Penryn Performance Preview

We have your first glimpse of Intel’s next-generation processor, Penryn. Intel claims it does pretty well in benchmarks. We agree that it’s more than just a die shrink.

The past year has been tremendous for the hardware world. Some of this is due to the new graphics subsystems that have come about, and some of it is due to new types of memory and advances in hard drive technology. However, the main reason the hardware world has been thriving for the last year is Intel’s Core 2 Duo. There are no two ways to describe the product: it single-handedly changed the way we look at processors. Recent developments bring news that not only is Intel quite happy with the success of the Core 2 family, they do not plan to sit back and relax as they have in the past. Intel is actively developing new and improved technologies to push the limits of desktop processing further than we could have imagined just a short year and a half ago. With Intel in full stride, the 2007 Intel Developer Forum (IDF) in Beijing has taken on a far more interesting plot of real estate in the trade show world.

A few weeks ago we brought you news of Intel’s new “Penryn” processor and the success they have been enjoying with it. Intel’s boss Paul Otellini did not pull any punches when he expressed his delight at seeing the early Penryn cores flawlessly running four operating systems. While Penryn is basically a die shrink of the current Kentsfield and Conroe cores, some added features have tossed in to supposedly boost performance significantly. Despite the tremendous success it has recently endured, Intel is no stranger to making claims that turn out lackluster when realized. For instance, we recall the Pentium 4 was supposed to hit 5 GHz a couple of years ago. As far as we know, that never happened. So with this claimed performance increase with Penryn, objective observers have reason to be skeptical. However, if there is one thing we have learned from Intel is the past year, it is that there are some pretty smart dudes over there. To combat any skepticism, Intel is bringing something new to the table at IDF: Cold hard proof.

We have received benchmark scores from Beijing that directly compare two new Penryn processors to Intel’s current second in command, the Core 2 Extreme QX6700. Enjoy.

Test Setup and Features
Hardware Configuration

Power Supply: Antec TruePower Trio 650
Motherboard: Intel Desktop Board D975XBX2 (Bad Axe 2) Rev. 505
Processor: Intel Core 2 Extreme QX6700 (2667 MHz) | 2x Pre-Production Penryn Processors (See below for details)
Hard Drive: Seagate Barracuda 320GB 7200.10 7200RPM 16MB Cache (Perpendicular Recording)
Video: ASUS NVIDIA GeForce 8800GTX
Memory: 2048MB 2×1024MB Corsair* TWIN2X2048-8500C5 @ DDR2 800 5-5-5-15
Optical Drive: Sony DVD-ROM DDU1615

Software Configuration

Motherboard BIOS: Pre-Production BIOS
Operating System: Microsoft Windows Vista Ultimate 32-bit
Video Driver: NVIDIA ForceWare Version 100.65 WHQL Certified (February 20 release)

Processors Used

We went over the new features to be included with the Penryn package in our “Intel’s Up and Coming: Penryn and Nehalem” article back in March. To briefly summarize what we learned about the new Penryn design:

A superior microarchitecture featuring many enhancements over the base Core architecture
Higher performance at current levels and even greater frequencies overall
SSE4 instructions for better gaming and graphics performance
Increased cache’s and quicker bus rates, up to 6MB and 12MB
Increased number division capabilities, doubling previous divider speed with Radix 16 45nm High-K gate technology which will allow for more transistors, a 20% increase in transistor switching which allows for more instructions per clock.

{mospagebreak heading=Introduction&title=3DMark06}

Tests
Unfortunately we did not get to pick and choose the tests to run on these processors. We didn’t even run the tests actually. All tests were run internally by Intel themselves. As a result, all numbers should be taken with a grain of salt. However, we wouldn’t expect Intel to exaggerate anything, since they really don’t have any reason to. Even if Penryn did not offer a sizable performance increase, it’s current lineup far surpasses all competition. That said, Intel chose to run several very good tests that test either processor performance or overall system performance, both of which are certainly good things to test!

3DMark06

In addition to the standard 3DMark score that gamers and enthusiasts have come to hold as the holy grail of system performance benchmarks, the suite from Futuremark also produces component specific scores. One of these scores is for the processor.

While the dual-core Penryn processor comes up short of the QX6700 from Intel, its quad-core sibling does quite well against the gold standard. When comparing the score of the quad-core Penryn against that of the QX6700, it appears that the difference tells two-sided story. This performance difference cannot quite be explained by the simple speed bump from 2.67 to 3.33GHz. We presume the SSE4 instructions capability of the Penryn processors, which aid in multimedia operations, play a significant roll here.

The differences evident from the overall test seem to back up the theory that the SSE4 capabilities of the Penryn chip offer a substantial performance boost. However, when testing 3DMark06 with a video card as capable as the GeForce 8800GTX, the final performance numbers are always going to be high and usually fairly consistent. Because of this, we don’t think 3DMark06 is the greatest test for comparing two micro-architecture generations, due to its heavy dependence on the graphics subsystem.
{mospagebreak title=Encoding}

Encoding

One thing we typically do not include in our processor reviews is video encoding. We have been searching for a really consistent means by which to test the encoding capabilities of processors for a while now, with some unimpressive results. However, we very recently came across a fairly good vehicle for testing H.264 encoding times, and we are pleased to see that Intel used the very one we had been considering. The Mainconcept encoder platform used here paints a good portrait of processor performance.

The numbers to pay attention to here are those of the quad-core chips. Because video encoding on this platform is multi-threaded, the dual-core processor is pretty much guaranteed to fall behind its quad-core counterparts. The performance difference here, though slightly more than what would be expected of a simple speed-bump, is nothing out of the ordinary.

Cinebench

We typically stay away from using the Cinebench test at GotFrag Hardware only because we feel we have a more competent sampling of synthetic benchmarks in our lineup already. However, Cinebench is a very good test of overall system performance and can be useful in illustrating differences in processor performance.

Here we see fairly dramatic performance increases over the QX6700 with the Penryn quad-core chip. Since Cinebench is a multimedia test that involves rendering three dimensional scenes, it is reasonable to suggest that the SSE4 capabilities of the Penryn processors play a huge part in the performance differences we see here. We also suspect that the increased cache sizes on the Penryn cores have something to do with, as calling textures from disk would seem to be a common occurrence when running Cinebench.

DivX with VirtualDub

Intel was pretty vague about the actual procedures they used when running these tests. Because of this, we have no idea what Intel did within VirtualDub to yield their results. We suspect there was some sort of post-processing or encoding to DivX format that went on, but this is just speculation. In any case, here are the results.

Here we see the most dramatic results of them all. Performance of even the dual-core Penryn chip is seemingly light-years ahead of the QX6700 in this test. Taking into consideration that this test was not run differently between the three processors (assumption, but we think Intel is good for it) and that there is a relatively small (4 seconds) performance difference between the two Penryn chips, it would seem that the SSE4 instructions really boosted performance exponentially here. The wider dynamic execution courtesy of the Radix-16 divider probably plays a fairly large role here as well.
{mospagebreak title=HL2, Conclusion}

Half-Life 2: Lost Coast

Last but definitely not least in Intel’s benchmark suite is VALVe’s venerable old HDR test Half-Life 2: Lost coast. This is the only real-world gaming benchmark that Intel conducted, or at least chose to publish, on the Penryn processors. While all previous tests do a good job at providing us with some idea of the synthetic application performance these processors can provide, gaming performance is the number we are most concerned with.

Here we see some fairly impressive yet somewhat suspicious results. There is a 57 FPS difference between the quad-core Penryn and the QX6700, which seems fair considering the extra 667MHz and the SSE4 capability (aids significantly in multimedia and gaming). However, when we ran our tests with the QX6700 back in November of last year, we found that there would always be an advantage for the quad-core processor over the similarly clocked dual-core one. We considered this to be the result of being able to offload background calculations to the other cores while the cores in use by the game chugged along just the same. With this in mind, the identical numbers between the quad-core and dual-core chips seems kind of fishy - but we did not run the tests so we cannot speculate as to the conditions under which they were run.

Conclusion
All of these tests seem to point to the fact that Intel is not lying about the success they have been having with Penryn. Although much of the evident performance increases can be explained through the 667MHz clock speed bump, some of the differences are greater than an increased speed can accommodate. The new features of on the Penryn core, namely SSE4, the Radix-16 Divider, 1333MHz FSB, and a larger L2 cache clearly boost performance of the new chip substantially.

Although we like to see genuine performance numbers coming out of Intel and IDF, we have to hold our applause on announcements like these because in the end, the numbers are indeed coming straight from Intel. While the engineers at the world’s largest chip maker are having a good time testing out their new design, reviewers are patiently waiting for their own turn with Penryn. No matter how legitimate Intel’s numbers may be, it is not until hardware review publications get a hold of the chip that the true performance benefits of Penryn will be understood.

Popularity: 4% [?]

Support FPSLabs! Share and Enjoy: