Sign in with
Sign up | Sign in
Your question

"Intel Spins Woodcrest Launch"

Last response: in CPUs
Share
June 27, 2006 3:28:48 PM

Quote:
INTEL HAD THE Woodcrest love in, and once again, I got a lot of good Windows spider games in, but somehow was left unfulfilled. The visual presentations were stunning*, and the news was a little stale, in fact, I can't think of anything that was said that has not been common knowledge since spring IDF.
But there were a few points that were odd for Intel to mention, and at least one that was flat-out wrong. The first was its spinning of the IMC. Intel officially poo-poos the lack of an IMC on the server, mainly because their first few stabs at it fell really flat.

Today, they spun it as a win because they could put 4MB of cache on the chip to allow for 50 per cent less memory accesses than AMD. Very true, but there are two problems: cost and cost.

The first cost is die area on your most expensive parts. Intel does not have an IMC, so it has to eat the cost on a cutting-edge process instead of on a -1 or a -2 process that it builds the chipsets on, or an IMC on the chip itself.

The one I thought was flat-out wrong was Intel's claim of virtualisation leadership. Intel does not have an IMC, AMD does. Intel can not virtualise memory for the time being, AMD can. Guess what the computer does a lot of? Hint, see the last paragraph. I doubt this claim to the point of almost ruining a monitor by coughing up Diet Berries and Cream Dr. Pepper when I heard it.

Another problem was Pixar, or at least Intel's touting of Pixar, and again, there were two problems there. The first is that Pixar uses P4s, not because they are better, but because Intel gave them such a sweatheart deal on them that AMD couldn't compete. It may have been the worst solution possible, but money talks.


http://www.theinquirer.net/?article=32649
June 27, 2006 3:40:29 PM

June 27, 2006 3:41:35 PM

Why do you keep posting this cr@p from The Inquirer here?

It's easier for you to read it on their own website, where the words are bigger than here. If you're doing it to help the rest of us who can read small words and joined up writing now, save yourself the trouble and go back to the alphabet.
Related resources
June 27, 2006 3:46:26 PM

Quote:
INTEL HAD THE Woodcrest love in, and once again, I got a lot of good Windows spider games in, but somehow was left unfulfilled. The visual presentations were stunning*, and the news was a little stale, in fact, I can't think of anything that was said that has not been common knowledge since spring IDF.
But there were a few points that were odd for Intel to mention, and at least one that was flat-out wrong. The first was its spinning of the IMC. Intel officially poo-poos the lack of an IMC on the server, mainly because their first few stabs at it fell really flat.

Today, they spun it as a win because they could put 4MB of cache on the chip to allow for 50 per cent less memory accesses than AMD. Very true, but there are two problems: cost and cost.

The first cost is die area on your most expensive parts. Intel does not have an IMC, so it has to eat the cost on a cutting-edge process instead of on a -1 or a -2 process that it builds the chipsets on, or an IMC on the chip itself.

The one I thought was flat-out wrong was Intel's claim of virtualisation leadership. Intel does not have an IMC, AMD does. Intel can not virtualise memory for the time being, AMD can. Guess what the computer does a lot of? Hint, see the last paragraph. I doubt this claim to the point of almost ruining a monitor by coughing up Diet Berries and Cream Dr. Pepper when I heard it.

Another problem was Pixar, or at least Intel's touting of Pixar, and again, there were two problems there. The first is that Pixar uses P4s, not because they are better, but because Intel gave them such a sweatheart deal on them that AMD couldn't compete. It may have been the worst solution possible, but money talks.


http://www.theinquirer.net/?article=32649You forgot this part, intentionally i'm sure.
"All this said, Intel does have a great CPU. It was delivered early, will ramp really fast, and has a stunningly good chipset that no one seems to notice. For the next few quarters, it has the lead in several key areas, no question there. If it could just lose the spin, I would be much happier."

http://www.theinquirer.net/?article=32649

Hmmm, what would you do without Charlie and Sharikou Demerjian's constant articles of denial?
June 27, 2006 4:10:28 PM

Quote:
INTEL HAD THE Woodcrest love in, and once again, I got a lot of good Windows spider games in, but somehow was left unfulfilled. The visual presentations were stunning*, and the news was a little stale, in fact, I can't think of anything that was said that has not been common knowledge since spring IDF.
But there were a few points that were odd for Intel to mention, and at least one that was flat-out wrong. The first was its spinning of the IMC. Intel officially poo-poos the lack of an IMC on the server, mainly because their first few stabs at it fell really flat.

Today, they spun it as a win because they could put 4MB of cache on the chip to allow for 50 per cent less memory accesses than AMD. Very true, but there are two problems: cost and cost.

The first cost is die area on your most expensive parts. Intel does not have an IMC, so it has to eat the cost on a cutting-edge process instead of on a -1 or a -2 process that it builds the chipsets on, or an IMC on the chip itself.

The one I thought was flat-out wrong was Intel's claim of virtualisation leadership. Intel does not have an IMC, AMD does. Intel can not virtualise memory for the time being, AMD can. Guess what the computer does a lot of? Hint, see the last paragraph. I doubt this claim to the point of almost ruining a monitor by coughing up Diet Berries and Cream Dr. Pepper when I heard it.

Another problem was Pixar, or at least Intel's touting of Pixar, and again, there were two problems there. The first is that Pixar uses P4s, not because they are better, but because Intel gave them such a sweatheart deal on them that AMD couldn't compete. It may have been the worst solution possible, but money talks.


http://www.theinquirer.net/?article=32649
Good find 9nm, now you can take your pills and try to sleep, after the Woodcrest release trauma.
June 27, 2006 4:15:17 PM

Quote:
Why do you keep posting this cr@p from The Inquirer here?


Arrggggg the theinquirer again... damn it all. they need to make a sticky about using good sources for posts.

Well known opinions about theinquirer ... :
"The Inquirer's articles are written with a subjective and opinionated tone" also known as bad journalism.

"mimics the style of reporting common in the British tabloid newspapers (for reference check out The Sun)."

"The Inquirer has been accused of manufacturing rumours or speculating aimlessly in order to inflate its number of visitors."

------------

Do you know what a tabloid newspapaer is and why we dont respect them or use them as credible sources?
"used to refer to newspapers focusing on less "serious" content, especially celebrities, sports, sensationalist crime stories and even hoaxes, "

"Tabloid is also known as the gutter press by people who wish to express it in a negative manner."

p.s. and yes I realize that you could argue wikipedia.org is a bad source that is constantly wrong. but if you like I can find other sources that corroborate with wikipedia. (if you didn't already realize wikipedia authors use sources for their material.)
June 27, 2006 4:53:16 PM

AMD are always going to show benches where large bandwidth communication between multiple CPUs is required (because of their strong multi-processor links) and Intel are always going to show benchmarks which require large caches to perform well.

They'd be totally stupid if they did otherwise, I don't see why you have such a problem with it.
June 27, 2006 5:41:53 PM

Quote:
AMD are always going to show benches where large bandwidth communication between multiple CPUs is required (because of their strong multi-processor links) and Intel are always going to show benchmarks which require large caches to perform well.

What you've said is contrary. More cache means less memory traffic. More memory badnwidh does not mean more performance. We have seen that twice with the K8. The s939 with twice the bandwidth(2 channels instead of one) of s754 has 0-5% better performance, the sAM2 with twice bandwidth(800MHz DDR2 instead of 400MHz DDR) is performing same as same clocked s939. So AMD have no card to play on bandwidth.
June 27, 2006 6:44:47 PM

I'm going on information from the recent 'Woodcrest vs. Opteron' article:

Quote:
HyperTransport interconnects processors and core logic. This point-to-point interconnect scales much better than bus interfaces.


and

Quote:
The Front Side Bus is the interface and a potential bottleneck between the processor(s) and the chipset northbridge.


And not even shakira could argue that Opteron's cache is better than Woodcrest's.
June 27, 2006 7:11:21 PM

God 9-inch your like a media pundit, enough is enough!
June 27, 2006 7:13:15 PM

Quote:
AMD are always going to show benches where large bandwidth communication between multiple CPUs is required (because of their strong multi-processor links) and Intel are always going to show benchmarks which require large caches to perform well.

What you've said is contrary. More cache means less memory traffic. More memory badnwidh does not mean more performance. We have seen that twice with the K8. The s939 with twice the bandwidth(2 channels instead of one) of s754 has 0-5% better performance, the sAM2 with twice bandwidth(800MHz DDR2 instead of 400MHz DDR) is performing same as same clocked s939. So AMD have no card to play on bandwidth.

Servers use bandwidth differently than desktop. Your FUD has been bursted. :wink:
June 27, 2006 7:43:58 PM



Cliff exit stage right!
June 27, 2006 8:03:09 PM

Quote:
I'm going on information from the recent 'Woodcrest vs. Opteron' article:

HyperTransport interconnects processors and core logic. This point-to-point interconnect scales much better than bus interfaces.


and

Quote:
The Front Side Bus is the interface and a potential bottleneck between the processor(s) and the chipset northbridge.


And not even shakira could argue that Opteron's cache is better than Woodcrest's.


Woodcrest problems starts when you add more than 8 cores, then you need some kind of interconnec. Thats why Intel works in CSI.
June 27, 2006 8:09:16 PM

Quote:
INTEL HAD THE Woodcrest love in, and once again, I got a lot of good Windows spider games in, but somehow was left unfulfilled. The visual presentations were stunning*, and the news was a little stale, in fact, I can't think of anything that was said that has not been common knowledge since spring IDF.
But there were a few points that were odd for Intel to mention, and at least one that was flat-out wrong. The first was its spinning of the IMC. Intel officially poo-poos the lack of an IMC on the server, mainly because their first few stabs at it fell really flat.

Today, they spun it as a win because they could put 4MB of cache on the chip to allow for 50 per cent less memory accesses than AMD. Very true, but there are two problems: cost and cost.

The first cost is die area on your most expensive parts. Intel does not have an IMC, so it has to eat the cost on a cutting-edge process instead of on a -1 or a -2 process that it builds the chipsets on, or an IMC on the chip itself.

The one I thought was flat-out wrong was Intel's claim of virtualisation leadership. Intel does not have an IMC, AMD does. Intel can not virtualise memory for the time being, AMD can. Guess what the computer does a lot of? Hint, see the last paragraph. I doubt this claim to the point of almost ruining a monitor by coughing up Diet Berries and Cream Dr. Pepper when I heard it.

Another problem was Pixar, or at least Intel's touting of Pixar, and again, there were two problems there. The first is that Pixar uses P4s, not because they are better, but because Intel gave them such a sweatheart deal on them that AMD couldn't compete. It may have been the worst solution possible, but money talks.


http://www.theinquirer.net/?article=32649You forgot this part, intentionally i'm sure.
"All this said, Intel does have a great CPU. It was delivered early, will ramp really fast, and has a stunningly good chipset that no one seems to notice. For the next few quarters, it has the lead in several key areas, no question there. If it could just lose the spin, I would be much happier."

http://www.theinquirer.net/?article=32649

Hmmm, what would you do without Charlie and Sharikou Demerjian's constant articles of denial? What do you have to say about this, 9-Inch? You and Sharikou would be excellent politicians, spinning everything into your favor.
June 27, 2006 8:19:01 PM

Quote:
Servers use bandwidth differently than desktop. Your FUD has been bursted. :wink:

Cool, no go take your pills and remain being sad becouse Woodcrest wipes the floor with Opteron
June 27, 2006 11:10:16 PM

Quote:
Servers use bandwidth differently than desktop. Your FUD has been bursted. :wink:


You use words different to the rest of us. You must be a financial analyst too. :roll:
June 28, 2006 1:47:48 AM

[/quote]
Quote:
Woodcrest problems starts when you add more than 8 cores, then you need some kind of interconnec. Thats why Intel works in CSI.


Corrected quote:
Quote:
Woodcrest problems starts when you add more than 4 cores, then you need some kind of interconnec. Thats why Intel works in CSI.

:wink:
June 28, 2006 2:05:00 AM

Quote:
Woodcrest problems starts when you add more than 8 cores, then you need some kind of interconnec. Thats why Intel works in CSI.


Corrected quote:
Quote:
Woodcrest problems starts when you add more than 4 cores, then you need some kind of interconnec. Thats why Intel works in CSI.

:wink:



Considering the fact that the preliminary Kentsfield @ 1066Mhz benchmarks haven't shown a measurable FSB bottleneck, I doubt it would be much of a problem with an 8-core system. Remember, it will be two FSB's operating at 1333Mhz. That's 2.5x the theroetical bandwidth of a single Kentsfield, with only twice as many cores. Anything past a 4S server though, and Intel is pretty much screwed.
June 28, 2006 5:06:42 AM

The sheer amount of pwnage in that post is almost painful.
June 28, 2006 7:53:00 AM

Why does this kid still have the ability to post here? When he runs out of things to say and suspects there might be some pro-AMD something in an article, he throws it up. Seriously, every other thread he makes is about finding some benchmark somewhere that says Intel is better and AMD sucks and attempts to spin it the other way.

You're like liberal America kid, you find half the story and if it fits your agenda, then by God you'll use it. Screw the truth.
June 28, 2006 11:46:28 AM

Are you SM?
Because I get the idea that you like being spanked...
June 28, 2006 12:07:57 PM

Couldn't sleep Jack? Good post. Bit of a waste on 9-inch tho (he still wont understand).
June 28, 2006 3:47:47 PM

This is what G0jD0 said
Quote:
What you've said is contrary. More cache means less memory traffic. More memory badnwidh does not mean more performance. We have seen that twice with the K8. The s939 with twice the bandwidth(2 channels instead of one) of s754 has 0-5% better performance, the sAM2 with twice bandwidth(800MHz DDR2 instead of 400MHz DDR) is performing same as same clocked s939. So AMD have no card to play on bandwidth.


Socket AM2 does benefit more from lower latency modules and a "little" from more bandwidth (I'm talking desktops here) so maybe he's right.

And since we're talking about servers here (just look at the topic's name), this is what I've said:
Quote:
Servers use bandwidth differently than desktop. Your FUD has been bursted


Well, you're right since the proper name is not FUD. Maybe we should call it bullzhit. :wink:

Maybe you own me an apology. 8)
June 28, 2006 3:51:58 PM

Oh, I thought it was you whose username was BullSh!tter and RettihslluB
!