The Phenom II X4 810 & X3 720: AMD Gets DDR3 But Doesn't Need It
by Anand Lal Shimpi on February 9, 2009 12:00 AM EST- Posted in
- CPUs
Adobe Photoshop CS4 Performance
To measure performance under Photoshop CS4 we turn to the Retouch Artists’ Speed Test. The test does basic photo editing; there are a couple of color space conversions, many layer creations, color curve adjustment, image and canvas size adjustment, unsharp mask, and finally a gaussian blur performed on the entire image.
The whole process is timed and thanks to the use of Intel's X25-M SSD as our test bed hard drive, performance is far more predictable than back when we used to test on mechanical disks.
Time is reported in seconds and the lower numbers mean better performance. The test is multithreaded and can hit all four cores in a quad-core machine.
While our Photoshop CS4 benchmark clearly prefers Intel's CPUs, the results aren't so clean cut. The entire Phenom II 800 and 900 lineup falls short of their competitors, but the Phenom II X3 720 can equal its rival: the Core 2 Duo E7500. It's slower than the more expensive E8400, but it's at least competitive.
This is where AMD has some potential to do well. As a result of its larger die, the 720 has one more core than Intel's svelte Core 2 Duos. In well threaded applications, the X3 could come out ahead.
DivX 8.5.3 with Xmpeg 5.0.3
Our DivX test is the same DivX / XMpeg 5.03 test we've run for the past few years now, the 1080p source file is encoded using the unconstrained DivX profile, quality/performance is set balanced at 5 and enhanced multithreading is enabled:
Both the Phenom II 900 and 800 series parts are competitive here. The Phenom II X3 720 actually beats out both competitors here, the E7500 and E8400 are close but no faster than the 720. Here's a benefit of that extra core at work.
x264 HD Video Encoding Performance
Graysky's x264 HD test uses the publicly available x264 codec (open source alternative to H.264) to encode a 4Mbps 720p MPEG-2 source. The focus here is on quality rather than speed, thus the benchmark uses a 2-pass encode and reports the average frame rate in each pass.
The first pass of the x264 benchmark is actually an analysis pass, the real encoding is done in the second pass. There are apparently quite a few unaligned cache accesses in this test which severely penalize the Core 2 processors; Core i7 is unaffected. Because of the unaligned cache access performance penalty, AMD owns the price/performance comparison here.
The actual encode workload is much more competitive. The Phenom II 800 and 900 series chips cluster around each other, remaining competitive but not really any faster than the equivalently priced Core 2 Quads. The Phenom II X3 720 continues to be faster than its competitor thanks to its core count advantage.
Windows Media Encoder 9 x64 Advanced Profile
In order to be codec agnostic we've got a Windows Media Encoder benchmark looking at the same sort of thing we've been doing in the DivX and x264 tests, but using WME instead.
Here we've got, once again, a competitive AMD but there's something very interesting going on with the triple core 720. WME9 doesn't make use of the third core as it is hard coded to use powers of 2 for the number of processor cores. There are some applications that will exhibit this sort of behavior, although in our testing the numbers are limited. With its third core out of the picture, the 720 is slower than Intel's Core 2 Duo E7500.
77 Comments
View All Comments
7Enigma - Monday, February 9, 2009 - link
Ditto the power consumption at OC'd levels. I always get annoyed to see these fantastic OC results but then fail to see whether we've doubled the power consumption. It would certainly allow us to see a potential benefit if one or the other uses significantly less power under OC load conditions.Gary Key - Monday, February 9, 2009 - link
We will have power consumption and temps up tomorrow for the OC results along with a few benchmarks. It was difficult to get stable (true) volt readings with the X4 810 installed, so I spent my weekend with the DMM on the boards.Kaleid - Monday, February 9, 2009 - link
Great! Possible to also add difference with non-overclocked and overclocked power consumption at idle?7Enigma - Monday, February 9, 2009 - link
At the wall (total system) or just for the CPU? Do you mean the total system power was fluctuating with the 810? That seems really really wierd.Gary Key - Monday, February 9, 2009 - link
The power was fluctuating on the board at the Core VID side. I should have a BIOS spin tonight to fix it. The 720 was perfect but the 810 had a few problems that have been identified now.7Enigma - Monday, February 9, 2009 - link
*weird, please give us an edit function.OCedHrt - Monday, February 9, 2009 - link
Interestingly, the 810 vs 910, there is no real performance difference outside the margin of error. In some cases, the 810 is faster and in some, the 910 takes the lead. Something I noticed though is that the 810 is faster than the 910 at more times, and faster by a larger amount (~3% when it is faster) as well. Seems like the reduced cache does not actually hamper performance.Moorbo - Friday, April 24, 2009 - link
For most applications it seems you're correct that the smaller cache makes little difference. However if you look at the gaming performance the 2MB/core L3 cache of the 720 and 710 clearly makes a big difference despite their slower clock and lack of an additional core. What would the numbers look like with only two cores and 3MB/core and a higher clock?johnsonx - Monday, February 9, 2009 - link
indeed, the 710 is also often faster than the 910, but usually slower than the 810.I'm a bit surprised we the readers have to thumb through all the benchmark charts to see the 710/810/910 comparison. 3 Phenom II's at the same speed, one 'standard', one with less cache and one with a missing core; that is something AT should have dedicated a page to.
stmok - Monday, February 9, 2009 - link
Yeah, I noticed that as well. It looks like 4MB L3 cache is sufficient with Phenom II. (Any less, it'll start hurting...Any more, you're just wasting silicon space).