprime destruction at stock clocks.

I love it when you talk dirty. :rofl: :D :eek:

[duploxxx]

xbitlabs just mentioned everything in there last few lines what there has to be said about this chip:

Given that Advanced Micro Devices will be unable to build its own 6-core processor code-named Istanbul until the second half of 2009, Intel will have a very comfortable lead on the market of high-end enterprise servers due to lack of competition. Nevertheless, an advantage that AMD still has is integrated memory controller as well as more advanced bus for its AMD Opteron processors which may translate in better performance scalability of AMD-based MP servers.

its a lead marketing wise for sales.... not for real performance figures, lots of OEM didn't want to sell the 2,93ghz quad because it had no added value on the pumped fsb bus, lets just say this 6 core only makes it worse....

tigertown was dead platform from the moment amd released there barcelona, even with the small 2,3ghz cpu, now with shangai on its way its total ownage. Intel can only attack this with there own 4s nehalem but that is on the roadmap for q32009

[savantu]

xbitlabs just mentioned everything in there last few lines what there has to be said about this chip:

Given that Advanced Micro Devices will be unable to build its own 6-core processor code-named Istanbul until the second half of 2009, Intel will have a very comfortable lead on the market of high-end enterprise servers due to lack of competition. Nevertheless, an advantage that AMD still has is integrated memory controller as well as more advanced bus for its AMD Opteron processors which may translate in better performance scalability of AMD-based MP servers.

its a lead marketing wise for sales.... not for real performance figures, lots of OEM didn't want to sell the 2,93ghz quad because it had no added value on the pumped fsb bus, lets just say this 6 core only makes it worse....

tigertown was dead platform from the moment amd released there barcelona, even with the small 2,3ghz cpu, now with shangai on its way its total ownage. Intel can only attack this with there own 4s nehalem but that is on the roadmap for q32009

Your alternate reality is so far off that it is laughable..Enjoy it !

[qurious63ss]

xbitlabs just mentioned everything in there last few lines what there has to be said about this chip:

Given that Advanced Micro Devices will be unable to build its own 6-core processor code-named Istanbul until the second half of 2009, Intel will have a very comfortable lead on the market of high-end enterprise servers due to lack of competition. Nevertheless, an advantage that AMD still has is integrated memory controller as well as more advanced bus for its AMD Opteron processors which may translate in better performance scalability of AMD-based MP servers.

its a lead marketing wise for sales.... not for real performance figures, lots of OEM didn't want to sell the 2,93ghz quad because it had no added value on the pumped fsb bus, lets just say this 6 core only makes it worse....

tigertown was dead platform from the moment amd released there barcelona, even with the small 2,3ghz cpu, now with shangai on its way its total ownage. Intel can only attack this with there own 4s nehalem but that is on the roadmap for q32009

I take it you haven't seen any of the dunnington benchmarks have you? That dead platform by the way has been eating up alot of the 4P market share from AMD. Shanghai is a shrink of barcelona and the only thing that has been getting owned lately has been AMD.

[m^2]

riptide;3293752']ya... but just think. wprime destruction at stock clocks.

Probably it's not the top for wprime. Under Windows you can use max 60 cores @ 2.66, while with the quad core version: 64 @ 2.93. Better communication between cores might help though.

[KeZzZu]

I take it you haven't seen any of the dunnington benchmarks have you? That dead platform by the way has been eating up alot of the 4P market share from AMD. Shanghai is a shrink of barcelona and the only thing that has been getting owned lately has been AMD.


Please go ahead and show real benches? not just superpi or 3dmark.

ehh yes i know server side. boinc ? cinebench, database benches?

dunno there's lot of benches. but i dont see anything here yet :(

[Zytek_Fan]

Just need to win the lottery... :D

[Jacky]

Please go ahead and show real benches? not just superpi or 3dmark.

ehh yes i know server side. boinc ? cinebench, database benches?

dunno there's lot of benches. but i dont see anything here yet :(
Here you go. Next time you can believe savantu, even though he may be a lil' intel biased.
http://realworldtech.com/forums/index.cfm?action=detail&id=92678&threadid=92678&roomid=2

[qurious63ss]

its a lead marketing wise for sales.... not for real performance figures, lots of OEM didn't want to sell the 2,93ghz quad because it had no added value on the pumped fsb bus, lets just say this 6 core only makes it worse....

