backtop

Print E-mail del.icio.us 46 comment(s) - last by retrospooty.. on Jul 23 at 11:49 PM
Intel gears up for AMD's Barcelona launch with a slew of new quad-core CPUs

Hot on the heels of the second generation Intel Core 2 Duo launch, an internal memo began making the rounds with Intel's 45nm Xeon core frequencies. 

These 45nm Xeons, based on the Penryn generation of processors, also feature the 1333 MHz front-side bus found on the new Core 2 Duo stepping.  Each die features 6MB of L2 cache -- giving the two dice quad-core Harpertown Xeons a total of 12MB of L2 cache. 

45nm Quad-Core Xeons
Model
 Core
Frequency 		TDP
FSB L2 Cache
X54603.16 GHz 120W1333 MHz
12MB
E54503.00 GHz 80W 1333 MHz
12MB
E5440
2.83 GHz 80W 1333 MHz
12MB
E54302.66 GHz 80W 1333 MHz
12MB
E5420
2.50 GHz 80W 1333 MHz
12MB
E54102.33 GHz 80W 1333 MHz
12MB
E5405
??? 80W???
12MB
L54302.66 GHz 50W 1333 MHz
12MB
L5410
2.33 GHz 50W 1333 MHz
12MB

Unlike the desktop roadmap, dual-core Xeons take a more auxiliary role.  Intel only laid out plans for three dual-core SKUs at launch; one of which is a low-power CPU.  These dual-core 45nm processors, dubbed Wolfdale, will only consist of a small amount of shipped 45nm Xeon processors.

Intel also uses the Wolfdale codename to describe dual-core 45nm desktop offerings.

45nm Dual-Core Xeons
Model
 Core
Frequency 		TDP
FSB L2 Cache
E52603.33 GHz 65W1333 MHz
6MB
E52051.86 GHz 65W 1066 MHz
6MB
L5250
3.16 GHz 40W 1333 MHz
6MB

One Intel engineer, who asked to not be named, claimed this focus on quad-core is a typical reaction to the market in general.  "In the server space, there isn't much need for dual-core when we can go quad ... If your [applications] are threaded, there's no reason to use two cores when four are available."

Intel did not release pricing on the new Xeons, but comparable 65nm CPUs are expected to get as low as $266 over the next few weeks when Intel officially cuts prices on all its existing dual and quad-core processors.

The next-generation AMD Barcelona family is expected to directly compete with the Intel 45nm quad-core Xeons.  AMD is expected to launch Barcelona in the last week of August, with top-out frequencies of 2.0 GHz.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
Holy consumption!
By mdogs444 on 7/16/2007 1:29:16 PM , Rating: 1
The X5640 looks quite tasty, but 120W TDP? I know at 45nm it would reduce heat, but not sure how much at 120w!

No doubt I'd rather have the 5410 at 50w w/ 1333 fsb & 12MB cache. Should OC much higher than 3.16 GHz anyway.




RE: Holy consumption!
By Operandi on 7/16/2007 1:39:02 PM , Rating: 2
quote:
The X5640 looks quite tasty, but 120W TDP? I know at 45nm it would reduce heat, but not sure how much at 120w!


120 watts peak if you manage to load all four cores to 100%. Under lower loads the power consumption will be far more reasonable.


RE: Holy consumption!
By omnicronx on 7/16/2007 1:56:24 PM , Rating: 1
still the jump of 0.13ghz does not justify the 40w tdp difference. It could theoretically also draw up to 1/3 more power even not at idle. Although with intels stupid tdp system they probably measure the tdp by the average of the whole series of cpu's (excluding the X version) so the chances are the 3ghz version draws a higher tdp than it is rated.

also on another not.. i doubt quad cores will overclock as easily is dual core systems. 4 cores = more incosistancies between cores.


RE: Holy consumption!
By Thorburn on 7/16/07, Rating: 0
RE: Holy consumption!
By omnicronx on 7/16/2007 3:21:38 PM , Rating: 2
Heh i am in no way disputing the way tdp is rated or what it means. What i am saying is the way intel measures their tdp between different models is flawed.

and the point i was trying to get across is the 40wtdp for a 0.13ghz increase makes no sense, the meaning of tdp aside, the processor with 120tdp will probably draw more power than its 80tdp counterpart. just go look at wiki :
quote:
The TDP is typically set not to be the most power the chip could ever draw , but rather the maximum power that it would draw when running real applications


RE: Holy consumption!
By Thorburn on 7/16/2007 3:32:45 PM , Rating: 2
It makes sense in the sense that VR-Zone's numbers aren't accurate.

