Socket 1567 Xeon Nehalem-EX

[duploxxx]

Another wide CPU, huh?
Interesting, not all the elements are under the heat spreader...

that is beckton, paper launch on 30 March and availability expected somewhere in June.

The elements under the heat spreader are fb buffers, while they used to have fbdimm they now moved the buffers to the mobo to reduce the dimm cost and can work with legacy Rdimm ddr3. But now you get a very expensive mobo and some additional HOT spots as a trade off for some additional RAM features.

price will decide if this system will shine or not.

So if I do the math right, you're saying that 256GB = 100W. So if they shut down the rest of the memory, that would be 800W for CPUs, fans, power and a drive or two?

Wow, that is a power-hungry CPU.

they use max pc8500 ddr3 rdimm which consumes about 2.1W idle and 3.5W high load no LP, we don't know the amount of dimms in there and we have no idea what the buffers will consume, although my guess is about 4-5W idle and upto 8-10W load looking at those heatsinks. In the end whatever we try to figure out what exactly remains on or off, the fact indeed remains that there is still a lot of power consumed, on a not so high loaded system

[Nedjo]

well, let s be gentlemen ... you do the math on your own processors, and you avoid to mess up the math on mine. Thanks!

Oh now only Intel's employes have exclusive rights to comment their product on public places like forums?

And when it comes to YOUR math, we all know that you do have special TDP math that only applies to Intel processors, so instead of 800W or 200W per CPU you'll say that TDP of these behemoths is actually 150W

So your "advice" to JF was in place - only Intel holds rights to Intel's TDP math!

BTW
I hope that you're not Rolling On The Floor so much in some data-center filled with cables

Your Timex watch is using less power than YOUR processor, it does not mean Timex CPU is a good server chip. Thanks!

Well apparently you think that 200W CPU that requires new socket on the new $XXXX mobo is good server chip!

You might not know this but I have no doubt that your business managers do know - This is just paper-launch chip useful only for pissing contest on forums, not something that conscious corporate buyer will consider, taking in to account power bills, infrastructure price and probably something else I can't think of it

On Xtremesystem, it is about horse power, world grid as fast as you can with awesome efficiency ... but 1st, you want to top of the list ...

Apparently you somehow selectively filler all those posts about power consumption of GPUs and CPUs... so yeah it is about the power, but it's also about power consumed

[Sn0wm@n]

45nm 8 cores 16 threads... ouch about the tdp.... must be near 200w if it runs at some decent clock .....

[Micutzu]

Even this is entertaining for us, come on guys, don't use flawed technical arguments. I said in a previous post that if shutting down half the memory reduced the power from 1000W to 900W, shutting down the rest won't reduce it by just another 100W but more, because idle memory consumes much less than active memory. Other than that, let's assume a system with 4 CPU's, a server motherboard, a few drives and a few fans consume 800W. A SCSI drive takes up to 20W, a high-power fan up to 15W, substract a few of each from the 800, substract a few more W for the various motherboard integrated stuff, then multiply everything with 0.9 (wich is a fair efficiency figure for the CPU PWM circuit), and you might end up just where you should.

On the other hand, in the testing done by us and also other hardware review websites, i haven't seen figures for the power consumption that would be far off the rated TDP.

[gallag]

Yet another Intel thread ran down by AMD marketing, Come on.

[saaya]

socket 1567
Next year, Intel bring socket 2011 for sever segment !
Now we buy Nehalem EX system, next year we can't update CPU from Nehalem to Sandy Bridge !

wow is that true? socket 1567 is a dead end?

well, let s be gentlemen ... you do the math on your own processors, and you avoid to mess up the math on mine. Thanks!
Your Timex watch is using less power than YOUR processor, it does not mean Timex CPU is a good server chip. Thanks!

On Xtremesystem, it is about horse power, world grid as fast as you can with awesome efficency ... but 1st, you want to top of the list ...

whats the big deal? its hot, yes, but its fast... you always have to balance the two and every chip will be different... and francois, as long as you guys dont try to hammer this thing into a laptop, i dont think anybody cares about the heat hahahah

at least here on xs...