tigertown was dead platform from the moment amd released there barcelona, even with the small 2,3ghz cpu, now with shangai on its way its total ownage. Intel can only attack this with there own 4s nehalem but that is on the roadmap for q32009

This was your quote. You are implying that 6 cores will not help because of the FSB. You are implying that Tigerton was "owned" as soon as barcelona was released. Do you have proof of this,

[qurious63ss]

Please go ahead and show real benches? (

Please contact IBM, HP, Cray, Dell, etc. to get the platform that suits your enterprise needs. They will be more then happy to show you some "real benches"

Here is a nice article on what server vendors are saying. ENJOY!

http://online.wsj.com/article/SB122150144796537293.html?mod=googlenews_wsj

[zanzabar]

wait intel has a lead in 4P systems, the core xeons couldent match operon 8xx and they couldent best the phenoms in the 8xxx so did they need 8 more cores to beat amd

and with no QPI thees will be outdated in NOV

[[XC] gomeler]

Uhh... this is old news and a throwback to the s604 server platform which itself is old as the hills. Now this would have been way cool if they had done a 1600FSB 2400MHz hex-core system. Still.. not that impressive but for a select few applications that can't be cloned/distributed amongst many cheap boxes.

[qurious63ss]

gomeler;3294348']Uhh... this is old news and a throwback to the s604 server platform which itself is old as the hills. Now this would have been way cool if they had done a 1600FSB 2400MHz hex-core system. Still.. not that impressive but for a select few applications that can't be cloned/distributed amongst many cheap boxes.

You say that it's a throwback and I would say that it's a great drop in upgrade for current s604 platforms.

[003]

WTF? $2700 for a 6-core chip? And 1/10th that for a 4-core? What am I missing?

All six cores are on a single die.

[Movieman]

wait intel has a lead in 4P systems, the core xeons couldent match operon 8xx and they couldent best the phenoms in the 8xxx so did they need 8 more cores to beat amd

and with no QPI thees will be outdated in NOV

I beg to differ with your statements.:D
I'll toss my 8 core Intel up aganst any Opty 8xx system you want to bring here.
SuperPi
Cinebench
wPrime
encoding app?
Let me know when you have your Opty system ready ok?

[wickedld9]

I beg to differ with your statements.:D
I'll toss my 8 core Intel up aganst any Opty 8xx system you want to bring here.
SuperPi
Cinebench
wPrime
encoding app?
Let me know when you have your Opty system ready ok?

It all depends on what you're doing with it. VM farm or VM benchmarks? Seriously, an 8 core Intel vs. a 32 core Opty???! I might have to unbox the DL785 G5 tomorrow.

[Asgard_thor]

It all depends on what you're doing with it. VM farm or VM benchmarks? Seriously, an 8 core Intel vs. a 32 core Opty???! I might have to unbox the DL785 G5 tomorrow.

bring it AMD FAN boy

GO movieman!

6cores sound great :D

[zanzabar]

I beg to differ with your statements.:D
I'll toss my 8 core Intel up aganst any Opty 8xx system you want to bring here.
SuperPi
Cinebench
wPrime
encoding app?
Let me know when you have your Opty system ready ok?

i remember at school we had a 604 and an optie 8 dual core and the optie smoked the intel, things may be different now with the quads but im sure that the phemons would out perform, once intel gets the HTT (oh wait QPI) then im sure they will be the undisputed server champs

im just not seing the viability no1 would buy servers now unless they had a bunch of deaths

[wickedld9]

bring it AMD FAN boy

GO movieman!

6cores sound great :D

Where in my post do you get that I'm an AMD fan boy?

For the record, I believe MovieMan is talking about a dual Nehalem system which should give the Opty's a run for the money. It all depends on which speed chips he got from Intel.
I've got my benchmarks for the dual Nehalem and Dunnington but just got the 785 a few days ago, still sitting in the box.

[Asgard_thor]

Where in my post do you get that I'm an AMD fan boy?

For the record, I believe MovieMan is talking about a dual Nehalem system which should give the Opty's a run for the money. It all depends on which speed chips he got from Intel.
I've got my benchmarks for the dual Nehalem and Dunnington but just got the 785 a few days ago, still sitting in the box.

it was a joke

[NapalmV5]

:up: :up: these are way more interesting than nehalem crap

if only they would offer these for desktop aswell :(

2 x 6-core @ skulltrail ?

