Print 23 comment(s) - last by Dactyl.. on Aug 9 at 1:00 AM

A CPU-Z screen shot of Wolfdale  (Source: HKEPC)
"Penryn's" first showing reveals impressive results

Intel expects to begin releasing its 45nm Penryn family next quarter, beginning with the Xeon family. The company plans to release the desktop Core 2 variants in early 2008. The desktop variants of Penryn include the quad-core Yorkfield and dual-core Wolfdale processors.

Aside from the obvious 45nm die shrink, Penryn features a few additional core enhancements. Despite the enhancements, Penryn is still a Core 2 architecture processor. Penryn is more or less a mid-life cycle refresh of the Core 2 architecture, similar to how car manufacturers update the front fascia and give a car more power halfway through its life cycle.

With the core enhancements, a Penryn should theoretically perform faster than a Conroe or Kentsfield, given the same clock speeds. HKEPC managed to obtain a dual-core Wolfdale for some benchmarking. HKEPC pits a 2.33 GHz, 1333 MHz front-side bus with 6MB L2 cache Wolfdale up against a 2.33 GHz, 1333 MHz front-side bus with 4MB L2 cache Core 2 Duo E6550.

The site benchmarks the two processors in various benchmarks such as CineBench 9.5, ScienceMark 2.0, DivX 6.6 Alpha, Far Cry, Half Life 2 and more. The Wolfdale sample is at least 5% faster in most tests. There is a notable 19% performance improvement in the CineBench 9.5 OpenGL software test.

The biggest performance improvement occurs with SSE4 optimized applications. HKEPC obtained an SSE4 optimized version of DivX 6.6, in alpha form. Wolfdale manages a 116% performance lead over the Core 2 Duo E6550 when converting a 1080p MPEG2 file to MPEG4.

Intel has managed to reduce power consumption with Wolfdale as well. HKEPC tested Wolfdale and Conroe with Enhanced Intel Speed Step disabled. The results show Wolfdale consuming 18 watts less while running 10 degrees Celsius cooler. Under load, Wolfdale consumes 24 watts less power while running 12 degrees Celsius cooler.

The early performance and power results of Wolfdale are quite impressive. If Intel can deliver the same results with Yorkfield, the CPU wars will be quite interesting – especially with Agena right around the corner and Deneb following shortly afterwards.

Comments     Threshold

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

Better than a refresh, IMO
By therealnickdanger on 8/8/2007 8:17:03 AM , Rating: 3
Incredible, really:

Divx 6.6 Alpha w/SSE4
1080p Mpeg2 to Mpeg4
65nm 2.3GHz C2D - 69s
45nm 2.3GHz C2D - 32s

An improvement of +115.63%.

The majority of gains appear to be between 6-15% across the board. 30% gain in Half-Life 2.

Really impressive power/heat numbers too!

65nm 2.3GHz C2D - 61W/41C
45nm 2.3GHz C2D - 41W/31C

65nm 2.3GHz C2D - 83W/49C
45nm 2.3GHz C2D - 59W/37C

RE: Better than a refresh, IMO
By omnicronx on 8/8/2007 10:51:14 AM , Rating: 2
correct me if i am wrong, but do not all the tests run, except for the games make use of sse4? how much of improvement would you see in normal situations i wonder? A small 10% or less theoretical improvement in most normal applications is nothing to write home about. Guess well have to wait for agena too see..

By therealnickdanger on 8/8/2007 11:11:08 AM , Rating: 2
Should tests involving SSE4 be excluded? Are they less valuable?

RE: Better than a refresh, IMO
By Spivonious on 8/8/2007 12:43:21 PM , Rating: 2
Seeing a 10% improvement solely due to a die shrink would be incredible IMO.

By therealnickdanger on 8/8/2007 4:09:57 PM , Rating: 2
HKEPC shows that SSE4 enhancements were only used for the DivX 1080p compression, which shows the most dramatic improvement. I believe we can accept all the other benchmarks as "normal applications". The MainConcept 1080p compression was only 12% and Windows Movie Maker was on the very low end of 3%.

Keep in mind that this CPU has 50% more cache... (from 4MB to 6MB)

By JumpingJack on 8/8/2007 11:32:01 PM , Rating: 2
correct me if i am wrong, but do not all the tests run, except for the games make use of sse4?

You are wrong -- the SSE4 features are new instructions and will only show performance gains for apps that have been recompiled to take advantage of those instructions. SSE4 consists of 47 new instructions themselves targeted mostly for gaming and multimedia.

The enumeration of the instructions and what they are used for can be found in the table in this PDF:

Of the benchmarks, only the DivX alpha has been identified as being compiled to use SSE4, all the other benchmarks are standard fare -- Cinbench 9.5, for example, was released before the SSE4 instructions were announced.

RE: Better than a refresh, IMO
By maroon1 on 8/8/2007 2:43:35 PM , Rating: 2
And don't forget that those games don't utilize two cores....

If each core is about 10-15% percent faster, then games that utilize two cores will show around 20-30% improvement. Correct me if I'm wrong !!

RE: Better than a refresh, IMO
By smitty3268 on 8/8/2007 11:04:00 PM , Rating: 2
You're wrong. If it's already using 2 cores and each core adds 10-15% then the entire game is sped up by 10-15%. It's the average of the two, not the sum. That also assumes the game is CPU limited and balanced well enough to completely use both cores. In general one of those cores isn't going to add as much, and most gaming situations today are more limited by the graphics card.

By 3kliksphilip on 8/8/2007 7:36:20 PM , Rating: 2
How long will it be before AMD brings out 45 nm products?

AMD is so dead in the water...
By aveegun on 8/7/2007 9:13:50 PM , Rating: 5
The lack of competitive product release is not helping either...

Even though I prefer AMD over Intel, I'll have to jump over to intel side if this keeps up.

Where are you AMD? Quit slacking off! You guys need to enforce mandatory overtime over there. No weekends for you!

RE: AMD is so dead in the water...
By leidegre on 8/8/2007 1:01:39 AM , Rating: 3
I'm pretty sure that they'r allready doing all they possibly can. Luckly, I'm buying new hardware before the end of this year, or at least that is my plan. But it's going to be Intel/nVidia stuff if AMD doesn't deliver on thier promise anytime soon.

RE: AMD is so dead in the water...
By ScythedBlade on 8/8/2007 9:05:01 AM , Rating: 3
Wow, Intel really is sticking to its plans ... they're shoving the tick-tock slide and reminding everyone that they're true to their word ... XD ...

AMD, better get your game up ... this is what you wanted wasn't it? Poking Intel until it woke up to give you a real fight ... But I think you might have done it too soon and was a bit too cocky ....

By One43637 on 8/8/2007 5:32:58 PM , Rating: 2
Agreed. My current setup is an Opteron 175, but I'm starting to get that itch of upgrading again and the benchmarks on this chip are very nice indeed.

By techfuzz on 8/8/2007 10:01:41 AM , Rating: 2
Looks like I may be going back to Intel for my next computer unless AMD can make pigs fly.

By Believer on 8/8/2007 4:56:42 AM , Rating: 3
Am I assuming correctly that the reduced power consumption and lower thermal values are because of the new high-K gate dielectrics and metal gate electrodes introduced by Intel some time ago? As in, beside the obvious other beneficial factors such as the die shrink and lower core voltages.

Or are these things that Intel plan to implement in the Nehalem architecture as the following blog quote from DT vaguely hints at:
Multi-level Cache Confirmed for Nehalem

Intel's guidance for Nehalem, the next-generation 45nm successor to Penryn, claims the following:
- Leverages 4-issue Intel Core microarchitecture technology
- Simultaneous multi-threading (SMT)
- Multi-level shared cache architecture
- Performance-enhanced dynamic power management
- Fully unlocks Intel 45nm Hi-K silicon process benefits

So are we to expect even further drops in power consumption and temperature when they reduce electrical leakage even further or is this already implemented to full extent in Penryn? ('Cause I haven't seen any mentioning of it in any review so far)

By Anh Huynh on 8/8/2007 4:02:01 PM , Rating: 2
Nehalem benefits from a more mature 45nm process, not to mention experience and knowledge gained from working on Penryn.

In the past, Intel usually shrunk the die of the current architecture, work out all the kinks, before it debuts a chip with major changes. Each fab process typically has two generations of cores.

We saw that with Prescott (90nm) -> Smithfield (90nm) -> Presler (65nm) -> Conroe (65nm) -> Penryn (45nm) -> Nehalem (45nm)

By Dactyl on 8/9/2007 1:00:12 AM , Rating: 2
Yes, that way Intel only has to deal with 1 type of problem at a time:

1 - die shrinks cause new problems, because smaller process technologies introduce new things that can go wrong. The physics changes as you get smaller and smaller. The electromagnetic fields around the tiny wires linking the cells interfere with each other more and more (crosstalk delay, antenna effect, etc.)

via wikipedia, here is an Intel slideshow about this phenomenon:

Designs that were fine at 130nm did not work at 90nm. 90nm designs don't work at 65nm. 65nm designs don't work at 45nm. It didn't used to be this way (it used to be you could move down a process technology without redesigning your whole CPU) but at about 100nm and below (aka "deep submicron") things got a little crazy.

2 - moving to a new architecture means there is tremendous room for bugs, because there are literally hundreds of millions of transistors that need to interact correctly.

Even Intel doesn't want to debug a new architecture on a new process! They would never figure out where the bugs were.

By Treckin on 8/7/2007 5:48:46 PM , Rating: 1
All I can say is that Agena better be REALLY fucking fast, for DAMMIT's sake (pardon the Inq ref).

RE: Well
By deeznuts on 8/7/2007 6:33:00 PM , Rating: 2
Yup. Especially since it appears its faster clock for clock (even with non-SSE4 code) and will launch at high clock frequencies.

RE: Well
By mrdelldude on 8/8/2007 12:14:49 PM , Rating: 2
Not only with a higher clock, but also 50% more cache per core.

By smitty3268 on 8/7/2007 8:21:52 PM , Rating: 2
SSE4 is looking very impressive. The rest of the benchmarks are really about what I expected, though.

By Captain Orgazmo on 8/8/2007 2:31:50 AM , Rating: 2
I'd love to see the performance gain given to an SSE4 optimized H.264 encoder... Realtime x264 encoding anyone? :D

AMD Fans Dead?
By LTG on 8/8/2007 4:44:11 PM , Rating: 2
It's gotten so bad for AMD that for the first time ever 17 comments went by without anyone saying "we still have to see how barcelona does".

By the way it is August right? When will AMD at least give an exact date for when benchmarks will be released?

They are scared to death to give out press samples.

"A lot of people pay zero for the cellphone ... That's what it's worth." -- Apple Chief Operating Officer Timothy Cook
Related Articles

Latest By Anh Tuan Huynh
More NVIDIA SLI Chipsets Around the Corner
September 25, 2007, 7:40 AM
NVIDIA Prepares New AMD IGPs
September 25, 2007, 7:39 AM
NVIDIA Launches Intel IGP
September 25, 2007, 7:25 AM
Intel Shows Off 32nm Test Shuttle
September 18, 2007, 4:34 PM
Intel Sets Official "Penryn" Launch Date
September 18, 2007, 1:17 PM
VIA Launches EPIA SN With 1.8 GHz Processor
September 17, 2007, 4:00 PM
Freescale Licenses AMD Technologies
September 17, 2007, 3:43 PM
AMD Adds Triple-Core Processors to Roadmap
September 17, 2007, 2:45 PM
Lenovo Announces Solar Power Capable Desktop
September 13, 2007, 2:00 PM

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki