backtop


Print 64 comment(s) - last by hstewarth.. on Jun 29 at 12:56 AM

SMP version of "Conroe" takes flight, but availability seems scarce

This Monday, June 26, 2006, Intel will launch its new Woodcrest processor, a server processor based on Intel's Core architecture. The new processor family, dubbed Xeon DP, will use Socket LGA-771 and include support for multiple-CPUs -- which enables a system to support more than one Woodcrest processor, as opposed to being limited to just a single Core 2 processor per system. Despite the significant difference, every other aspect of Woodcrest is virtually identical to Core 2 Duo for the desktop.

Intel's new Woodcrest processor comes at us with the intention to compliment and ultimately replace Dempsey, Intel's latest dual-core Xeon processor based on the NetBurst microarchitecture. While both Dempsey and Woodcrest are based on 65nm processes, Woodcrest will focus a great deal on power consumption efficiency.  Both Dempsey and Woodcrest have a TDP envelope of 130W, but the performance is much higher on Woodcrest giving better performance-per-Watt. Being a NetBurst component, Dempsey scaled high in clockspeed, reaching up to 3.73GHz on a 1066MHz FSB. It also employed a 2x2MB L2 cache structure versus Woodcrest's shared 4MB L2 cache. Another detail to note is that Dempsey is capable of processing up to four threads simultaneously thanks to Hyper-Threading, versus Woodcrest's maximum of two -- Hyper-Threading is not enabled on the first Core 2 Duo nor Woodcrest Xeons.

During some demonstrations of Woodcrest
, Intel stated that when compared to AMD's Opteron processor, Woodcrest can be up to 33% more efficient in power consumption. Actual tests however indicated that Woodcrest is about 10% to 15% lower in some situations.  As for evolution, Intel says that the new Woodcrest is two to three times better in power efficiency over the previous Dempsey core.  AMD also recently published its opinion on server power consumption.

Unfortunately, here's the kicker.  Numerous channel vendors have contacted DailyTech claiming that availability of Woodcrest will not be for another two weeks.  Synnex and ASI will be the only US distributors with any quantity, and then general availability begins WW31. One vendor sent DailyTech a memo claiming "[the] second week of August is when we start to get box [processors] in volume." Several vendors will announce system builds with the processors immediately, but there will be no channel availability.  Motherboards are already widely available as every Socket 771 motherboard that supports Dempsey also supports Woodcrest.  Intel will ship Woodcrest Xeon DP in the following configurations:

Intel Xeon Processor
Processor
Brand
Processor
No.

Core /
FSB
L2 Cache
Price @
Launch
Xeon Processor DP
(Woodcrest Bin-0)
5160   3.0GHz /
1333MHz
4MB
$850
Xeon Processor DP
(Woodcrest Bin-1)
5150   2.66GHz /
1333MHz
4MB
$690
Xeon Processor DP
(Woodcrest Bin-2)
5140   2.33GHz /
1333MHz
4MB
$455
Xeon Processor DP
(Woodcrest Bin-3)
5130   2.0GHz /
1333MHz
4MB
$320
Xeon Processor DP
(Woodcrest Bin-4)
5120   1.86GHz /
1066MHz
4MB
$260
Xeon Processor DP
(Woodcrest Bin-5)
5110   1.60GHz /
1066MHz
4MB
$210

All Woodcrest processors will sport 4MB of L2 cache and are manufactured at 65nm. Prices will start at $210 and increase up to $850 in batches of 1000. With Woodcrest it's evident that most if not all of the world's top server companies will be shipping systems with the new processor. One of the most anticipated uses of Woodcrest will be Apple, which is expected to be releasing Woodcrest based systems later this year.

Woodcrest
will be accompanied by DDR2 memory, running at either DDR2-533 or DDR2-667. On Intel's Bensley platform, Xeon DP systems will be limited in the graphics department -- systems will only support a single-lane PCI Express setup. However, memory performance should see a nice boost thanks to the use of fully-buffered DIMMs (FB-DIMMs).

Looking further down the road, Intel's Clovertown will feature quad-cores -- two Woodcrest dice stamped onto a single package. This gives Clovertown systems the ability to scale up to eight CPU cores. Intel also says that Clovertown will deliver power consumption levels on par with Woodcrest. Later this year, Intel will also release Xeon MP Tulsa, the final processor based on the NetBurst architecture. Tulsa may be the last NetBurst processor from Intel, but it will be the company's first x86 processor to support shared L3 cache.  Intel's Itanium 2 Montecito processor will be the company's first shared-L3 processor.

Looking through into 2007, Intel is expected to introduce Tigerton, a new Xeon MP processor set to replace Tulsa. Tigerton is expected to contain at least four cores and have support for SMP configurations of four or more processors per system. With Tigerton, Intel is also expected to include a technology it currently calls "dedicated high-speed interconnect." The new technology gives each processor a direct pathway to the chipset. This will prove to be much faster than today's front-side bus technology. Actual launch dates for Tigerton are still unknown.

AMD also has a big server release around the corner on August 1, 2006. The new Opteron "Revision F" -- based on LGA 1207-pin Socket F -- will mark a significant evolution in multi-core CPUs for AMD.

Update 06/26/2006: The Intel press relese for Woodcrest was released today.


Comments     Threshold


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

Supply shortgage probbaly FUD
By caboosemoose on 6/23/2006 8:16:56 PM , Rating: 3
Personally, I reckon the stories of tight supply on NGMA chips is probably FUD. For starters, the world is absolutely drowning in leaked engineering samples and the press are currently being buried in samples, too. When chips are in short supply, it's almost always reflected in these areas, too.

Thing is, the sort of channel companies that talk to Daily Tech are at the bottom of the food chain. Intel could frankly give a shit if some crappy middle man online retailer has boxed processors or not. The big money is with the system integrators - especially for servers and especially for this chip - I mea it's not as is Woodcrest is going to be a drop upgrade for any company's warehouse full of servers is it? So, if Dell launches Woodcrest servers with short deliveries times, I for one will judge that Woodcrest supply to be just fine - especially for a brnad new architecture.




RE: Supply shortgage probbaly FUD
By KristopherKubicki (blog) on 6/23/2006 9:05:06 PM , Rating: 2
quote:
Thing is, the sort of channel companies that talk to Daily Tech are at the bottom of the food chain

What makes you say that?


RE: Supply shortgage probbaly FUD
By caboosemoose on 6/24/2006 9:38:28 AM , Rating: 2
It's not a slight on Dailytech in particular. But on the whole, the big, serious players like Dell, HP or whoever don't talk to news websites about such things. There's nothing in it for them so there's no reason to take any risks by doing so. Broadly speaking, the people who really know about the state of NGMA chip suply won't be talking about it.


RE: Supply shortgage probbaly FUD
By Shadowmage on 6/24/2006 9:49:24 AM , Rating: 2
Sorry, but you're completely wrong. Although the companies themselves won't officially speak to review websites as regards to unreleased products, nothing stops the engineers or other people from leaking.

I should know, from first hand experience ;)


By caboosemoose on 6/24/2006 12:28:40 PM , Rating: 3
Yeah - I know from firsthand experience too - and the people at the top with the real info largely don't leak. The wage slaves nearer the bottom who have but a partial view of the whole story occasionally do, in the right circumstances. But the big players at the big companies don't get involved most of the time. If you think you're privvy to all kinds of super secret yield info, then good for you. My opinion is that the FUD surrounding NGMA yields is just that - FUD.

Another reason why I say that is that whenever I am, say, under NDA I am then in a position to judge the veracity of the ruomour mill. And for stories like this, the hit rate is pretty low on the whole.


RE: Supply shortgage probbaly FUD
By hstewarth on 6/24/2006 12:47:20 AM , Rating: 2
Yep and probably started from the # tech fud site below...

http://sharikou.blogspot.com/2006/05/intel-woodcre...

Woodcrest demand is probably going the strongest server chip yet. Over at 2cpu.com, there is one desirng over clocking and such with them. So some are not just thinking of highend workstations or server for this chip. Also discussion of using SLI on these beast.

But its very possible there maybe a shortage of chip, not because of supply but because of demand!