Of course people are right in saying that with TDP being set at certain levels (eg. 50w, 80w, 120w) that even if its only 1 or 2w above it goes up to the next grade, the actual rating isn't given (and this allows say a core which would be marginal for an 85w TDP to still fit comfortably within the 120w window).

What is wrong is saying that a chip will draw above its TDP, and as for that Wikipedia quote, I'd dispute it.


RE: Holy consumption!
By omnicronx on 7/16/2007 4:14:19 PM , Rating: 2
and thats why i did not want to get into tdp ;) every source I have seen has something different to say.

but as far as i understand it, a chip can draw above its tdp as they never give out the maximum draw (whether or be over a millisecond or a minute).

i think what you are basing your statement on is sustained wattage not burst. so in effect a 80wtdp chip will at load over a period of time handle up to 80wtdp safely but could burst higher.


RE: Holy consumption!
By PlasmaBomb on 7/16/2007 5:26:56 PM , Rating: 4
Intel stat that there TDP is "Not 100% tested. Specified by design/characterization." It is what is expected to be consumed at 100% load under specified conditions.


RE: Holy consumption!
By Shintai on 7/16/2007 5:32:26 PM , Rating: 1
Ye right, you cant find a single P-M, Core or Core 2 that is even near their TDP under max load. The only ones you can is the P4s.

http://www.xbitlabs.com/articles/cpu/display/core2...

And this is even with the VRM aswell.


RE: Holy consumption!
By GlassHouse69 on 7/17/07, Rating: 0
RE: Holy consumption!
By Hawkido on 7/18/2007 1:15:40 PM , Rating: 2
Doesn't Intel advertise Average Load TDP Ratings, whereas, AMD always uses Full Load TDP Ratings. That's what I have been told. Can anyone confirm this? If this is true the the TDP could be Higher under Max Load.


RE: Holy consumption!
By Roy2001 on 7/16/2007 1:52:18 PM , Rating: 2
Right now 2.66Ghz quad-core TDP is 125W. 45nm 3.0Ghz quad-core TDP is just 80W. Hope you can realize the fact.


RE: Holy consumption!
By mdogs444 on 7/16/2007 1:59:51 PM , Rating: 2
Are you sure the 3.0GHz is 45nm? I think thats 65nm.


RE: Holy consumption!
By kostas213 on 7/16/2007 2:12:15 PM , Rating: 2
It is surely 45 nm. The TDP of this processor is 80 W and it also has a 12 MB L2 cache. In contrary the 65 nm Xeon 3.0 GHz has 120 W TDP (not 100% sure about the TDP) and 8 MB L2 cache.


RE: Holy consumption!
By kostas213 on 7/16/2007 2:10:36 PM , Rating: 2
quote:
No doubt I'd rather have the 5410 at 50w w/ 1333 fsb & 12MB cache. Should OC much higher than 3.16 GHz anyway.

I thought that a 3.33 GHz Penryn would show up. As far as I remember, Intel has demoed a dual-socket configuration with Penryn running at 3.33 GHz.


RE: Holy consumption!
By coldpower27 on 7/16/2007 2:27:55 PM , Rating: 2
Yeah I think I remember that too, though I guess Intel reevaluated their need to bring such a power SKU to market after seeing what AMD has to offer, or just keeping it for a later refresh of the line.


RE: Holy consumption!
By retrospooty on 7/23/2007 11:49:39 PM , Rating: 2
"I thought that a 3.33 GHz Penryn would show up. As far as I remember, Intel has demoed a dual-socket configuration with Penryn running at 3.33 GHz."

This is just the Xeon roadmap. Penryn includes Xeon Server chips and and Core2(or3)quad desktop chips. I am sure there will be a 3.33ghz version, and likely a 3.66ghz dual core as well, with 4ghz coming soon after.


RE: Holy consumption!
By Samus on 7/16/07, Rating: -1
RE: Holy consumption!
By Shintai on 7/16/2007 3:18:55 PM , Rating: 2
Cache dont eat much power. On the old 3.73Ghz P4 with 16MB l3. The cache was 0.75W per MB.


RE: Holy consumption!
By Thorburn on 7/16/2007 3:19:10 PM , Rating: 2
Cache isn't a big power consumer. Infact its one of the lowest power regions of a CPU.

It's also very easy to shutdown cache to save power and there is relatively low leakage current.


RE: Holy consumption!
By mdogs444 on 7/16/2007 3:26:14 PM , Rating: 2
What you seem to forget is that the AMD & Intel CPU's are a little bit different on how much they perform per clock cycle. The Cache is also affected in much the same way.


RE: Holy consumption!
By maroon1 on 7/16/2007 6:23:19 PM , Rating: 1
X5460 has unlocked multipliers because it is an extreme edition

And extreme edition usually have higher TDP than non-extreme edition


RE: Holy consumption!
By coldpower27 on 7/21/2007 12:18:38 PM , Rating: 2
These are server parts, they don't have unlocked multipliers they aren't designed to be overclocked.


.5 Multipliers
By Zandros on 7/16/2007 1:36:24 PM , Rating: 2
Seems like 1/2 multipliers are back in fashion, now both with AMD and Intel.

Quite a massive jump in TDP over just 160 MHz in core frequency. (Yes, even with Intel's method of measuring TDP)




RE: .5 Multipliers
By TomZ on 7/16/2007 1:52:30 PM , Rating: 2
You have to measure actual power dissipation if you want to evaluate that aspect of performance. TDP ratings don't tell you much, as those figures illustrate.


RE: .5 Multipliers
By coldpower27 on 7/16/2007 2:29:29 PM , Rating: 4
Intel only uses certain TDP levels on the their Quad Core Xeon's, 50W 80W 120W, so even if a processor just barely crosses the 80W barrier and say is 85W it will be slapped with a 120W TDP.


RE: .5 Multipliers
By Andrwken on 7/16/2007 2:34:51 PM , Rating: 3
It's probably just powerhungry enough to get over the 80w tdp spec. So they upped the tdp to the 120w range. My guess is that it probably is not going to push the 120w threshold with that little of a clock increase.


Typo
By Visk on 7/17/2007 3:50:48 AM , Rating: 1
quote:
features 6MB of L2 cache -- giving the two dice quad-core


should be

quote:
features 6MB of L2 cache -- giving the two die quad-core




RE: Typo
By decapitator666 on 7/17/2007 8:45:58 AM , Rating: 2
From wikipedia.
http://en.wikipedia.org/wiki/Die_%28integrated_cir...

Die (integrated circuit)
From Wikipedia, the free encyclopedia
Jump to: navigation, search
A medium-scale integrated circuit die, with bond wires attached.
A medium-scale integrated circuit die, with bond wires attached.
A VLSI integrated-circuit die
A VLSI integrated-circuit die

A die in the context of integrated circuits is a small block of semiconducting material, on which a given functional circuit is fabricated. Typically, integrated circuits are produced in large batches on a single wafer of electronic-grade silicon (EGS) through processes such as lithography. The wafer is cut into many pieces, each containing one copy of the ciruit. Each of these pieces is called a die.

There are three commonly used plural forms: dice, dies, and die.

So if you trust wikipedia, dice is OK


RE: Typo
By phusg on 7/18/2007 7:08:32 AM , Rating: 2
Typical wikipedia to compromise and state all 3 forms are ok! Personally I agree with Visk, I found the sentence read wrong too. I would go with:

The quad-core has 2 dice. It is double diced. Each die features 6MB of L2 cache; giving the two diced quad-core Harpertown Xeons a total of 12MB of L2 cache.

Compare with leg. You don't say two legs person. It's a two legged person.


RE: Typo
By Hawkido on 7/18/2007 1:24:36 PM , Rating: 2
Those little cubes you toss down the craps table...

1 = Die
2 or more = dice

Dice is nice.

Live with it.


No more desktop chips?
By podknocker on 7/16/2007 3:02:54 PM , Rating: 2
I hope Intel decide to roll out some Socket 775 processors. These new 45nm parts will be Socket 771 and need expensive FB-DIMM memory and the use of server boards.

I considered a new system based on an X3210 or X3220 CPU on a new P35E motherboard but these are 1066FSB

I'd love to see 12MB Socket 775 Xeons with 1333MHz FSBs running at these speeds. A low voltage part with this spec would be a bonus.

Has anyone heard a rumour that suggests Intel will release such CPUs?