3 power phases per cpu looks low indeed...
loooots of io on that board... pciE, memory... very impressive...that is one huge motherboard... :o

[Katanai]

I'm kinda curious how this will turn out. By the way I see it this platform sits between the current x86 setups and the Itanium systems. I think it will do well but only time will tell. HP has some nice systems coming up on this platform, on which I cannot really say much, other than they are true beasts!!! Will try to post some info on XS when the NDA lifts on them...

[ajaidev]

socket 1567
Next year, Intel bring socket 2011 for sever segment !
Now we buy Nehalem EX system, next year we can't update CPU from Nehalem to Sandy Bridge !

Not totally true

Also take a look at this

[duploxxx]

Yet another Intel thread ran down by AMD marketing, Come on.

as if you are any different in other threads

Even this is entertaining for us, come on guys, don't use flawed technical arguments. I said in a previous post that if shutting down half the memory reduced the power from 1000W to 900W, shutting down the rest won't reduce it by just another 100W but more, because idle memory consumes much less than active memory. Other than that, let's assume a system with 4 CPU's, a server motherboard, a few drives and a few fans consume 800W. A SCSI drive takes up to 20W, a high-power fan up to 15W, substract a few of each from the 800, substract a few more W for the various motherboard integrated stuff, then multiply everything with 0.9 (wich is a fair efficiency figure for the CPU PWM circuit), and you might end up just where you should.

On the other hand, in the testing done by us and also other hardware review websites, i haven't seen figures for the power consumption that would be far off the rated TDP.

Fully agree, but then again the system was more or less idle, only running few VM's, perhaps they totally disabled any other power saving feature in the bios.

I really wonder if there are already so many review sites that have Nehalem EX for testing purpose, highly doubt it

I'm kinda curious how this will turn out. By the way I see it this platform sits between the current x86 setups and the Itanium systems. I think it will do well but only time will tell. HP has some nice systems coming up on this platform, on which I cannot really say much, other than they are true beasts!!! Will try to post some info on XS when the NDA lifts on them...

looking at the HP lineup you don't need a wizard to try and find out what types they will bring with EX.. very obvious, those "beasts" are already existing yet off course with different chipsets and architectures, EX will just replace existing intel 4s+ offerings and create new ones where these days only AMD was offered within the HP portfolio.

[gallag]

as if you are any different in other threads

Show me?? I could show you a lot more where you are worse, I also praise AMD products when they deserve it.

[Micutzu]

...
I really wonder if there are already so many review sites that have Nehalem EX for testing purpose, highly doubt it
...

Nedjo was talking generally about how Intel rates their processor TDP, and so was I. Plus, I'm thinking you can't fool around with ratings for the kind of customers the EX is targeted at, 30% more power consumption per CPU means much more costs overall in that enviroment.

[Hornet331]

Regarding power consumption, I'll wait for real numbers (review or spec). Total power consumption of a system is quite meaningless wihtout knowing what performance it does deliver.

Also I guess people and companies really considering such systems prioritize other things, like higher reliability or on the fly error correction, rather then power conumption.

[Movieman]

Oh now only Intel's employes have exclusive rights to comment their product on public places like forums?

And when it comes to YOUR math, we all know that you do have special TDP math that only applies to Intel processors, so instead of 800W or 200W per CPU you'll say that TDP of these behemoths is actually 150W

So your "advice" to JF was in place - only Intel holds rights to Intel's TDP math!

BTW
I hope that you're not Rolling On The Floor so much in some data-center filled with cables


Well apparently you think that 200W CPU that requires new socket on the new $XXXX mobo is good server chip!

You might not know this but I have no doubt that your business managers do know - This is just paper-launch chip useful only for pissing contest on forums, not something that conscious corporate buyer will consider, taking in to account power bills, infrastructure price and probably something else I can't think of it

Apparently you somehow selectively filler all those posts about power consumption of GPUs and CPUs... so yeah it is about the power, but it's also about power consumed

Before I clean the living crap out of this thread I want to say to you that this time Francois was right and fair in what he said.
He was replying to JF-AMD's comment and he works for AMD.
So do you from what I've been told.
XS is NOT the place for company employees to drop trou and have a "Whose is bigger" biatching match.
As I said politely earlier last night to one gentleman from a company: Tout your own product, don't dump on the other guy's.
I've seen this crap go on since I got here over 4 years ago.
When the hell are you children going to grow up?
It's a frigging processor not life or death and no matter how much either company dumps on the other it won't make a tinker's dam in your sales.
The public will decide that.
Honest to God, I'd like to reach out and slap some damned sense into some of you and it's always the same people.I think I'll grab some coffee before I lose my temper and start handing out bans!

[Movieman]

Thread reopened but I'll say it plain, keep this crap up and I won't say another thing. I'll just ban the guilty parties and it won't be for a week.
Act like adults please and leave the fanboyism where it belongs, in the barrel.

[Hornet331]

ouch.. has been a long time since I'v see you so pissed movieman.

[-Sweeper_]

socket 1567
Next year, Intel bring socket 2011 for sever segment !
Now we buy Nehalem EX system, next year we can't update CPU from Nehalem to Sandy Bridge !

i thought socket 2011 was 1366 replacement, not 1567....

[savantu]

Nedjo was talking generally about how Intel rates their processor TDP, and so was I. Plus, I'm thinking you can't fool around with ratings for the kind of customers the EX is targeted at, 30% more power consumption per CPU means much more costs overall in that enviroment.

Both Intel and AMD apply the same principles in measuring TDP, there's no measurable difference between the 2.
The TDP for both parties is defined as " max power while running power hungry commercial apps ". The cooling solutions are designed to handle this.

CPUs from both companies can exceed this TDP in special cases like a thermal virus. Typically TDP is reached while running Linpack, everything else drops below it.

So how exactly is Intel's TDP misleading as you imply ?

What is truly misleading is ACP; a pure marketing invention that nobody who actually uses the systems cares about. It is basically a weighted mean given that CPUs also spend time idle. Which is pure bull because :
-cooling designers need to target TDP, cannot assume idle time in utilization
-users need to assign both power and cooling based on TDP, since you need to cover the peak

[p2501]

Now, where's one 8-socket test system for each dedicated WCG/Folder on XS? ME WANTS!

[ajaidev]

ouch.. has been a long time since I'v see you so pissed movieman.

It must have touched his blue heart

Just kidding pls dont ban me...

[JF-AMD]

Both Intel and AMD apply the same principles in measuring TDP, there's no measurable difference between the 2.
The TDP for both parties is defined as " max power while running power hungry commercial apps ". The cooling solutions are designed to handle this.

CPUs from both companies can exceed this TDP in special cases like a thermal virus. Typically TDP is reached while running Linpack, everything else drops below it.

So how exactly is Intel's TDP misleading as you imply ?

What is truly misleading is ACP; a pure marketing invention that nobody who actually uses the systems cares about. It is basically a weighted mean given that CPUs also spend time idle. Which is pure bull because :
-cooling designers need to target TDP, cannot assume idle time in utilization
-users need to assign both power and cooling based on TDP, since you need to cover the peak

It is far more complicated than you think. Our TDP is max power, which is the most power that the processor can consume. For Intel, this is "max power", not TDP.

Real power at the wall is what will matter. For example:

Xeon X5570:
TDP: 95W
Max power: 155W (sustained)
Highest power 197 (I would generally ignore this number, but it is an interesting data point)
http://www.cpu-world.com/CPUs/Xeon/I...602X5570).html
Power at the wall @100% utilization: ~254W

Opteron 2435:
ACP: 75W
TDP: 115W (this is max power)
http://www.cpu-world.com/CPUs/K10/AM...S6DGNWOF).html
Power at the wall: ~258W

Power at the wall for both (I just grabbed the first result, thus the ~):
http://www.spec.org/power_ssj2008/re...r_ssj2008.html

So, what you really see, is at the wall, you have to plan for about the same amount of power for both. So you can't compare TDP only (95W to 115W).

Nor, can you compare our ACP to their TDP because again there is a 20W delta.

What you really need to look at is the power at the wall, and you can see that they are both about the same.

That is a pure power discussion. You could layer in performance to try to argue it one way or the other, but you have to be really careful in doing that as both products are going to rev in 2 weeks. Unless you know where both are going to end up you might not want to put your stake in the ground just yet. (also don't forget that this is a $989 part vs. a $1386 part)

[Drwho?]

