Tom's Guide Forum
  Tom's Guide Forums » CPU & Components » CPUs » Intel says Penryn "complete"
 

Add a reply



Word :   Username :  
 
 Page : 1 2 3
Next
Previous
Bottom 
Author
 Thread : Intel says Penryn "complete"
 
Barcelona_Xtreme
More Information
n°1457467
01-19-2007 at 03:45:58 AM

Last message on previous page:

Quote :

True - It's starting to look like x86 is currently the only way to go at any platform level.



I do agree with you on this one.

x86 is the driving force on desktops, laptops and servers. Even Power and itanium are being displaced by x86 processors in the HPC segment.

Related Product

Register or log in to remove.

Barcelona_Xtreme
More Information
n°1457534
01-19-2007 at 04:13:29 AM

And talking about x86 in HPC, AMD scored another win with this supercomputer sold to the UNAM (Mexico's national Autonomous university):
AMDZone

epsilon84
More Information
n°1457544
01-19-2007 at 04:19:01 AM

Quote :

The extra cache, faster bus, double pumped ALU, vector processing execution units, high K dielectric junctions, reduced leakage, and smaller features likely slow it down too...



Intel won't be upgrading the NGMA's core with penryn. As I said, it's only a die shrink with some SSE4 instructions and some cache. That's all.

Expect a core upgrade in 2008 with Nehalem (which will be competing with upgrades made to K8L and DC 2.0)

It'll also scale to higher clockspeeds than Conroe at lower wattages. A miracle aside, I think its safe to say that Penryn will scale to much higher clockspeeds than Barcelona.

Let's assume Penryn will have a 25% clock for clock advantage against K8, it would mean Barcelona will have to improve over K8 >30% to have a notable lead over Penryn clock for clock.

From the various roadmaps I've seen, the highest rumoured clockspeed for Barcelona is 2.9GHz, while Penryn has been touted to be clocked between 3.4 to 3.73GHz for QC and 3.5 - 4GHz for DC.

If these rumoured clockspeeds are remotely accurate, Barcelona will have to be at least 40% faster than K8 to be competitive, and 50% faster to be outright outperforming Penryn. That is quite an ask, despite all the enhancements Barcelona has over K8.

Barcelona_Xtreme
More Information
n°1457625
01-19-2007 at 04:59:59 AM

Quote :

It'll also scale to higher clockspeeds than Conroe at lower wattages. A miracle aside, I think its safe to say that Penryn will scale to much higher clockspeeds than Barcelona.



I don't count higher clock speed as an upgrade. Indeed, AMD will launch 90nm K8s at 3.2GHz. If clock speed is all that important to you, expect more refinements to AMD's 65nm process. They already have the low power crown, now they need to enhance the clock speeds.

And about intel's 45nm offerings, I'd like you to take a read of this (it's from AMDZone's Scientia):

Quote :

In 2006 Intel had roughly a 9 month or 3 quarter lead on AMD for 65nm. And, although some have went on and one about 45nm for Intel it looks like Intel will lose ground in terms of lead time.

To the best of my knowledge, AMD will release 45nm mid 2008. This means that AMD has to start production about the beginning of Q2 08 to have cpu's in the channel for release. Now, I have heard of estimates of later than this like mid Q2 which would mean a release about a month later like beginning of August.

So, if AMD is going to begin production at the beginning of Q2 08 then Intel would need to begin production at the beginning of Q3 07. This definitely will not happen. Much more likely is production at the beginning of Q4 07. This would mean that Intel's lead would have been reduced from 9 months to 6 months.

If you look at the statement that production begins in the 2nd half but that 3 FABs will be producing 45nm in 2008 then you can assume that initial production is D1D. This could very well mean that 45nm will only be low volume until Q1 08. So, it seems likely that Intel's lead time will be reduced from 9 months to 6 months or perhaps even 4 1/2 by the time production really starts.



And...

Quote :

I've only seen articles which brag about Intel's 45nm process which people (such as those on forumz) claim will be incredibly fast and low power. However, this is from the 13th of December 2006:

The continued enhancement of AMD and IBM’s transistor strain techniques has enabled the continued scaling of transistor performance while overcoming industry-wide, geometry-related scaling issues associated with migrating to 45nm process technologies. In spite of the increased packing density of the 45nm generation transistors, IBM and AMD have demonstrated an 80 per cent increase in p-channel transistor drive current and a 24 per cent increase in n-channel transistor drive current compared to unstrained transistors. This achievement results in the highest CMOS performance reported to date in a 45nm process technology.



Here is the link if in case: AMDZone

Quote :

If these rumoured clockspeeds are remotely accurate, Barcelona will have to be at least 40% faster than K8 to be competitive, and 50% faster to be outright outperforming Penryn. That is quite an ask, despite all the enhancements Barcelona has over K8.


K8L only needs to be 35% better than K8 to beat the C2D uArch. That's all.

HotFoot
More Information
n°1457667
01-19-2007 at 05:18:27 AM

Intel only had a 9-month lead in 65nm? What are the actual dates of retail availability of Intel's first 65 nm parts versus AMD's? It seems to me that Intel launched 65nm parts in Dec '05, while AMD "launched" theirs in Dec '06, while we're still waiting for parts availability. Maybe I'm recalling the dates wrong, but that's a full 12 month lead.

And in terms of AMD's plans to do some catching up with the 45 nm node, it seems more and more likely that they'd have to have parts out in Q1 or Q2 2008 to be only 6 months behind Intel this time around.

Barcelona_Xtreme
More Information
n°1457682
01-19-2007 at 05:28:51 AM

Quote :

Intel only had a 9-month lead in 65nm? What are the actual dates of retail availability of Intel's first 65 nm parts versus AMD's? It seems to me that Intel launched 65nm parts in Dec '05, while AMD "launched" theirs in Dec '06, while we're still waiting for parts availability. Maybe I'm recalling the dates wrong, but that's a full 12 month lead.


He answered that also:

Quote :

abinstein wrote: ›Apple using 65nm Core Duo was shipping in January 2006. I think it's about a year ahead of AMD's 65nm shipping, not 9 months.



Yes, begining of Q1 to beginning of Q4 is 3 quarters or 9 months.

HotFoot
More Information
n°1457694
01-19-2007 at 05:34:37 AM

Isn't the beginning of Q4 October? I didn't think AMD had 65 nm parts available for sale in October.

enewmen
More Information
n°1457696
01-19-2007 at 05:35:44 AM

What's a VRM? Sorry for the stupid question.

cwj717
More Information
n°1457707
01-19-2007 at 05:42:04 AM

Quote :

Conroe is no Netburst. Conroe is already 20-30% faster per clock than K8... Do you realize how much K8L would have to improve over K8 to make it as fast as, let alone faster than a higher clocked Conroe?



LOL. :)

Where do you get your numbers from?
And please, don't tell me that JumpingJack (or some other intel troll) have told you so).

Conroe is only 10-15% better than K8 OVERALL!!!
Also, don't count K8L yet. LMAO. I was kind of expecting you to say that...so instead of pulling numbers out of your a$s, why don't you back it up with some data?

EDIT: Here is a link... http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2795&p=1

Ok, well after look over these benchmark results... I guess its more like 15-66% :) If you would like I could make a table and graph the results...

Judging from clairvoyant's post, C2D is 18% better in real world usage than K8 (what I've been saying), but you have to take that chart with a grain of salt since it was made from an intel fanboy and with THG's benchmark scores. I wasn't talking about that chart... Look at the data yourself. There are plenty of C2D reviews online. After looking at the chart posted here, it looks fine to me... although this data was not taken from THG but from Anandtech. Also, the average in that chart is a bit off because the benchmark results with the lowest difference are mostly IO bound... that throws off the mean average some. BTW, you still have not provided any data to back what you have said.

cwj717
More Information
n°1457715
01-19-2007 at 05:50:47 AM

Voltage Regulator Module...

enewmen
More Information
n°1457730
01-19-2007 at 05:56:32 AM

Can the VRM be upgraded (or downgraded) for a p965?

enewmen
More Information
n°1457803
01-19-2007 at 07:14:39 AM

I was hoping that was something I can just pop out.
Thanks for the reply...

enewmen
More Information
n°1457809
01-19-2007 at 07:18:32 AM

Thanks. I learn more every day.

qcmadness
More Information
n°1457811
01-19-2007 at 07:18:45 AM

Quote :

I was hoping that was something I can just pop out.
Thanks for the reply...



Sadly, no --- it was the older VRM's that prevented Core 2 Duo from being a drop-in on existing 775 MBs.... I have not heard if Penryn will require new VRMs, I doubt it will but that is speculation.

Jack

I guess Intel will require us to upgrade our boards for 45nm products :?
Especially if it wants to bundle Bearlake chipset with 45nm CPUs.

enewmen
More Information
n°1457818
01-19-2007 at 07:23:59 AM