Socket 1567 Xeon Nehalem-EX

[Fuji]

meaning = S2011 is no MP socket and will be a different one than S2011. So S1567 will stay for some time, even when S2011 (or whatever is the socket for sandybridge) is around. When ever intel decides to release a replacement for nehalem EX i guess well see a new socket, but i don't think that will happen before 2013-14. It took nehalem ex 2 years and several months to be available after nehalem was released. Sof if it follows the same pattern and sandybridge comes beginning 2011 the eraliest date we can expect MP sandybridge will be 2013.

It's hard to believe that Nehalem has been out for a year and like four months.

I think the Nehalem-EX platform is coming out later than planned. I remember reading about a delay although i can't remember when.


I say this b/c generally, the XeonMP line is delayed by a year.

http://en.wikipedia.org/wiki/Xeon

There are exceptions, but generally, there's a 6 month-1 year delay.

If you look at the XeonMP release date, you'll see that Intel has been somewhat silent about that platform. You have a year or two between releases but between 2005-2008, we had 7 XeonMP CPU releases.

Nehalem EP was released in march of 2009 so it would make sense that given the delay that it would come out around this time of year.

Having said that I think we'd see a Nehalem-EX successor sometime next year ish based on the Westmere design. And then a year or so later, transition to Sandybridge.

[SocketMan]

TDP,ACP,PCP,KGB...anyone heard of PPD (Point per day) ?
That and the electric bill that comes in is what matters to me,in other words how much an X amount of PPD ( running WCG or FAH) will cost me daily,assuming 24/7/365 full load.
It all comes down to PPP -Price,Performance and Power consumption (based on the above scenario).

Who works for who, and who sleeps with who I really don't care about

[savantu]

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

Nice job of comparing some "measured" numbers ( who knows how they were actually measured) with official figures.
What's the max power and highest power for the Opteron then ?

Oh wait, should it be 115w ? Then how is it possible for both systems to end up with the same usage at the wall ? And more extraordinarly, the Opteron loses by 4w.
Assuming more or less similarly configured systems we have :
-155W Xeon + rest of system = 254W
-"115W" Opteron + rest of system = 258W.

This makes me believe the 115W is pure BS. AMD's official definition is identical to the one Intel uses. So either the Xeon doesn't use 155W continous, which I find ridiculous for a part with a cooling system designed for 95W as per Intel's datasheet ( haven't heard of Nehalem datacenters going up in smoke or throttling down ) or that figure is grossly innacurate.
Based on what sites like Anantech, Hexus,Techreport have measured, I believe it is the later.

[duploxxx]

well, you are taking about the "powercaculation" that are by OEM for price negociation ... lol ... you are more naive than I though if you believe those ... They are design to get your price down my friend ... wake up

I have a story for you ...

It is a man who go to a watch store, he ask to the clock master, do you have a clock that is red with green dot in the back ground ... then, he master says "hell no!"
Then , the customer says: "So, it is 10% off on the others" ...

lol man ... you are so funny some time.

now who's naive here , I don't give a what your company has written on a so called L,E,W,X serie which is rated at a certain TDP. Those TDP/ACP figures changed into a marketing sales argument with retared ratings, for me that is something a hardware vendor should take care about to calculate system air flow, electronics etc. Perhaps your company could add additional star rating for your server cpu's just like in your stupid consumer brand logo's

I don't care what OEM company A rates as total system power or B, wouldn't even think of a fact that would drive me to distribute other vendors just for that, have any idea what a world wide service organisation means? All we need to know what that total system is expected to consume on idle and max load for BTU calculations and power calculation. What kind of cpu we will take will not be depending on TDP or decided by again marketing hyped spec calculations although those are good indicators, that is where those spec and manipulated compilers end.

Again try to be neutral, For a certain OEM with the same server platform type but different cpu manufacturer they do calculations on power consumption then you can take conclusions what they are really consuming.
Those vendors should be neutral on that! accept for the fact that there are always people who give certain rebates to buy and sell crap remember?

Weather or not a certain system is more powerfull, gets more don in a certain time, less total cost or wathever is another discussion, you dragged in the basic powerconsumption linked to TDP.
Don't forget it is just about 1 year that your company can drag about lower power consumption on some parts, for another 6 years you could only eat paper, now lets see again what will happen in a few days.

[Movieman]

now who's naive here , I don't give a what your company has written on a so called L,E,W,X serie which is rated at a certain TDP. Those TDP/ACP figures changed into a marketing sales argument with retared ratings, for me that is something a hardware vendor should take care about to calculate system air flow, electronics etc. Perhaps your company could add additional star rating for your server cpu's just like in your stupid consumer brand logo's

I don't care what OEM company A rates as total system power or B, wouldn't even think of a fact that would drive me to distribute other vendors just for that, have any idea what a world wide service organisation means? All we need to know what that total system is expected to consume on idle and max load for BTU calculations and power calculation. What kind of cpu we will take will not be depending on TDP or decided by again marketing hyped spec calculations although those are good indicators, that is where those spec and manipulated compilers end.

Again try to be neutral, For a certain OEM with the same server platform type but different cpu manufacturer they do calculations on power consumption then you can take conclusions what they are really consuming.
Those vendors should be neutral on that! accept for the fact that there are always people who give certain rebates to buy and sell crap remember?

Weather or not a certain system is more powerfull, gets more don in a certain time, less total cost or wathever is another discussion, you dragged in the basic powerconsumption linked to TDP.

Now since there's no PHD behind my name but instead almost 45 years of working out in the real world here's a common sense way to end all these arguments over whose system draws what,when and under what load:
Buy a damned $20.00 Kill-A-Watt like I did, plug into wall, turn on machine, record the numbers on the little LED at idle, run it up to 50% load, record again, run to 100% load, record again. Shut down machine, test next machine, compare numbers, toss in the variable as to machines capabilities, and bingo, decision made..
Cut's right thru everyone's claims and settles all..
Numbers don't lie, marketing depts on all companies at the very least bend the hell out of the facts and quite often lie thru their teeth.

[kl0012]

More info about Nehalem-EX architecture from ISSCC 2010:
http://translate.google.com/translat...n&hl=&ie=UTF-8
The new thing here is that Nehalem-EX has ring bus L3 architecture. It is interesting how it will perform since upcoming Sandy Bridge also has a ring bus.

[JF-AMD]

Buy a damned $20.00 Kill-A-Watt like I did, plug into wall, turn on machine, record the numbers on the little LED at idle, run it up to 50% load, record again, run to 100% load, record again.

Numbers don't lie, marketing depts on all companies at the very least bend the hell out of the facts and quite often lie thru their teeth.

That is why I am a fan of "at the wall" power. I have a bunch of kill-a-watt meters at home and we have lots of them at the office.

[kl0012]

That is why I am a fan of "at the wall" power. I have a bunch of kill-a-watt meters at home and we have lots of them at the office.

Wow. You must be a fan of Intel's cpu then

[flippin_waffles]

Now since there's no PHD behind my name but instead almost 45 years of working out in the real world here's a common sense way to end all these arguments over whose system draws what,when and under what load:
Buy a damned $20.00 Kill-A-Watt like I did, plug into wall, turn on machine, record the numbers on the little LED at idle, run it up to 50% load, record again, run to 100% load, record again. Shut down machine, test next machine, compare numbers, toss in the variable as to machines capabilities, and bingo, decision made..
Cut's right thru everyone's claims and settles all..
Numbers don't lie, marketing depts on all companies at the very least bend the hell out of the facts and quite often lie thru their teeth.

Pretty sure this is what John has been advocating all along, but good ol' Francois didn't want to play.

[v0dka]

Wow. You must be a fan of Intel's cpu then
http://www.techreport.com/r.x/core-i...-sysprices.gif

So the power consumption and the price are higher, yet the efficiency is better. Of course that is in a 100% multitask scenario with HT enabled. Not the kind of comparison that I call useful for consumer cpu's.

[kl0012]

So the power consumption and the price are higher, yet the efficiency is better. Of course that is in a 100% multitask scenario with HT enabled. Not the kind of comparison that I call useful for consumer cpu's.

Well, you can choose wathever you want from the following scenarios:
idle, load, task energy, overtime energy. They have it all:
http://techreport.com/articles.x/18581/4
I would say that i5-750 is the best thing you can find if you need performance and relative low price & power consumtion. Otherwise i3-530 is doing its job well while using impressive low power.

[Katanai]

Well, you can choose wathever you want from the following scenarios:
idle, load, task energy, overtime energy. They have it all:
http://techreport.com/articles.x/18581/4
I would say that i5-750 is the best thing you can find if you need performance and relative low price & power consumtion. Otherwise i3-530 is doing its job well while using impressive low power.

But wait, the AMD chips draw power smoother than Intel chips.

Anyway this is a none-issue right here. AMD has nothing in this market segment so all comparisons are pointless and only serve to derail this thread even more.

[JF-AMD]

AMD has nothing in the 4P market?

[BeepBeep2]

But wait, the AMD chips draw power smoother than Intel chips.

Anyway this is a none-issue right here. AMD has nothing in this market segment so all comparisons are pointless and only serve to derail this thread even more.

lolwut?

http://products.amd.com/en-us/Optero...il.aspx?id=567

And

http://www.xtremesystems.org/forums/...d.php?t=234105
http://en.wikipedia.org/wiki/Socket_G34

[Movieman]

But wait, the AMD chips draw power smoother than Intel chips.

Anyway this is a none-issue right here. AMD has nothing in this market segment so all comparisons are pointless and only serve to derail this thread even more.

Let's keep this friendly ok?
It is very fair to say that AMD is very strong in the 4 core market right now and should continue to be.
Thier 6 core chip more than matches the current Intel Dunnington platform.
Then yes, Intel will be bringing out the 4P Beckton and AMD will be countering with the 4P 48 core MC system.
It's give and take and yes, one can spin this any way one wants but reality is they both have excellent systems coming to market.

[PytonOrm]

hmmm 8 cores & 16 Threads per cpu thats 64 cores total and 128 Threads runs out on 4 cpu´s =)

[Florinmocanu]

no, that's 32 cores and 64 threads out of 4 CPUs

[Movieman]

no, that's 32 cores and 64 threads out of 4 CPUs

Correct, but I've seen a 8 socket Beckton board with 64 cores and 128 threads.. Impressive but I'd hate to write the check for it..
Just the cpu's is well over $25,000.00

[savantu]

Correct, but I've seen a 8 socket Beckton board with 64 cores and 128 threads.. Impressive but I'd hate to write the check for it..
Just the cpu's is well over $25,000.00

The SW cost makes the HW look like pennies at that level.

[ghost_recon88]

Correct, but I've seen a 8 socket Beckton board with 64 cores and 128 threads.. Impressive but I'd hate to write the check for it..
Just the cpu's is well over $25,000.00

$3k for a server level octo-core chip with HT? Really not that bad to be honest.

[overclocking101]

dear god there goes HWBOT!

[Otis11]

...

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

...

Um... So what about the company that has 10,000 of these in one room (and has plans to expand?)

Sometimes floorspace matters more than power usage.

Correct, but I've seen a 8 socket Beckton board with 64 cores and 128 threads.. Impressive but I'd hate to write the check for it..
Just the cpu's is well over $25,000.00

Now do that math... $25K x 10,000 systems...

Now figure each one does 560K ppd for WCG...