[duploxxx]

I take it you haven't seen any of the dunnington benchmarks have you? That dead platform by the way has been eating up alot of the 4P market share from AMD. Shanghai is a shrink of barcelona and the only thing that has been getting owned lately has been AMD.

no haven't seen those synthetic test result only had the dunnington here at my desk end of july for testing from large OEM. And we are talking about 4S here... not your plain daily desktop all of you on XS are thinking about.

some synthetic vm bench results....
http://www.vmware.com/products/vmmark/results.html

a 4s 2,5ghz quad opty is faster then a 4s 2,93 tiger... not to mention what difference shanghai will bring here and yes the dunnington gives a better score here, but i did the tests already with dunnington in real world VM apps its not so good if they have to talk much outside there own cores. And that's the big issue in VM, you can't control affinity unless you want to loose some functionality.

Your alternate reality is so far off that it is laughable..Enjoy it !

you probably have never worked on a 4s machine did you :confused:

I beg to differ with your statements.:D
I'll toss my 8 core Intel up aganst any Opty 8xx system you want to bring here.
SuperPi
Cinebench
wPrime
encoding app?
Let me know when you have your Opty system ready ok?


perhaps something real world in stead of some stupid benches and synthetic apps?

lets create a 7-8 VM 2-4CPU /4-6gb ram environment with database around 80-100GB and a total throughput of 5-600MB/s network and 4-500MB/s storage, lets see how your toss will go against an opty 4s system...

[informal]

I found some interesting information regarding the TDP of the X7460 6 core part here (http://download.intel.com/design/xeon/datashts/32033501.pdf).

This is similar to ACP rating of Opterons.We see new 6 core part can pull 170W in some border line case(max. power) but it's rated at 130W(ACP-like) due to:
Analysis indicates that real applications are unlikely to cause the processor to consume
maximum power dissipation for sustained time periods. Intel recommends that
complete thermal solution designs target the Thermal Design Power (TDP) instead of
the maximum processor power consumption. The Intel Thermal Monitor feature is
intended to help protect the processor in the event that an application exceeds the TDP
recommendation for a sustained time period.

The Intel® Xeon® Processor E7400 Series (see Figure 7-2; Table 7-3) and Intel®
Xeon® Processor L7400 Series (see Figure 7-4; Table 7-5) supports a single Thermal
Profile. The Thermal Profiles are indicative of a constrained thermal environment (Ex:
1U form factor). Because of the reduced cooling capability represented by this solution,
the probability of TCC activation and performance loss is increased. Additionally,
utilization of a thermal solution that does not meet the Thermal Profile will violate the
thermal specifications and may result in permanent damage to the processor. Refer to
the Intel® Xeon® Processor 7400 Series Thermal Mechanical Design Guide for details
on system thermal solution design, thermal profiles and environmental considerations.
The upper point of the thermal profile consists of the Thermal Design Power (TDP) and
the associated TCASE value. It should be noted that the upper point associated with
Intel® Xeon® X7460 Processor Thermal Profile (x = TDP and y = TCASE_MAX P @ TDP)
represents a thermal solution design point. In actuality the processor case temperature
will not reach this value due to TCC activation (see Figure 7-1 for the Intel® Xeon®
X7460 Processor).
Analysis indicates that real applications are unlikely to cause the processor to consume
maximum power dissipation for sustained time periods. Intel recommends that
complete thermal solution designs target the Thermal Design Power (TDP) instead of
the maximum processor power consumption. The Intel Thermal Monitor feature is
intended to help protect the processor in the event that an application exceeds the TDP
recommendation for a sustained time period. For more details on this feature, refer to
Section 7.2. To ensure maximum flexibility for future requirements, systems should be
designed to the Flexible Motherboard (FMB) guidelines, even if a processor with lower
power dissipation is currently planned. Intel Thermal Monitor or Intel Thermal
Monitor 2 feature must be enabled for the processor to remain within its
specifications.

[Shintai]

And for Informal..no..its not like ACP....Also if you actually read the paper you cant exceed the TDP. Since it will throttle.

The Intel Thermal Monitor feature is intended to help protect the processor in the event that an application exceeds the TDP recommendation for a sustained time period.

For short time its common to draw more power.

Also since you dont understand ACP obviously. Its what AMD says the CPU will use in average. Not peak. They still give the TDP numbers to OEMs and such for the cooling solution.