RE: Supply shortgage probbaly FUD
By Viditor on 6/25/2006 12:21:37 AM , Rating: 2
quote:
Personally, I reckon the stories of tight supply on NGMA chips is probably FUD. For starters, the world is absolutely drowning in leaked engineering samples and the press are currently being buried in samples, too


Some problems with this assessment...
1. Intel's total 65nm production capacity has only just crossed the 50% mark according to Paul Otellini. This includes all 65nm chips (Yonah, Pressler, Cedar Mill, etc...)

http://www.technologynewsdaily.com/node/3453

2. Only 2 Fabs (large ones, D1D and F12) are currently running and ramping 65nm, with 2 smaller Fabs just now coming online (F24 and D1C). However D1D is already cutting back on 65nm in order to get ready for 45nm at the end of next year.

3. Producing Engineering Samples says nothing about yields. They can produce amazing chips that are low yield and still come up with 10k+ just from their attempts to tweak the yield process. However, for production that isn't cost effective...

4. Merom has a HUGE die size (almost twice that of Yonah) because it's going from 2MB to 4MB of cache. This means that they can only produce half the number with the same wafer space.

So, tight supply should be expected...and remember that it takes ~3 months to complete a wafer.


RE: Supply shortgage probbaly FUD
By caboosemoose on 6/25/2006 9:21:52 AM , Rating: 2
Yup, all fair points, although my experience in the past is that sample availability has in past definitely reflected general availability. When yields and are very low, the sample number do tend to go down and it's the press that feel it first - as you would expect, customers / system integrators are first priority for samples.

I also don't think that Intel having only 50% 65nm capacity is a major impediment. there are plenty of netburst chips being made on 65nm that I presume are being dropped to make way for NGMA chips.

In short, I'm fairly sure that as these things go (ie the launch of a brand new CPU architecture) NGMA chips will have perfectly good availability. I'm sure that anyone who really cares about these things will be able to get hold of one. And givne the great pricing Intel is they'll be getting them for surprisingly little money. From an end user perspective NGMA is all good, plain and simple.


RE: Supply shortgage probbaly FUD
By Viditor on 6/25/2006 10:25:34 AM , Rating: 2
quote:
my experience in the past is that sample availability has in past definitely reflected general availability


You are forgetting that these are unique times...

1. Intel has never been so far behind for so long before
2. Intel has never waged a preview campaign so early and with so much marketing before.
3. Intel has never lost so much revenue share (as opposed to market share) before, let alone in such a short period of time.

quote:
I also don't think that Intel having only 50% 65nm capacity is a major impediment. there are plenty of netburst chips being made on 65nm that I presume are being dropped to make way for NGMA chips

Actually, the Netburst chips won't be dropped this year...in fact Intel just released another new stepping for them.

According to Intel's own numbers, NGMA will certainly be in tight supply...but this should not come as a surprise to anyone considering that they have pushed it's release ahead so much from it's initially planned launch (Q1 07). Rembember that the equipment and everything else has to be designed and ordered years in advance...


RE: Supply shortgage probbaly FUD
By zsdersw on 6/25/2006 5:19:58 PM , Rating: 2
quote:
You are forgetting that these are unique times...

1. Intel has never been so far behind for so long before
2. Intel has never waged a preview campaign so early and with so much marketing before.
3. Intel has never lost so much revenue share (as opposed to market share) before, let alone in such a short period of time.


And what specific relevance to availability do those three things have?


RE: Supply shortgage probbaly FUD
By Viditor on 6/26/2006 2:41:56 AM , Rating: 2
quote:
And what specific relevance to availability do those three things have?

They aren't relevant to supply, they are relevant to cabooses observations on the supply of Engineering Samples (read the previous few posts...).


RE: Supply shortgage probbaly FUD
By zsdersw on 6/26/2006 6:38:56 AM , Rating: 2
Actually, you responded to his assertion that availability of engineering samples correlates to general availability, which means that your 3-item list was attempting to disprove that assertion (that the relative wide availability of engineering samples does not, in this case, indicate wide availability of the chips when they ship). In effect, you are talking about the general availability.


RE: Supply shortgage probbaly FUD
By Viditor on 6/26/2006 6:54:21 AM , Rating: 2
quote:
you responded to his assertion that availability of engineering samples correlates to general availability, which means that your 3-item list was attempting to disprove that assertion

Because he based that assertion on past experience...
My point was that times have changed as have Intel's marketing strategies (shipping of Engineering Samples in this case is a Marketing Division responsibility).


RE: Supply shortgage probbaly FUD
By zsdersw on 6/26/2006 8:46:35 AM , Rating: 2
The marketing division cannot ship what isn't available. A change in marketing strategy doesn't necessarily have anything to do with actual available supply.


RE: Supply shortgage probbaly FUD
By Viditor on 6/26/2006 9:42:58 AM , Rating: 2
quote:
The marketing division cannot ship what isn't available. A change in marketing strategy doesn't necessarily have anything to do with actual available supply

Uh-huh...but I don't get your point. caboose wasn't saying that they didn't ship ESs because they couldn't produce them, he just made the point that Intel has previously tended not to ship many ES chips when the production run was in low volume...
Producing all of the ES chips to date is like 20-30 wafers worth (out of 100s-1000s), even if the yields are terrible. Supply has nothing to do with it.

Marketing is the division that decides how many ES chips to ship and who to ship them to. Even with an absolutely horribly yielding chip, manufacturing would be able to supply whatever they need for that...


RE: Supply shortgage probbaly FUD
By zsdersw on 6/26/2006 9:49:47 AM , Rating: 2
I don't get yours either. You said that these were different times.. but everything you listed has nothing to do with available supply or is relevant to your point.


RE: Supply shortgage probbaly FUD
By Viditor on 6/26/2006 1:04:26 PM , Rating: 2
quote:
I don't get yours either

OK, let me see if I can help you...

1. Caboose said that he figures that supplies will be higher than expected on Conroe because Intel has so many Engineering samples out there. His reasoning is that the only time Intel has sent out numerous Engineering Samples in the past, they have had a large volume of production chips stockpiled.

2. I pointed out that the decision of how many ES chips to ship is a marketing decision, and it has nothing to do with what kind of inventory is available.

3. I also pointed out that even though the marketing people have always followed the pattern he suggest in the past, the situation Intel has found itself in over the last few years is probably responsible for them changing this strategy.


RE: Supply shortgage probbaly FUD
By zsdersw on 6/26/2006 5:02:12 PM , Rating: 2
The three things you listed, that I quoted above in my first comment, don't seem relevant to any of that.. which is why I originally asked of you what I did.


RE: Supply shortgage probbaly FUD
By Viditor on 6/26/2006 7:09:36 PM , Rating: 2
Their relevance is only to part 3 above...


RE: Supply shortgage probbaly FUD
By zsdersw on 6/26/2006 7:35:21 PM , Rating: 2
quote:
My point was that times have changed as have Intel's marketing ...


But then what's this about? You're apparently referring to something other than Intel's marketing strategies.


RE: Supply shortgage probbaly FUD
By hstewarth on 6/25/2006 6:57:56 PM , Rating: 2
I would expect that Woodcrest shortage is already fud, there is already an article stated that supple of Core 2 DUO will be great supply.. I would assume that Woodcrest would also be.

http://www.neoseeker.com/news/story/5889/


RE: Supply shortgage probbaly FUD
By Viditor on 6/26/2006 2:44:11 AM , Rating: 2
quote:
there is already an article stated that supple of Core 2 DUO will be great supply

Ummm... the article you linked states that:

"Taiwan-based motherboard makers doubt Intel's ability to supply a sufficient quantity of Core 2 Duo processors. Sales of the Core 2 Duo line may account for less than 10% of Intel's total CPU sales and that the first batch of the Core 2 Duo processors may be limited and will be delivered mainly to international vendors"


confused?
By maevinj on 6/23/2006 2:43:30 PM , Rating: 2
I know I'm a dumba$$, but I'm confused why intel is releasing so many different server cores in a relatively short time. I could just imagine the confusion on the dell website trying to by a server and thinking you're getting x core and actually get y core




RE: confused?
By TheDoc9 on 6/23/2006 3:09:57 PM , Rating: 3
There trying to erase their past mistakes. Problem is obviously availability. We'll have to wait a few more weeks, but it wouldn't surprise me to see extreemly short supplies and high markups for these chips.


Dedicated HS Interconnect
By vgermax on 6/23/2006 3:13:25 PM , Rating: 2
Starting with the i5000P (Blackford) chipsets, Intel now has two independent FSBs available, one for each socket. This alleviates, to a certain degree, the limitations of shared bus architecture. However, when there are multiple die per socket all the die present share one bus from the socket. AMDs implementation for multiple core CPUs remains superior in this regard. I would think this approach of multiple FSBs from a single northbridge is a stop-gap solution, although apoint-to-point interconnect from Intel seems terminally delayed.


RE: Dedicated HS Interconnect
By hstewarth on 6/23/2006 3:31:38 PM , Rating: 2
The article states that high speed interconnect is for Tigerton which comes in 2007. This is called CSI in intel terms.


RE: Dedicated HS Interconnect
By Zanfib on 6/23/2006 3:42:17 PM , Rating: 2
For sure.

AMD has the advantage of the onboard memory controller too, meaning there is quite a bit less traffic on the FSB/HT links then if the memory controller is on the motherboard. I agree HT is the more elegant solution though, even if memory traffic doesn't always need to go over it, when it does (to memory located on another socket's controller) HT is nice and fast, and scaleable.

Woodcrest has a fair amount of FSB to play with though, 1333 MHz when the cores are running a fair amount slower (compared to the 3.73 GHz Xeons) will help, but certainly only until clock speeds start increasing or both cores really want to use the FSB at the same time).


RE: Dedicated HS Interconnect
By hstewarth on 6/23/2006 3:53:36 PM , Rating: 2
But there is more than just the high speed bus of the Woodcrest. It also has FB-Dimm memory which reduces the need for memory controller - because of interface to memory.. which is suppose 4x the normal memory speed.
I belive the FB-Dimm serial memory is designed to reduce the overhead on the bus on memory - just like SATA compared to PATA does.

Intel does have in plan Intergrate memory controller which combine with higher speed bus and FB-Dimm memory should help a lot.

I think memory test have shown that its almost as good as AMD's.

The fact that the woodcrest cores are slower compared to 3.73 ghz xeon is a good thing, on 1333Mhz bus, this means that we have are going to have a lot faster Woodcrest in the future.


RE: Dedicated HS Interconnect
By Viditor on 6/24/2006 11:48:03 PM , Rating: 2
quote:
It also has FB-Dimm memory which reduces the need for memory controller - because of interface to memory.. which is suppose 4x the normal memory speed

I think you are confusing bandwidth with throughput.
FBDs are capable of writing and reading simultaneously, but there is no reduced need for the memory controller, and it certainly doesn't reduce the "overhead on the bus" (not really sure what that means though...).


RE: Dedicated HS Interconnect
By Viditor on 6/25/2006 10:28:40 AM , Rating: 2
quote:
The article states that high speed interconnect is for Tigerton which comes in 2007. This is called CSI in intel terms

Nope...the CSI interface is coming out in 2008/9. The high-speed interface they are talking about is the dual FSB.


The AM2 surprise
By crystal clear on 6/24/2006 9:29:23 PM , Rating: 2
This may not be the right article to mention but still it
is important,so read on..
The Inquirer has an article- AM2 has a secret weapon hidden.
The surprise being REVERSE HT-This theory has being floating
around for quite sometime,many have rejected outright.
I think we should look into this-if true then ofcourse it is
competition head on for Conroe.




RE: The AM2 surprise
By hstewarth on 6/24/2006 10:29:50 PM , Rating: 3
This sounds like FUD especially after reading the article. I think the idea is to supply people with misinformation to delay them to going to Core2 architexture products.

Just by name of "Reverse HT" sounds like a joke anyway.


RE: The AM2 surprise
By Viditor on 6/25/2006 12:27:00 AM , Rating: 2
quote:
Just by name of "Reverse HT" sounds like a joke anyway

Well, I don't know if they are developing it or not, but it's NOT a joke or FUD...and Intel is developing it as well.
Intel's program is called Mitosis.
http://www.intel.com/technology/magazine/research/...


RE: The AM2 surprise
By crystal clear on 6/25/2006 5:03:32 AM , Rating: 2
So we have crossed the stage-whether its a FUD/joke/misinformation/gimick etc.
ITS A REALITY-from this point,the question in my mind is
What does this do to CPU technology & where does it lead
us to?


RE: The AM2 surprise
By hstewarth on 6/25/2006 10:53:39 AM , Rating: 2
Well for one thing its not a secret weapon, if Intel also is planning on it - who knows who started it first.

Its still maybe a FUD for AM2, AMD CEO stated that the next generation of AMD Processors is coming out in 2008.

It also sounds like programs have to be compiled in specific method to handle it.

Anyway - this topic is not part of Woodcrest, unless Woodcrest has Mitosis in it.


RE: The AM2 surprise
By del on 6/25/2006 4:05:02 PM , Rating: 2
that's old news. =P And it's not officially called Reverse Hyper-Threading.

I guess one could call it single thread, multiple data, because it seems similar to what SIMD does for instructions.


Dedicated High-Speed Iinterconnect = FSB
By Jellodyne on 6/23/2006 2:35:20 PM , Rating: 2
"dedicated high-speed interconnect"

Oh, wait, intel is limited to 4 cores on their FSB, so each 4 core chip neads it's own dedicated FSB. But if they state it like that it almost sounds like they're running a modern point to point high speed interconnect, doesn't it?




By shadowzz on 6/23/2006 2:37:45 PM , Rating: 2
Yeah it's kind of a joke. All it means is indepdent FSBs per CPU. Two northbridges.


By epsilonparadox on 6/23/2006 3:12:26 PM , Rating: 2
Isn't that Hyper-Transport, a dedicated high speed interconnect?


RE: Dedicated High-Speed Iinterconnect = FSB
By hstewarth on 6/23/2006 3:29:10 PM , Rating: 2
I think there is some consfusing here, the article is talking about next generation MP processor with high speed interconnect - which I believe is called CSI in intel terms.
which would have seperate connection for each core

But the Woodcrest has seperate bus for each socket.

I believe the memory is point to point high speed interface.

Also note: the Bus/Memory interface on these chips is not the same as Conroe. It has a much more advance interface.


By Viditor on 6/24/2006 10:50:19 AM , Rating: 2
quote:
I believe the memory is point to point high speed interface

Not quite...there are seperate connections from each CPU socket to the Northbridge, but a single connection from the Northbridge to memory. So while bandwidth to the NB (FSB) is increased, bandwidth to system memory remains the same.


Seems kinda relevant
By Fnoob on 6/24/2006 9:13:58 AM , Rating: 2
Another detail to note is that Dempsey is capable of processing up to four threads simultaneously thanks to HyperThreading, versus Woodcrest's maximum of two -- HyperThreading is not enabled on the first Core 2 Duos.

Would seem like a waste not to enable this - or not to wait for processors where this is actually turned on. Perhaps I'm way off here, but isnt this akin to training a world class boxing octopus, and then tying half his arms behind his back?




RE: Seems kinda relevant
By mino on 6/24/2006 12:48:47 PM , Rating: 2
It's NOT present so it isn't enabled :) Sipmle as that.

IMHO the "no enabled" quote comes from the likes who write about "C7 has extended multimedia functionality enabled". It means it does not support it, just said in a more polite way fo intel.

Not that HT is missed on MCW core by anyone into CPU's.


RE: Seems kinda relevant
By Fnoob on 6/25/2006 9:27:32 AM , Rating: 2
Thanks for the clarification....

But I still stand by the fact that an evolutionarily "improved" CPU should be able to handle MORE threads, not less. The server we are running now is bottlenecked more by excessive threading demands than say, FSB, clock, etc.



RE: Seems kinda relevant
By Scrogneugneu on 6/25/2006 11:08:46 AM , Rating: 2
HyperThreading made little to no difference. Overall, it was more a marketing technology than a real boost in performance.

Having a processor able to perform 2 threads simultaneously, instead of 2 threads + 2 fake-threads, won't hurt performance that much. And since the new cores should be much more efficient, you'll most likely find that a 2 threads processor will perform better than the old 4 threads one.


RE: Seems kinda relevant
By hstewarth on 6/25/2006 2:15:44 PM , Rating: 2
Actually hyperthreading did make a difference - but only really before Dual core came out. On a hyperthreading machine, any you doing a long intensive task such as compiling a large code, you can smoothly do something else while the compile is going on. On the same situation, its much hard to do anythings.

Also in some situations that you can run two instanse of same application and achieve faster overall output. I notice this with graphics application call Mojoworld - I go do six renders with hyperthreading in same time as doing 4 without.

Also its important to understand that the OS is not single threaded, so multiple threads will help in this situation which could explain some of smoothness of HT machine verse a single core non HT machine.

Stating this, Dual core is much better than HT. But I would rather have HT over non HT-single core anyday.


problems w/ the article, etc.
By del on 6/25/2006 4:45:54 PM , Rating: 2
1. I always thought SMP meant multiprocessing no matter whether you have two CPUs in a system from two boxes or not. So, SMP could be one dual-core (or multi-core) CPU or a CPU with HT technology.

2. Hyper-Threading should not be written as HyperThreading. You're confusing its punctuation with HyperTransport.

3 & 4. Intel's new micro-architecture is called Core, not Core 2. Core 2 is the name of the CPU.

5. Woodcrest does not consume 65 W of electrical power, nor does Dempsey consume 130 W of power. They both dissipate 65 W and 130 W of thermal power, respectively.

6.
quote:
Another detail to note is that Dempsey is capable of processing up to four threads simultaneously thanks to HyperThreading, versus Woodcrest's maximum of two -- HyperThreading is not enabled on the first Core 2 Duos.


Woodcrest will never be known as Core 2 Duo. It will be called Xeon DP. You said it yourself.

7.
quote:
Intel stated that when compared to AMD's Opteron processor, Woodcrest can be up to 33% more efficient in power consumption. Actual tests however indicated that Woodcrest is about 10% to 15% lower in some situations.


8. "up to" isn't a guarantee. So Intel's statement already accounts for the 10 to 15% margin of error. Maybe you should try phrasing it differently.

to the posters:

The 1333 MHz FSB is actually two 667 MHz FSBs because there are two Xeon DP CPUs per system.

What does FUD mean?




RE: problems w/ the article, etc.
By hstewarth on 6/25/2006 7:02:28 PM , Rating: 2
quote:
The 1333 MHz FSB is actually two 667 MHz FSBs because there are two Xeon DP CPUs per system.


Not true: The Woodcrest platform has 2 indepentent 1333Mhz bus - 1 for each processor - but they are running at 1333Mhz speeds.

The memory however is similar to Pentium in which it has Quad channel support - so yes it has 667Mhz memory support - but that is seperate from the Bus.

I believe it actually it has 4 channels of 333Mhz to memory - DDR2 has 2 channels of 333Mhz.


RE: problems w/ the article, etc.
By del on 6/27/2006 7:00:27 PM , Rating: 2
Cool! Actually, I meant to say that's what I heard about the FSB, but I forgot to type that. So, I guess one could say two 667 MHz FSBs is "FUD."


RE: problems w/ the article, etc.
By Viditor on 6/26/2006 2:56:09 AM , Rating: 2
quote:
I always thought SMP meant multiprocessing no matter whether you have two CPUs in a system from two boxes or not. So, SMP could be one dual-core (or multi-core) CPU or a CPU with HT technology

SMP stands for Symmetric Multi-Processing. It specifies multiple processors.
SMT stands for Simultaneous Multi-Threading and incorporates HT.
The difference between the 2 is that with SMT, you can start 2 threads at the same time, but the CPU only has the resources to complete 1 at a time.
With SMP (multi-core), the CPU can completely process as many threads as it has cores for simultaneously.
quote:
They both dissipate 65 W and 130 W of thermal power, respectively

Actually, they don't do that either (though you are much closer). The TDP is a suggested guideline for OEMs to design for. Intel and AMD have totally different ideas of what is safe to guide for...Intel guides to estimated highest useage, and AMD guides to absolute theoretical maximum. This is why TDP should never be used to guage differences between power consumption (though even some of the really good review sites make this mistake).

quote:
What does FUD mean?

FUD is a phrase that was originated by the IBM marketing division many years ago...it stands for Fear, Uncertainty, and Doubt.


By hans007 on 6/23/2006 8:21:35 PM , Rating: 2
honestly this seems retarded. so the 2.93/1066 conroe is $999 and the FASTER 3ghz/1333 (both by bus and core clock) will be $850.

doesnt that just seem awful?

some mobo vendor needs to build a socket 771, single socket, unbuffered ddr2 using board just for this.




By KristopherKubicki (blog) on 6/23/2006 9:06:36 PM , Rating: 2
quote:
honestly this seems retarded. so the 2.93/1066 conroe is $999 and the FASTER 3ghz/1333 (both by bus and core clock) will be $850.

Look in the disparity with memory support though.


By hstewarth on 6/24/2006 12:52:51 AM , Rating: 2
One thing to think about the motherboards are actually about 2 to 3 times the price of desktop - but they do a lot - more - IO systems are usually a lot more power, have more memory and of course support dual cpus.

As for single socket 771, unbuffered ddr board, you will probably be destorying one of the advantages of the platform. Woodcrest platform has high speed dual bus on it and supports FB-Dimm - this is likely to be common memory for desktop by 2008 or so.. according to memory sites, it supports to be 3 to 4 times speed of normal dimm and very important to up and coming 64 bit age, its supports a lot more memory because of its serial interface.


Memory figures look wrong.
By smilingcrow on 6/23/2006 2:15:11 PM , Rating: 2
"Woodcrest will be accompanied by DDR2 memory, running at either 1066MHz or 1333MHz."

If you mean dual channel 533 or 667 then Iā€™d agree, otherwise this is baloney.




RE: Memory figures look wrong.
By hstewarth on 6/23/2006 2:22:11 PM , Rating: 2
Actually the Bus is 1066Mhz and 1333Mhz.. it has a Quad Pump bus.. Memory speed is 533Mhz and 667Mhz.


WW31 and Availability
By hstewarth on 6/23/2006 8:06:54 PM , Rating: 2
What does WW31 mean?

Also I question the validity of this available issue. I would expect the importance of these chips, that Intel will have made a major effort of making sure these chips are available.




By KristopherKubicki (blog) on 6/23/2006 9:05:53 PM , Rating: 2
WW31 means Work Week 31 -- the 31st week of the year starting on Monday.


Woodcrest availability
By bozilla on 6/28/2006 4:57:05 AM , Rating: 2
So any sightings of Woodcrest no matter whether they are from a builder or as a peripheral?

I've been waiting for this to upgrade my Xeon Noconas 3.6.




RE: Woodcrest availability
By hstewarth on 6/29/2006 12:56:10 AM , Rating: 2
I called Acme Micro and they stated early July for availability of the Woodcrest. I personaly waiting on the Supermicro X7DA3/i motherboard ( SAS Workstation ).


"dies" is the correct term
By cjc1103 on 6/23/06, Rating: 0
RE: "dies" is the correct term
By PT2006 on 6/23/2006 11:41:56 PM , Rating: 3
God I hope dailytech stops using that term. I'll just point it out now. "Dies" or "dice" is fine. DailyTech uses it in every article. It turns into a flamewar, and neither side is wrong or right.

http://www.dailytech.com/article.aspx?newsid=2977&...
http://www.dailytech.com/article.aspx?newsid=909&c...
http://www.dailytech.com/article.aspx?newsid=1761&...
http://www.dailytech.com/article.aspx?newsid=2649&...


106mhz.
By Myrandex on 6/23/2006 1:58:35 PM , Rating: 2
...should be 1066mhz.




Typo
By Nocturnal on 6/23/2006 5:57:17 PM , Rating: 2
"Is for evolution"

I'm guessing it should say "As for evolution".




"Spreading the rumors, it's very easy because the people who write about Apple want that story, and you can claim its credible because you spoke to someone at Apple." -- Investment guru Jim Cramer














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