It is far more complicated than you think. Our TDP is max power, which is the most power that the processor can consume. For Intel, this is "max power", not TDP.

Real power at the wall is what will matter. For example:

Xeon X5570:
TDP: 95W
Max power: 155W (sustained)
Highest power 197 (I would generally ignore this number, but it is an interesting data point)
http://www.cpu-world.com/CPUs/Xeon/I...602X5570).html
Power at the wall @100% utilization: ~254W

Opteron 2435:
ACP: 75W
TDP: 115W (this is max power)
http://www.cpu-world.com/CPUs/K10/AM...S6DGNWOF).html
Power at the wall: ~258W

Power at the wall for both (I just grabbed the first result, thus the ~):
http://www.spec.org/power_ssj2008/re...r_ssj2008.html

So, what you really see, is at the wall, you have to plan for about the same amount of power for both. So you can't compare TDP only (95W to 115W).

Nor, can you compare our ACP to their TDP because again there is a 20W delta.

What you really need to look at is the power at the wall, and you can see that they are both about the same.

That is a pure power discussion. You could layer in performance to try to argue it one way or the other, but you have to be really careful in doing that as both products are going to rev in 2 weeks. Unless you know where both are going to end up you might not want to put your stake in the ground just yet. (also don't forget that this is a $989 part vs. a $1386 part)

Can you please keep talking about your own product and avoid distributing false information about your competitor's product ... you are mixing Max electrical and max thermal numbers to confuse people, this is not a gentleman like attitude. I don't think you have a grip on Intel methodologies:
I propose that you talk about what you understand and know, and ONLY this.

Your little FUD on the power stuff should be kept on your own processors, you obviously don't understand the difference between electrical max power and Thermal max power, and I know why, your processors don't have a PCU, (power control unit), and you are trying to confuse people on it. Reality is that Nehalem and Westmere are able to make the difference between Electrical and thermal power, and we exploit this to provide maximum performance for a given TDP. Again, stay away from what you don t understand.

So, please, keep marketing your platform, but only speak about what you know and don't spit in the soup of your neighboor (with false claims and banana to apple comparaisons) ... thanks!

Francois

[ajaidev]

*Dangerous tone*

XS is the place where fight of the marketing people takes place....



JF vs Who, place ur bets gentlemen fight of the year rite here

[Katanai]

Can you please keep talking about your own product and avoid distributing false information about your competitor's product ... you are mixing Max electrical and max thermal numbers to confuse people, this is not a gentleman like attitude. I don't think you have a grip on Intel methodologies:
I propose that you talk about what you understand and know, and ONLY this.

Your little FUD on the power stuff should be kept on your own processors, you obviously don't understand the difference between electrical max power and Thermal max power, and I know why, your processors don't have a PCU, (power control unit), and you are trying to confuse people on it. Reality is that Nehalem and Westmere are able to make the difference between Electrical and thermal power, and we exploit this to provide maximum performance for a given TDP. Again, stay away from what you don t understand.

So, please, keep marketing your platform, but only speak about what you know and don't spit in the soup of your neighboor (with false claims and banana to apple comparaisons) ... thanks!

Francois

This, this is a flamers wet dream. It has certainly brought a tear in the corner of my troll eye.


Now on a more serious note: why has everything here degenerated this much and every thread has to be only about AMD??? For example the reply I got to my previous post was a flamebait ending in AMD something...

[Drwho?]

*Dangerous tone*

XS is the place where fight of the marketing people takes place....

JF vs Who, place ur bets gentlemen fight of the year rite here

I am not fighting with him, we all need this forum to stay clean, no false claim, no "below the belt", and his posting are full of misleading and miss understood comparaisons ... , and I see him doing this in many forums.
Take the high road!

I was ask to refrain to post in AMD related threads, I think JF need to refrain to post on Intel Threads. I look at the title and if it is AMD related, I stay away, even in EXtreme News area. (because otherwise, it was creating too much flaming, Charles and I agreed on this)
I did post in the past in AMD threads, and with the experience, it was a mistake.

[JF-AMD]

I broke the argument down to say that in the end power at the wall is the best way to measure this. That has always been my position.

The rest of it IS confusing. Power at the wall is pretty straightforward, wouldn't you agree that it is a better way to compare?