[informal]

Like a clockwork Shintai..Priceless.

Anyhow,the max. power cons. for the the said 6 core part obviously isn't 130W.It is 170W and the chip will throttle for some period of time if it exceeds the 130W mark(which means with lower clock it will perform slower than at its rated clock speed).This is unlike ACP which means it is a common and expected power draw and if it was exceeded chip will not throttle and reduce performance AFAIK.From Phenom reviews we know that the TDP rating is overrated a bit(140W is around 100-110 for 9950 ie.-Lost Circuits,thus confirming the ACP rating works for desktop parts too but is not used).

So you have 2 cases if the TDP ratings are exceeded: throttle and stay below the average TDP mark in certain situations and reduce perf.(intc) or just draw more power if ACP is exceeded and retain the perf.(amd),with no throttling until max. TDP is exceeded(which is likely never to happen since it's max. current x rated Vcore;throttling can happen if cooling device fails and critical temp. is reached).

[Shintai]

Like a clockwork Shintai..Priceless.

Keep your personal attitude problems far away. Unless thats all you can do...

Anyhow,the max. power cons. for the the said 6 core part obviously isn't 130W.It is 170W and the chip will throttle for some period of time if it exceeds the 130W mark(which means with lower clock it will perform slower than at its rated clock speed).This is unlike ACP which means it is a common and expected power draw and if it was exceeded chip will not throttle and reduce performance AFAIK.From Phenom reviews we know that the TDP rating is overrated a bit(140W is around 100-110 for 9950 ie.-Lost Circuits,thus confirming the ACP rating works for desktop parts too but is not used).

So you have 2 cases if the TDP ratings are exceeded: throttle and stay below the average TDP mark in certain situations and reduce perf.(intc) or just draw more power if ACP is exceeded and retain the perf.(amd),with no throttling until max. TDP is exceeded(which is likely never to happen since it's max. current x rated Vcore;throttling can happen if cooling device fails and critical temp. is reached).

You also dont seem to understand what you read. Basicly no application yet created (Even torture tests) can push it beyond 130W, but it is technical possible to make it consume 170W.

[informal]

That wasn't anything personal.Your humor level is low or what?

The delta is quite large between the 130 and 170W(30%).I didn't say it will even consume the 130W.It probably won't.But this is a first time i see the Xeon paper listing the max. TDP,which i may missed in the past.

[savantu]

You're funny Shintai, you can't refute duploxx's statements, so you just pretend that he's making things up? How's it like in your fantasyland, do they hand out free meatballs and ice-cream at mealtime?

A link was posted previously that showed Dunnington based servers scoring world records in a lot of industry standard benchmarks by huge margins versus the competition.

It's up to Duploxx to bring up evidence ( that means numbers not his fantasies in written mode ) to the contrary.

Necessitas probandi incumbit ei qui agit. - ring any bells ?

[duploxxx]

A link was posted previously that showed Dunnington based servers scoring world records in a lot of industry standard benchmarks by huge margins versus the competition.

It's up to Duploxx to bring up evidence ( that means numbers not his fantasies in written mode ) to the contrary.

Necessitas probandi incumbit ei qui agit. - ring any bells ?

right and those industry standard benchmarks also give a clear lead to the tigertown 2,93ghz quad against the barcelona 2,5ghz, but yet semi synthetic application tests like the VMmark already provide the opposite .... not to mention real world testing... done my share already on sql/oracle databases and data throughput to test all different platforms. As i said cpu might be good but whole platform is not up to it.

like i said before fouls follow marketing.

[savantu]

That wasn't anything personal.Your humor level is low or what?

The delta is quite large between the 130 and 170W(30%).I didn't say it will even consume the 130W.It probably won't.But this is a first time i see the Xeon paper listing the max. TDP,which i may missed in the past.

Happy as the little kid that discovered his shadow ?

Unlike you , everyone here knew that Intel's TDP is not maximum power draw ; but the highest power than can be reasonably expected while running a wide array of commercial apps.As a result , there is a delta ( always has been ) between published TDP ( which is the reference number ) and maximum power draw which happens in 2 ways :
-transient and doesn't matter , these are spikes lasting a few microseconds
-under power virus

It is almost impossible in real world to exceed the TDP while running commercial software.

AMD measures its TDP nowadays just like Intel , based on what can be expected by running normal apps.

ACP however is completely different from TDP since it measures average power , it might be good to calculate ones electricity bill but has no relevance in the sense TDP has.

You can go into your corner now.

[savantu]

right and those industry standard benchmarks also give a clear lead to the tigertown 2,93ghz quad against the barcelona 2,5ghz, but yet semi synthetic application tests like the VMmark already provide the opposite .... not to mention real world testing... done my share already on sql/oracle databases and data throughput to test all different platforms. As i said cpu might be good but whole platform is not up to it.

like i said before fouls follow marketing.

Post numbers ; what are your credentials or do you expect to be taken at face value ?

One can come and claim the exact opposite.Without numbers or hard evidence , you can't be expect to be trusted while the server makers wet their pants for the new parts.
AMD's dwindling share in large servers is a clear example that your claims are full of hot air.While there might be apps where K10 based servers beat equivalent Xeons due to IMC and HT , those are mainly FP heavy apps used in HPC.The large majority of commercial apps aren't FP heavy.


TPC-E
Dell PowerEdge R900, 4S/24C/24T, Intel Xeon X7460, 2.66GHz, Ws2008, MS-SQL 2008, 64 GB, 344 HDs - 671.35 tpsE at 500.55 $/tpsE
http://www.tpc.org/results/individual_results/Dell/dell_R900_Dun_2008_081908_TPCE_es.pdf

For comparison:
Dell PowerEdge R900, 4S/16C/16T, Intel Xeon X7350, 2.93GHz, Ws2008, MS-SQL 2008, 64 GB, 344 HDs - 451.29 tpsE at 734.25 $/tpsE
IBM System x3850 M2, 4S/16C/16T, Intel Xeon X7350, 2.93GHz, Ws2008, MS-SQL 2008, 128 GB, 384 HDs - 479.51 tpsE at 1,591.20 $/tpsE
IBM System x3950 M2, 8S/32C/32T, Intel Xeon X7350, 2.93GHz, Ws2008, MS-SQL 2008, 256 GB, 544 HDs - 804.00 tpsE at 1,450.05 $/tpsE

TPC-C
HP ProLiant DL580G5, 4S/24C/24T, Intel Xeon X7460, 2.66GHz, Ws2008, MS-SQL 2008, 256 GB, 1034 HDs - 634,825 tpmC at 1.10 $/tpmC
IBM System x3950 M2, 8S/48C/48T, Intel Xeon X7460, 2.66GHz, RHEL 5.2, DB2 ESE 9.5, 512 GB, 1931 HDs - 1,200,632 tpmC at 1.99 $/tpmC
http://www.tpc.org/results/individual_results/HP/HP_ProLiant_DL580G5_2.67GHz_es.pdf
http://www.tpc.org/results/individual_results/IBM/IBM-x3950M2-Linux-DB2-ES.pdf

For comparison:
HP ProLiant DL580G5, 4S/16C/16T, Intel X7350, 2.93GHz, Ws2005, MS-SQL 2003, 128 GB, 1048 HDs - 407,079 tpmC at 1.71 $/tpmC
HP ProLiant DL585G5, 4S/16C/16T, AMD Opteron 8360SE, 2.50GHz, Ws2005, MS-SQL 2003, 256 GB, 725 HDs - 471,883 tpmC at 1.17 $/tpmC
IBM System x3850 M2, 4S/16C/16T, Intel Xeon X7350, 2.93GHz, RHEL 5, DB2 9.5 EE, 256 GB, 775 HDs - 516,752 tpmC at 2.59 $/tpmC
IBM Power 550, 4S/8C/16T, IBM Power6, 4.2 GHz, AIX 5L V5.3, DB2 9.5 EE, 256 GB, 1553 HDs - 629,159 tpmC at 2.49 $/tpmC

IBM System x3950 M2, 8S/32C/32T, Intel Xeon X7350, 2.93GHz, Ws2003, MS-SQL 2005, 512 GB, 1361 HDs - 841,809 tpmC at 3.46 $/tpmC
IBM System p5 570, 8S/16C/32T, IBM Power5+, 2.2 GHz, AIX 5L V5.3, DB2 UDB 8.2, 512 GB, 1694 HDs - 1,025,170 tpmC at 4.42 $/tpmC
IBM System p 570, 8S/16C/32T, IBM Power6, 4.7 GHz, AIX 5L V5.3, DB2 Enterprise 9, 768 GB, 3482 HDs - 1,616,162 tpmC at 3.54 $/tpmC

A full list of benchmarks can be seen here :

http://www.intel.com/performance/server/xeon_mp/summary.htm?iid=perf_server+mp_sum

And regarding virtualization , while K10 beat Tigerton , the situation reversed with Dunnignton :

http://www.intel.com/performance/server/xeon_mp/virtualization.htm?iid=perf_server_lhn+mp_vt

[Shintai]

no haven't seen those synthetic test result only had the dunnington here at my desk end of july for testing from large OEM. And we are talking about 4S here... not your plain daily desktop all of you on XS are thinking about.

some synthetic vm bench results....
http://www.vmware.com/products/vmmark/results.html

a 4s 2,5ghz quad opty is faster then a 4s 2,93 tiger... not to mention what difference shanghai will bring here and yes the dunnington gives a better score here, but i did the tests already with dunnington in real world VM apps its not so good if they have to talk much outside there own cores. And that's the big issue in VM, you can't control affinity unless you want to loose some functionality.



you probably have never worked on a 4s machine did you :confused:




perhaps something real world in stead of some stupid benches and synthetic apps?

lets create a 7-8 VM 2-4CPU /4-6gb ram environment with database around 80-100GB and a total throughput of 5-600MB/s network and 4-500MB/s storage, lets see how your toss will go against an opty 4s system...

Are we looking on the same thing here?

http://www.vmware.com/files/pdf/vmmark/vmmark_hp3.pdf
This is a 2.83Ghz Xeon system..and its the fastest 16 core setup.

Also with your "godly" server knowledge that the rest of us only can bow in awe for. Perhaps you should check the storage solutions and such for the systems aswell. The CPUs and Memory only do so little in an I/O congested system...

[duploxxx]

Are we looking on the same thing here?

http://www.vmware.com/files/pdf/vmmark/vmmark_hp3.pdf
This is a 2.83Ghz Xeon system..and its the fastest 16 core setup.

Also with your "godly" server knowledge that the rest of us only can bow in awe for. Perhaps you should check the storage solutions and such for the systems aswell. The CPUs and Memory only do so little in an I/O congested system...

sure you ever worked on server platforms? maybe u should read some more, might help a bit on knowledge. btw we only use EVA storage and each storage environment has its own blade enclosure, so no wrong again.

looks like you never looked at a Vmmark before, read again... top 3 on 16 cores are HP585 + dell M/R905 not sure if you are aware of OEM server branding but those look AMD systems to me :D

then look at the actual performance in each tile, for example between the 24 core r900 and 16 core 585 you will see that the performance/tile is way higher on the amd and that the other system just scores higher because it is able to run more tiles. Its only in mail server that the 4s intel is able to keep up overtake the amd all other tiles are clearly much faster on the amd system and that is what you see in real life virtualisation testing.
http://www.vmware.com/files/pdf/vmmark/vmmark_hp3.pdf
http://www.vmware.com/files/pdf/vmmark/vmmark_dell_R900.pdf

so concluesion, when to buy dunnington for virtualisation, when you have a huge amount of small Vm's that just need to be there but don't require high resources, all other apps, just take the AMD. System performance and accessibility is way better on those 4s barcelona systems.

[Shintai]

sure you ever worked on server platforms? maybe u should read some more, might help a bit on knowledge, we only use EVA storage and each storage environment has its own blade enclosure, so no wrong again.

looks like you never looked at a Vmmark before, read again... top 3 on 16 cores are HP585 + dell M/R905 not sure if you are aware of OEM server branding but those look AMD systems to me :D

Ye ok, I mixed its clients. Anyway. So if we aint doing synthetic virtualization benches. What else you got? Savantu´s numbers are still something you need to prove wrong..or beaten.

Oh and btw..I/O is still QUITE different.

http://www.vmware.com/files/pdf/vmmark/vmmark_hp3.pdf
http://www.vmware.com/files/pdf/vmmark/vmmark_hp5.pdf

AMD system uses 5 HBAs, no switch. the Intel system uses 1 HBA and a switch. Just as the most obvious.

AMD uses a 3400 series switch, Intel a 2900. Different NIC conf etc.

Gotta love those rigth? :D

How could that escape your trained eye? :p:

Would be a shame if things was actually tested with somewhat similar conf.

[[cTx]Riggs]

Are these native 6 cores?

[Shintai]

Are these native 6 cores?

Yup

[twilyth]

For the people who are saying that Intel is thrashing AMD in terms of market share, take a look at this (http://www.idc.com/getdoc.jsp?containerId=prUS21263208).

In terms of processor vendor shares, Intel experienced a small gain over AMD in the first quarter. On an overall unit basis, Intel earned 78.9% market share, a gain of 2.2%. AMD earned 20.9% market share, a loss of 2.2%.

Similar shifts in market share were evident by form factor. In the mobile processor segment, Intel earned 85.3% market share, a gain of 3.4%, and AMD captured 14.5%, a loss of 3.3%. In the PC server processor segment, Intel earned 87.0% share, a gain of 1.6%, while AMD earned 13.0%, a loss of 1.6% share. Finally, in the desktop PC processor segment, Intel captured 73.1% share, an increase of 1.0%, and AMD earned 26.7% share, a loss of 1.0%.

I don't consider those numbers as constituting a thrashing especially since we still need to see how Shanghai will compare to Nehalem.

[Chad Boga]

For the people who are saying that Intel is thrashing AMD in terms of market share, take a look at this (http://www.idc.com/getdoc.jsp?containerId=prUS21263208).



I don't consider those numbers as constituting a thrashing especially since we still need to see how Shanghai will compare to Nehalem.
Well there is market share and then there is profit share.

[Shintai]

For the people who are saying that Intel is thrashing AMD in terms of market share, take a look at this (http://www.idc.com/getdoc.jsp?containerId=prUS21263208).



I don't consider those numbers as constituting a thrashing especially since we still need to see how Shanghai will compare to Nehalem.

Compare the server numbers with the desktop numbers. I would say its a trashing. Also 4S servers might be 1% of the entire serversegment numbers.

[Hornet331]

plus thats only for Q108, go back to Q306 and look up how much server share amd lost from there to now (a hint in 2006 server ms was above 17% ;) )

[BrowncoatGR]

Post numbers ; what are your credentials or do you expect to be taken at face value ?

One can come and claim the exact opposite.Without numbers or hard evidence , you can't be expect to be trusted while the server makers wet their pants for the new parts.
AMD's dwindling share in large servers is a clear example that your claims are full of hot air.While there might be apps where K10 based servers beat equivalent Xeons due to IMC and HT , those are mainly FP heavy apps used in HPC.The large majority of commercial apps aren't FP heavy.


TPC-E
Dell PowerEdge R900, 4S/24C/24T, Intel Xeon X7460, 2.66GHz, Ws2008, MS-SQL 2008, 64 GB, 344 HDs - 671.35 tpsE at 500.55 $/tpsE
http://www.tpc.org/results/individual_results/Dell/dell_R900_Dun_2008_081908_TPCE_es.pdf

For comparison:
Dell PowerEdge R900, 4S/16C/16T, Intel Xeon X7350, 2.93GHz, Ws2008, MS-SQL 2008, 64 GB, 344 HDs - 451.29 tpsE at 734.25 $/tpsE
IBM System x3850 M2, 4S/16C/16T, Intel Xeon X7350, 2.93GHz, Ws2008, MS-SQL 2008, 128 GB, 384 HDs - 479.51 tpsE at 1,591.20 $/tpsE
IBM System x3950 M2, 8S/32C/32T, Intel Xeon X7350, 2.93GHz, Ws2008, MS-SQL 2008, 256 GB, 544 HDs - 804.00 tpsE at 1,450.05 $/tpsE

TPC-C
HP ProLiant DL580G5, 4S/24C/24T, Intel Xeon X7460, 2.66GHz, Ws2008, MS-SQL 2008, 256 GB, 1034 HDs - 634,825 tpmC at 1.10 $/tpmC
IBM System x3950 M2, 8S/48C/48T, Intel Xeon X7460, 2.66GHz, RHEL 5.2, DB2 ESE 9.5, 512 GB, 1931 HDs - 1,200,632 tpmC at 1.99 $/tpmC
http://www.tpc.org/results/individual_results/HP/HP_ProLiant_DL580G5_2.67GHz_es.pdf
http://www.tpc.org/results/individual_results/IBM/IBM-x3950M2-Linux-DB2-ES.pdf

For comparison:
HP ProLiant DL580G5, 4S/16C/16T, Intel X7350, 2.93GHz, Ws2005, MS-SQL 2003, 128 GB, 1048 HDs - 407,079 tpmC at 1.71 $/tpmC
HP ProLiant DL585G5, 4S/16C/16T, AMD Opteron 8360SE, 2.50GHz, Ws2005, MS-SQL 2003, 256 GB, 725 HDs - 471,883 tpmC at 1.17 $/tpmC
IBM System x3850 M2, 4S/16C/16T, Intel Xeon X7350, 2.93GHz, RHEL 5, DB2 9.5 EE, 256 GB, 775 HDs - 516,752 tpmC at 2.59 $/tpmC
IBM Power 550, 4S/8C/16T, IBM Power6, 4.2 GHz, AIX 5L V5.3, DB2 9.5 EE, 256 GB, 1553 HDs - 629,159 tpmC at 2.49 $/tpmC

IBM System x3950 M2, 8S/32C/32T, Intel Xeon X7350, 2.93GHz, Ws2003, MS-SQL 2005, 512 GB, 1361 HDs - 841,809 tpmC at 3.46 $/tpmC
IBM System p5 570, 8S/16C/32T, IBM Power5+, 2.2 GHz, AIX 5L V5.3, DB2 UDB 8.2, 512 GB, 1694 HDs - 1,025,170 tpmC at 4.42 $/tpmC
IBM System p 570, 8S/16C/32T, IBM Power6, 4.7 GHz, AIX 5L V5.3, DB2 Enterprise 9, 768 GB, 3482 HDs - 1,616,162 tpmC at 3.54 $/tpmC

A full list of benchmarks can be seen here :

http://www.intel.com/performance/server/xeon_mp/summary.htm?iid=perf_server+mp_sum

And regarding virtualization , while K10 beat Tigerton , the situation reversed with Dunnignton :

http://www.intel.com/performance/server/xeon_mp/virtualization.htm?iid=perf_server_lhn+mp_vt

The Opteron system is running different software. The results are therefore not comparable. Why dont you also list total cost as well? Dunnington costs quite a bit more

[gallag]

I have found on paper evidence that Intel is a power hog, Real world does not matter as this is a wee bit of ammo to make Intel look bad.

blah blah I found one test where intel is bandwidth limited so I am going to harp on like it is the most important thing in the world blah blah

Why you guys always raging on Intel threads. :ROTF:

[Movieman]

i remember at school we had a 604 and an optie 8 dual core and the optie smoked the intel, things may be different now with the quads but im sure that the phemons would out perform, once intel gets the HTT (oh wait QPI) then im sure they will be the undisputed server champs

im just not seing the viability no1 would buy servers now unless they had a bunch of deaths

See, now here's the difference.
I "don't remember from school."
I have a DP socket 604 Intel system here that I built with top parts back in August 2005..3600/2mb/800 Irwindales,Supermicro X6DA8-G2 MB
Corsair/Viking DDR2-400 ECC REG memory. Nothing but the best at the time.
I know what that system does and will not do as it was my "daily driver" for 2+ years.
My current Harpertown system here has SIX times the computational power of that DX xeon rig.
It will destroy a 16 core Opteron rig and that's no BS.
32 cores is asking a bit much but I'll plant money that a 24 core Dunnington will outdo a 32 core Opteron and EASILY!:D

[duploxxx]

Why you guys always raging on Intel threads. :ROTF:

where do you think 4s servers are mainly used for? so if you buy one you better be sure you buy it the correct one for your needs.

See, now here's the difference.
I "don't remember from school."
I have a DP socket 604 Intel system here that I built with top parts back in August 2005..3600/2mb/800 Irwindales,Supermicro X6DA8-G2 MB
Corsair/Viking DDR2-400 ECC REG memory. Nothing but the best at the time.
I know what that system does and will not do as it was my "daily driver" for 2+ years.
My current Harpertown system here has SIX times the computational power of that DX xeon rig.
It will destroy a 16 core Opteron rig and that's no BS.
32 cores is asking a bit much but I'll plant money that a 24 core Dunnington will outdo a 32 core Opteron and EASILY!:D

with your synthetic benches.... might be, I'll send you a pm when my 4s shanghai test unit arrives within a month :D only have 2s hyper + barca and 4s tiger at the moment