Analyst: Intel's Quad-Core Chip May Fall Short

Duke3d87

Limp Gawd
Joined
Jul 6, 2004
Messages
430
umner Lemon, IDG News Service
Wednesday, March 01, 2006

An upcoming Intel quad-core Xeon server processor, codenamed Clovertown, likely contains two dual-core chips in a single package, according to an industry analyst, who warned the processor may not offer a significant boost in performance.

"Intel hasn't said whether it architected Clovertown as a quad-core processor, or merely combined a pair of dual-core chips in a multi-chip package," said Nathan Brookwood, an analyst at Insight64, in a research note. Clovertown is expected to hit the market in early 2007.

While Intel has yet to offer details of the Clovertown design, Brookwood believes the chip uses two Woodcrest dice, or pieces of silicon, housed in a single package. Woodcrest is the codename of an upcoming dual-core Xeon chip developed by Intel, he said.

Intel used a similar "ad hoc" design for its first generation of dual-core processors, which included either two cores on a single die or two single-core dice housed in one package, Brookwood said.

"We've seen this movie and it didn't have a happy ending then, either," he wrote.

Performance Problems?

The problem is performance. Intel's earlier dual-core chips contained two cores that shared a single front-side bus, which is housed outside the chip, Brookwood said. The front-side bus connects a computer's processor with main memory. While Intel's first dual-core chips outperformed a single-core chip, a better dual-core design would likely have meant even greater performance, he said.

The Blackford chip set designed to be used with the dual-core Woodcrest chip addresses the performance issue in earlier chips by offering two independent front-side buses, one for each core. However, those same performance issues return with the Clovertown design, which has two cores sharing each of Blackford's two front-side buses, Brookwood said.

As a result of this design choice, Clovertown's performance will only be "marginally better" than Woodcrest in most applications, Brookwood said. The performance would be better if Intel used a better processor architecture, he said.

"We are confident that eventually Intel will introduce quad-core Xeon processors with competitive performance, but we doubt that Clovertown will be the vehicle that meets this target," he wrote.


Source

Intel employees any rebutal? Nathan posted one post on Geek.com and said that he was very dissapointed with the way that cloverton was made. I think we talked about some of the aspects of cloverton here, but we never really talked about the performance increases that Cloverton will include. Brookwood makes it sound like Cloverton will be very slow and get smashed by the opteron.
 
Brookwood isn't an engineer. Nor does he have access to Clovertown samples.

That's all that needs to be said.
 
Mako360 said:
Brookwood isn't an engineer. Nor does he have access to Clovertown samples.

That's all that needs to be said.
are you also Intel? Cause if so, that' makes three people here.
 
I really don't have much info on Clovertown other than they just did Power On with the platform so I doubt that this guy really knows what he's talking about, or at the very least it's VERY speculative. I can tell you that there are differences between the Bensley platforms that run clovertown and the ones that run dempsey/woodcrest, but I cannot speak to what those differences are. If I who works on enterprise platforms for Intel doesn't have much info on Clovertown, it's kinda amazing that this guy does. :D
 
Poncho said:
I really don't have much info on Clovertown other than they just did Power On with the platform so I doubt that this guy really knows what he's talking about, or at the very least it's VERY speculative. I can tell you that there are differences between the Bensley platforms that run clovertown and the ones that run dempsey/woodcrest, but I cannot speak to what those differences are. If I who works on enterprise platforms for Intel doesn't have much info on Clovertown, it's kinda amazing that this guy does. :D
Good point. Also, at last IDF, Intel showed the world a CPU that had an integrated NB and voltage regulators all within the CPU. I know CSI shoud be out in a two yearsish, but when does Intel plan on going full production with the integrated voltage regulator? And, would it really cut power consumption all that much?
 
rayman2k2 said:
lemme speculate too:

Intel's 16-core chip will suck.


Maybe... but their 32 core proc will kick some major ass, trust me I've seen it.... Otellini has one in his office. It ROXORS!@@!!!@!@!@!@!



LOL
 
Analyst: I'm pulling this out of my ass, yet all major networks and so called "Tech" publications will pick this up and report on it as if it's the gospel, instead of pwning my useless analysis on hardforums. :D
 
Poncho said:
Maybe... but their 32 core proc will kick some major ass, trust me I've seen it.... Otellini has one in his office. It ROXORS!@@!!!@!@!@!@!



LOL
are you for real or are you joking. If i was Ontellini, i would start folding with that box. But, do you have any non NDA bound info about the integrated voltage controllers? And are you going to IDF?
 
Analyst: Intel will place lasers on the heads of multiple sharks and take over the world.

Intel hasn't even beta tested the laser, god knows they haven't trained the sharks. :rolleyes: :)
 
Duke3d87 said:
are you for real or are you joking. If i was Ontellini, i would start folding with that box. But, do you have any non NDA bound info about the integrated voltage controllers? And are you going to IDF?


Yea, I was joking. Maybe I should have said so. :D As for the Voltage Controllers... I really don't know anything about them and no, I'm not going to IDF. :(
 
serbiaNem said:
Analyst: Intel will place lasers on the heads of multiple sharks and take over the world.

Intel hasn't even beta tested the laser, god knows they haven't trained the sharks. :rolleyes: :)

LMAO!!!!!!
 
everyone says "no one knows how good conroe is. there's no benchmarks, proof, etc.". why doesn't that same logic apply here?

analysts can make predictions all they want, but until the actual thing is out it's all speculation.
 
Actually, tehre are people that know how Conroe is, albeit preproduction versions. If Conroe is launching in Q3 as planned, some folks (MB manufactureuers, Intel internal labs, computer companies like Dell, etc) have to have samples. There'll be NDAs to keep them from posting detailed data though.
 
Legenic said:
everyone says "no one knows how good conroe is. there's no benchmarks, proof, etc.". why doesn't that same logic apply here?

analysts can make predictions all they want, but until the actual thing is out it's all speculation.

Here's the thing.... Clovertown JUST powered on and the boards that they run on were JUST built. So they are VERY limited at the moment which means that they will probably be only used internal for a while where as Merom has been available for months. And as for me... I've used Merom/Conroe/Woodcrest..... but not Clovertown so I won't draw any conclusions on it's performance.
 
Poncho said:
Here's the thing.... Clovertown JUST powered on and the boards that they run on were JUST built. So they are VERY limited at the moment which means that they will probably be only used internal for a while where as Merom has been available for months. And as for me... I've used Merom/Conroe/Woodcrest..... but not Clovertown so I won't draw any conclusions on it's performance.
if it's been availiable for a couple of months, why is Intel pushing its launch back to Q4?
 
Duke3d87 said:
if it's been availiable for a couple of months, why is Intel pushing its launch back to Q4?

If you are talking about Merom.... it taped out months ago and is still going through validation. Also you have to consider the chipset. Not sure if it's going to release on 945GM or Crestline. And beyond that... no idea. Those decisions are WAY above my pay grade. :D
 
Poncho said:
If you are talking about Merom.... it taped out months ago and is still going through validation. Also you have to consider the chipset. Not sure if it's going to release on 945GM or Crestline. And beyond that... no idea. Those decisions are WAY above my pay grade. :D
wait so i'll take it that Conroe and Woodcrest are still going through validation? How is it that Conroe and Woodcrest are supposed to be shipped in Junish and Merom late if they are all based on the Merom core?
 
well, supersevers need that crap, but thats a very small market, and if they do put 2 dual core xeons on a single mobo, its gonna gobble a helluva lot of power like more then most people need. like you would need HUGE domain for that power to b needed.

and my kitten is now sitting on the monitor :p
 
Duke3d87 said:
wait so i'll take it that Conroe and Woodcrest are still going through validation? How is it that Conroe and Woodcrest are supposed to be shipped in Junish and Merom late if they are all based on the Merom core?

I would imagine that it's more based on the chipset than anything. If they decide to launch on 945 then it's probably all good, though if they are planning on a release with crestline then it'll get pushed I'm sure. I know that the software validation doesn't have crestline yet, so I'm not sure where it's at. Conroe and Woodcrest already have platforms to run on... 975 and Bensley.
 
Poncho said:
I would imagine that it's more based on the chipset than anything. If they decide to launch on 945 then it's probably all good, though if they are planning on a release with crestline then it'll get pushed I'm sure. I know that the software validation doesn't have crestline yet, so I'm not sure where it's at. Conroe and Woodcrest already have platforms to run on... 975 and Bensley.
i think you're right. I remember tomshardware talking about the mobile chipset for Merom using too much power under vista so they are going to fix that. Another thing...if Paul Ontellini is on the board of Google, and he knows that woodcrest is coming out soon, why didn't he tell google to wait like three months before going out and buying purely opteron systems. That does not make sense.
 
Merom will be available in two versions: a] Calistoga chipset (667Mhz FSB) and b] Crestline chipset (800Mhz). The first platform is pin compatible with Yonah while the second one is of a different pin out.
 
Back
Top