A little over a year ago at the Intel Developers forum in San Francisco, Intel allowed a small group of members of the technology press to run benchmarks on a pre-configured Conroe-based system, well before processors based on the Conroe core had begun to ship. This was a new strategy for Intel, but a welcomed one in our opinion as it gave enthusiasts a chance to glean some real information about an upcoming product well before it actually hit store shelves.
Intel continued to employ this strategy in the weeks leading up to the launch of the Kentsfield core. Once again, in a closed door session in a briefing room at IDF in California, Intel gave is the chance to run some benchmarks on a then unreleased quad-core Kentsfield-based system. But at this particular juncture, IDF isn't taking place in San Francisco. At the moment, the Intel Developers Forum is underway in Beijing. Intel, however, still has plenty of Penryn-related details coming out at the show, and we've got some of it for you here today.
In January, we talked to you about Intel's 45nm high-k and metal gate transistors and posted some information regarding Penryn, the first processor core from Intel that will be manufactured using their 45nm process. Then a few weeks ago, we followed up with some more info regarding Penryn and snuck in some details related to its successor, Nehalem. And today, direct from Intel, we have some performance data comparing a couple of flavors of Penryn - Wolfdale and Yorkfield to be exact - to Intel's current flagship quad-core Core 2 Extreme QX6800 processor.
The test system configurations and specifications are detailed in the table below. And in case you're wondering, that picture up there is of Intel's Justin Ratner holding an entire wafer of 45nm Penryn dies. Samples are being manufactured already folks...
Test System Configuration
Pre-production 45nm Hi-k Intel®
core processor (6MB cache, 3.33GHz, 1333MHz FSB)
Pre-production 45nm Hi-k Intel®
core processor (12MB cache,3.33GHz, 1333MHz FSB)
Intel® Core™2 Extreme Processor QX6800 (8MB Cache, 2.93GHz, 1066MHz FSB)
Pre-Production Intel® BadAxe 2 Desktop Board D975XBX2 (Rev 505)
Intel® BadAxe 2 Desktop Board D975XBX2 (Rev 505)
Intel® Chipset INF 184.108.40.2060
ASUS* nVidia* Geforce* 8800GTX
nVidia* 100.65 graphics driver (http://www.nvidia.com/object/winvista_x86_100.65.html)
2x1GB Corsair* TWIN2X2048-8500C5 set to DDR2 800 5-5-5-15 in BIOS
Seagate* Barracuda* 320GB 7200.10 7200RPM
Intel® Matrix Storage Driver 220.127.116.112
Sigmatel* Audio Driver 6.10.5275
Sony* DVD-ROM DDU1615
Antec* TruePower Trio* 650
Microsoft* Windows* Vista Ultimate 32bit
Benchmark (or) Software Application
Test Results on System A
Test Results on System B
Test Results on System C
Quad-Core 2.93GHz QX6800
3DMark*06 V1.1.0 Pro - CPU (score)
3DMark*06 V1.1.0 Pro - Overall (score)
Mainconcept* H.264 Encoder (seconds)
Cinebench* R9.5 (CPU Benchmark)
Cinebench* R10 Beta (CPU Benchmark)
Half-Life* 2 Lost Coast Build 2707 (frames per second)
Divx* 6.6 Alpha with VirtualDub 1.7.1 (seconds)
There are a couple of things to note in the comparisons above. First, System A is equipped with an upcoming 45nm dual-core CPU; Systems B and C are equipped with quad-core processors. And Systems A and B are clocked at 3.33GHz, while System C which is based on the current flagship QX6800 is clocks in at 2.93GHz. That's a 400MHz clock speed disparity, which equated to roughly a 13.5% difference.
Although the upcoming dual-core processor has clock speed advantage that gives it an edge in the HL2 gaming tests, it can't quite keep up with the current QX6800 in most of the other multithreaded tests. We say most, because the SSE4 instructions incorporated into Penryn give these upcoming processors a huge boost in the Divx benchmark, where the dual-core CPU actually outpaced the QX6800 by 16 seconds.
Also note that although the 13.7% edge in clock speed accounts for some of the new quad-core chip's advantage over the QX6800 in all of the tests, it is actually between 18% and 111% faster than Intel's current CPU (disregarding the overall 3DMark06 score, which is determined more by GPU than CPU performance).
This information shows that early Penryn-based CPU samples are already running very well, and even in this early stage are showing major performance improvements over current products. We can't wait to get our hands on one and test it out for ourselves.