RE: No more desktop chips?
By PlasmaBomb on 7/16/2007 5:29:03 PM , Rating: 2
What's wrong with getting a X32** at 1066 FSB and upping it yourself to 1333 FSB, its been done many times.


RE: No more desktop chips?
By porkpie on 7/16/2007 9:45:59 PM , Rating: 2
So what mobo/chipset do I need to support the 1333mhz fsb?


X5640? Not X5460?
By kostas213 on 7/16/2007 2:08:20 PM , Rating: 3
Are you sure that the 3.16 GHz Penryn will be named X5640? If this model has to follow the whole Penryn model number scheme, it should be named X5460.




By KristopherKubicki (blog) on 7/16/2007 2:19:51 PM , Rating: 2
Yes -- thanks for the correction.


Comparable?
By darkpaw on 7/16/2007 1:37:21 PM , Rating: 2
There definately won't be anything comparable in price between the soon to be $266 quad and even what a low end Xeon will sell for.




RE: Comparable?
By coldpower27 on 7/16/2007 2:32:05 PM , Rating: 2
It will have to depend for Quad Core DP Xeons the lowest price per is going to be the E5310 price drop to $209, for 1.6GHZ Quad Core.


clock speeds
By FujiT on 7/16/2007 2:38:24 PM , Rating: 2
The Xeon clock speed does look a bit low when compared to the Yorksfield core which is expected to clock at least 3.3 GHz.

But the 45 watt reduction for the 3 GHz quad core is noice.




RE: clock speeds
By coldpower27 on 7/16/2007 5:06:13 PM , Rating: 2
Xeon values stability first over frequency, usually Xeons are clocked lower then what the desktop processors reach or reach it later, the Woodcrest was the exception rather then the rule.

It's only a margina clock decrease though 3.16 vs 3.33.


when is the 1600 fsb penryn coming?
By qnetjoe on 7/16/2007 3:52:37 PM , Rating: 2
it is great that the 45nm 1333 are leaked b ut does anyone know when the 1600mhz fsb one are coming or the release date of the seaburg chipset?




By FujiT on 7/16/2007 5:51:00 PM , Rating: 2
I was talking to Faud from Fudzilla and he said that Initially, the 45nm Xeons will launch with the 1.33 GHz FSB and then get a boost next year.

However, there's talk that Nehalem will be brought into Q2 2008.


45nm decent improvement on TDP
By PlasmaBomb on 7/16/2007 5:19:41 PM , Rating: 3
If you compare them to current 65nm quad core parts the improvement is pretty good. At the moment the 5365 operates at 3.0GHz and chews down on 150W TDP. The new 5450 manages 3.0GHz at 80W which is great.

Intel's thermals are what the chip can typically be expected to use at 100% load under normal operating circumstances, therefore if things go bad they could use more than the stated TDP.




TDP means different things to different companies.
By eickst on 7/16/2007 5:31:17 PM , Rating: 1
Intel's TDP ratings are about as accurate as our intelligence reports that said Saddam had WMDs.

"Thermal Design Power (TDP) should be used for processor thermal solution design targets. The TDP is not the maximum power that the processor can dissipate.?"

?"The numbers in this column reflect Intel?s recommended design point and are not indicative of the maximum power the processor can dissipate under worst case conditions.?"

Taken from processor datasheets.




By Khato on 7/16/2007 6:18:14 PM , Rating: 2
What's wrong with a little bit of legal disclaimer in the datasheet? Worst case conditions aren't nominal operating conditions under full load. No, that's what the TDP represents. Worst case conditions where the processor will dissipate more than the TDP is when you have a 'slow silicon' processor with motherboard VRM that's at the top end of the tolerated voltage margin - extra voltage on a 'slow' part means more heat.

Heh, it's easy to make a processor dissipate more heat than the TDP dictates it should...


TDP
By pauldovi on 7/16/2007 2:29:19 PM , Rating: 2
The X5460 has a significant just in TDP because it is in a different range compared to the others. The X5460 may just be a few watts higher than the others, but it managed to be higher enough to jump to the "range" or TDP. Just look at the TDP of the X6800 compared to the TDP of the E6300-E6700.




"I f***ing cannot play Halo 2 multiplayer. I cannot do it." -- Bungie Technical Lead Chris Butcher

DailyTech Poll
Which web browser do you use on your primary personal machine? 






44 Comments









botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki