backtop


Print 32 comment(s) - last by 3kliksphilip.. on Jul 20 at 8:01 AM


Intel demonstrates quad-core Xeon DP "Penryn" at Computex 2007 (Source: DailyTech, Anh T. Huynh)
Intel "Penryn" Xeon DP to start out at $177 and top out at $1,172

Intel’s latest roadmap reveals pricing for its upcoming Penryn-based Xeon DP processors. Yesterday, DailyTech unveiled the Xeon DP models set for launch.  The costs of entry for Penryn-based Xeon DP processors start at $177 and top out at $1,172.

Intel has two Penryn-based quad-core Xeon DP processors under $300. The quad-core Xeon DP E5405 starts out at $209. The $209 price tag yields a quad-core Xeon DP processor clocked in the 2.0 GHz range, but below 2.33 GHz, with 12MB of L2 cache. Putting forth an extra $47 gets you the quad-core Xeon DP E5410 with its 2.33 GHz clock speed.

PenrynQuad-Core Xeon DP

Model
Core
Frequency
TDP
L2 Cache
Q4'07 Launch
Price

X54603.16 GHz 120W12MB
$1,172

E54503.00 GHz 80W12MB
$851
E5440
2.83 GHz 80W 12MB
$690
E54302.66 GHz 80W12MB
$455
E5420
2.50 GHz 80W12MB
$316
E54102.33 GHz 80W 12MB
$256
E5405
2.xx GHz
80W12MB
$209

As the clock speed increases, the price gaps between quad-core Xeon models get larger. Intel prices the 2.50 GHz quad-core Xeon DP E5420 at $316. Stepping up to the next speed-bin comes at a $139 cost, for the 2.66 GHz quad-core Xeon E5430. The quad-core Xeon DP E5440 and its 2.83 GHz clock speed comes in at $690 -- $235 more than the E5430 with a 170 MHz lower clock speed.

Breaking 3.0 GHz will cost $851 for the quad-core Xeon DP E5450 with its 3.0 GHz clock speed. The quad-core Xeon DP E5450 is as high as the Penryn-based Xeon lineup will go, while retaining an 80-watt thermal envelope. If 3.0 GHz is not enough, $1,172 will get you the big daddy of the Penryn-based lineup, the quad-core Xeon DP X5460. The quad-core Xeon DP X5460 has a 3.16 GHz clock speed, but the high clock speed comes at a cost. Intel rates the quad-core Xeon DP X5460 with a 120-watt TDP – 40-watts higher than the rest of the lineup.

Penryn Quad-Core Xeon DP (LV)

Model
Core
Frequency
TDP
L2 Cache
Q4'07 Launch
Price

L54302.66 GHz 50W12MB
$519
L54102.33 GHz50W12MB
$320

If the 80-watt thermal ratings are a bit high, Intel has low voltage quad-core Xeon DP processors. The quad-core Xeon DP L5430 and L5410 cost $519 and $320, respectively. The low voltage processors have 50-watt TDP ratings and command an extra $64 over its 80-watt counterparts.

Penryn Dual-Core Xeon DP
Model
Core
Frequency
TDP
L2 Cache
Q1'08 Launch
Price
E52603.33 GHz 65W6MB
$851
E52051.86 GHz 65W6MB
$177

Intel only has two Penryn-based dual-core Xeon DP processors this time around. The two dual-core Xeon DP processors occupy the very bottom and the very top of the Penryn-based Xeon DP lineup, in terms of clock speed. The clock-topping 3.33 GHz dual-core Xeon DP E5260 costs $851 while the 1.86 GHz E5205 costs $177.

Expect Intel to introduce quad-core Xeon DP processors next quarter in normal and low voltage flavors. The dual-core Xeon DP processors are not Penryn launch products, but will arrive in Q1’08.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

Rock the boat
By TimberJon on 7/17/2007 5:57:59 PM , Rating: 2
Concerning a quad-core.

Isnt the number of manufactured cores always based on an Even number? 1 isnt, but it doesnt have to match itself with anything else.

If you have 2 cores on this side of the boat, and 2 cores on the other, and one falls overboard, would'nt that mess up everything?

I would believe that would definately create some inconsistencies in processing.




RE: Rock the boat
By GaryJohnson on 7/17/2007 7:11:00 PM , Rating: 2
The XBox 360's CPU is a tri-core.


RE: Rock the boat
By 3kliksphilip on 7/17/07, Rating: 0
RE: Rock the boat
By mechwarrior1989 on 7/17/2007 8:12:11 PM , Rating: 2
It doesn't get bigger because they can change the architecture to optimize them and also decreases in manufacturing size. 130nm -> 90nm -> 65nm -> 45nm -> etc etc. That's almost 3 times smaller than the early AMD Athlon 64 chips were created on so you could easily pack many times more transistors in the same space.


RE: Rock the boat
By firewolfsm on 7/17/2007 11:03:19 PM , Rating: 2
You'd think it's around three times smaller but it's actually closer to 8. Each full node change halves transistor size, I'm not sure why though.

That's why dual worked with 90nm and now quad works with 65nm. If Penryn was a native quad chip, Intel could easily put two on one package and make an octo core.


RE: Rock the boat
By Garreye on 7/18/2007 12:07:18 AM , Rating: 5
I believe it's because when thinking about chip size you have to think in terms of area. So the effect of each process shrink is squared...For example if a chip on 90nm process is 9cm by 9cm, the area for this chip is 9^2=81cm^2, when we move to 65nm, the same logic can (ideally) fit in 6.5cm by 6.5cm, and the area of this is 6.5^2=42.25cm^2. So now we can have (81-42.25) 38.73cm^2, almost half, of the area on our chip to add more logic, or we can shrink the chip size down and save on cost.

Of course, the shrinking of the area will not work exactly as mentioned above as that situation is the ideal. Some logic does not scale down as well others and we may not have as much free space or be able to shrink quite as might be expected, but this is still the general idea behind it.


RE: Rock the boat
By oTAL on 7/18/2007 11:30:08 AM , Rating: 2
Yup. You are correct.
And the shrink from 65 to 45 is even better since 65^2/45^2=2.09 and 90^2/65^2=1.91 - quite a respectable difference.
There are, however, certain key points in the IC that are not manufactured at the lowest dimensions available. Still, they can usually be shrunk anyway since, there were key points on the previous process anyway. Ex: when manufacturing a certain IC in 65nm some key transistors are kept at 90nm, but once the process is moved to 45, those parts are reduced to 65nm.


RE: Rock the boat
By Garreye on 7/18/2007 12:15:28 AM , Rating: 2
quote:
I thought it would go 1 core, 2 core, 4 core, 8 core etc.

That does seem logical...everything for computers should come in powers of 2!

quote:
Can't make them faster? Add more of them!

Although the idea seems pretty simple, I think the implementation must be fairly complicated. Getting the cores to communicate and share data efficiently would be very complex. I don't even want to think about all the possible data and branch hazards involved with multi-core, multi-issue pipelined processors!


RE: Rock the boat
By 3kliksphilip on 7/20/2007 8:01:50 AM , Rating: 2
Why did I get voted down for that? Was it something to do with the word 'cheating' ?


RE: Rock the boat
By TomZ on 7/17/2007 10:12:20 PM , Rating: 2
I'm not an expert in the hardware in that particular domain, but I can't imagine that they couldn't produce a tri-core processor. The geometry of the chip might be nicer with 1, 2, or 4 cores, but I doubt that can't be overcome.

Certainly the software wouldn't care if there were 3 cores.


Where's Tigerton?
By Viditor on 7/17/2007 11:26:21 PM , Rating: 3
Except for the press releases last year, we have yet to see any Xeon MP (greater than 2 sockets) C2D based chips announced...
Does this meen that Intel is leaving the enterprise boxes to Barcelona alone?




RE: Where's Tigerton?
By defter on 7/18/2007 1:16:16 AM , Rating: 2
Tigerton will be released in September. As been mentioned by Dailytech and other sites.


RE: Where's Tigerton?
By Viditor on 7/18/2007 8:52:33 PM , Rating: 2
quote:
Tigerton will be released in September

This article is showing models numbers and specs for Xeons being released into next year...
Could you please show me the article with the specs and prices on Tigerton


RE: Where's Tigerton?
By Visual on 7/18/2007 6:36:25 AM , Rating: 2
Dual-socket quad-core setups already mean 8 cores, which many might consider enough. Even many MP setups from just two years ago couldn't match this (dual core Xeons came 2005). MP also is considerably more expensive, and the bottleneck for sharing a single south bridge becomes more noticeable.


RE: Where's Tigerton?
By darkpaw on 7/18/2007 2:15:33 PM , Rating: 2
I agree with this. Barcelona should be very good in the 4+ socket configurations, but the vast majority of business class systems use 2 sockets. 4+ sockets starts to become very cost prohibitive in hardware and even more so often in software licensing. Very, very few applications make use of that many processors as well. I think AMD will continue to do extremely well in the 4+ niche, but it definately isn't a huge market.


RE: Where's Tigerton?
By Viditor on 7/18/2007 9:12:24 PM , Rating: 3
quote:
the vast majority of business class systems use 2 sockets

Actually, that's not true (the vast majority use one)...but whether it is or not doesn't really matter, it certainly won't be true in the future.
As an example, have a look at this article http://www.anandtech.com/IT/showdoc.aspx?i=3039&p=...

In the conclusion it states:
"Studies indicate that the average utilization of a PC based server is approximately 7%...Take a look at VMWare, XenSource, and other virtualization solutions, if you really want to lower operating costs relating to power consumption"

As servers move more and more into the VM world, the advantages of fewer machines with more sockets increases (most especially at the Enterprise and Server Farm level).


RE: Where's Tigerton?
By Phynaz on 7/18/07, Rating: 0
difference
By tacoburrito on 7/17/2007 11:06:15 PM , Rating: 2
Can someone fill me in on the difference between the Xeon and the Core 2, other than the fact that the Xeon has a much bigger cache? I do understand that the Xeon is made for servers while the Core 2 is for desktop. But other than the difference in cache sizes, why can't someone use the Xeon for desktop uses?




RE: difference
By HotFoot on 7/17/2007 11:31:24 PM , Rating: 2
The Xeon and C2D chips actually have the same amount of cache. The differences come in quality control, in that the Xeons are supposed to pass more rigorous testing than C2Ds, as well as the Xeons for DP/MP work with FB-DIMM memory rather than regular DDR2. The latter reason is why they fit into a different socket, I believe. Also, you can't get, as far as I know, DP/MP C2D motherboards. The only way to have more than one of these CPUs on the same board is to go with the Xeon.

The Xeon SP (3xxx) chips are even more similar to the C2D.


RE: difference
By colonelclaw on 7/18/2007 8:08:27 AM , Rating: 2
seriously? is that it?

i was hoping the difference would be something meaningful and performance related. how youve described it makes it sounds like marketing


RE: difference
By tygrus on 7/18/2007 9:47:19 PM , Rating: 2
quote:
... Xeons for DP/MP work with FB-DIMM memory rather than regular DDR2. The latter reason is why they fit into a different socket, I believe.
Wrong.

Intel still uses FSB to IO/Memory Controller (ie. northbridge). The chipset can be changed to support different memory without changing socket. Intel does try to limit avaialable choices of socket & chipset combinations. Intel also ties the CPU socket with the voltage regulators (lower voltage, higher current for new CPU requires new socket with more supply/ground pins or modifies old socket to force you to update mainboard to insure support).
They are tested for use in dual-socket configurations or above in the case of MP (extra load on FSB, data coherancy, electrical stability). They typically have a larger saftey margin (24/7 use in confined space in servers) and additional cost of longer support (keep several in stock for 3yr old server to replace).

Some of these costs and responsibilities (keeping replacement parts in stock, testing, replacement) are handled by OEM's.

But when it comes down to it, Intel can charge whatever they like for various Server and Desktop SKU's (lower for better market share, higher if the market will take it).


Nice Prices
By akugami on 7/17/2007 5:09:23 PM , Rating: 2
I was hesitant to get a new CPU after the C2D E6600 which does pretty much anything I need but a $200'ish quad core CPU is looking really really enticing. Judging by past Core 2 Duo's and Quads, I'm sure we can get some decent overclocks outta that sucker. Hopefully in the 3.0 to 3.4 GHZ range.

I do a bit of video encoding and transcoding so a quad core would make sense to me. Assuming it overclocks well of course.

AMD...you need to get your stuff out now. Show us something. These great prices are only because you are pressuring Intel. Lose that and we're back to sky high prices.




RE: Nice Prices
By mdogs444 on 7/17/2007 5:32:24 PM , Rating: 1
I doubt the overclocking ability of the quad cores are going to be as much as the dual cores. They generate much more heat, and with 4 cores instead of two, you get much more inconsistancies. Some of the cores may be able to clock higher than others, thus limiting you to the lowest clock for stability.

With then coming stock at 2.33GHz, 3.0 might be a good target to shoot for. It would be a nice to have at speeds greater than 3.0GHz, but I wouldnt get your hopes up just yet, especially on the first revision of them.


RE: Nice Prices
By nemrod on 7/17/2007 7:44:11 PM , Rating: 2
This is 45nm... Should be better. At least tdp is lower at stock.

(but for OP, this is server cpu (LGA771)), we have to wait Q1 2008 for desktop.


RE: Nice Prices
By omnicronx on 7/18/2007 12:42:19 PM , Rating: 2
3.0ghz from 2.3 is probably a stretch too. As you said with 4 cores you have more chance of inconsistencies between the cores. I would not be surprised if they did not oc more than a few hundred mhz in most cases (maybe 400).. well and remain stable.. although these will probably oc better than than normal c2d quadcores because of the quality control.


Top to bottom
By jhtrico1850 on 7/17/2007 5:20:39 PM , Rating: 2
That's quite a lineup there.

http://www.xbitlabs.com/news/cpu/display/200705211...
54% Clovertown and 25% Harpertown with Harpertown covering all price points, how is that going to work out?




RE: Top to bottom
By darkpaw on 7/17/2007 5:29:47 PM , Rating: 2
I have to say I'm shocked. Usually Xeon = overpriced, but these prices are very nice.

Looks like Intel wants to crush AMDs profit margins in the business class systems as well. I really hope AMD has a good answer for that. These prices already mean AMD won't be able to charge any kind of a premium for Barcelona CPUs.


RE: Top to bottom
By Assimilator87 on 7/17/07, Rating: -1
RE: Top to bottom
By TomZ on 7/17/2007 10:15:33 PM , Rating: 1
Sounds like idle dreaming to me. You must be a real AMD fan to make a statement like that, considering there are no benchmarks even close to being out.


RE: Top to bottom
By defter on 7/18/2007 1:30:20 AM , Rating: 2
Yes, 2.5GHz quad core at $316 will be very tough for AMD. Especially considering that Penryn will have clock-to-clock performance improvements compared to Conroe.


By Kevlar on 7/18/2007 11:29:10 PM , Rating: 2
A CPU is just TRANSISTORS.....HOW MANY?

isn't "CORE" just marketing hype?

I want TRANSISTOR COUNTS!

Show me that 2 cores is half the transistors as 4 Cores....

SHOW ME THE MONEY!

How many TRANSISTORS in each of these "MODELS"

How many TRANSISTORS in each of the Core2 Duo's that are coming out JULY 22nd?

Sick of the marketing hype.




By Visual on 7/19/2007 4:06:09 AM , Rating: 2
WTF?


"Young lady, in this house we obey the laws of thermodynamics!" -- Homer Simpson

Related Articles
Intel Leaks 45nm Xeon Clock Frequencies
July 16, 2007, 12:40 PM













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki