The Clarkdale Experiment: Mea Culpa
by Loyd Case on May 4, 2010 1:57 PM ESTSeveral months ago, I wrote about my little experiment with Clarkdale, where I built a small form factor system, based around a Core i5 661 CPU, an Asus H55 motherboard and a Radeon HD 5850 graphics card. That system also used a pricey, 250GB SSD, which was a little over the top for an otherwise modest system, but the idea was to make it small, quiet and low power.
Quite a few people critiqued the article, and most of the critiques revolved around the lack of performance data. After all, the general feeling went, how do we know this is really a good system? It’s pricey, to be sure, but we also have no way of judging performance.
As it happens, I have another small form factor system in the basement lab, which happens to be running a Core i5 750. Ignoring hard drive performance for the moment, all I really needed to do was swap out the graphics card, since the Lynnfield system was running an older Radon HD 4870. So I dropped in a Radeon HD 5850 and took both systems for a spin.
Price versus Performance
There’s a time and place for integrated graphics – but not for PC gaming. Dropping in a Radeon HD 5850 likely means the system will be used for PC gaming – which was my intent all along. The Core i5 661 is priced nearly identically with the Core i5 750. Beyond price, the differences are pretty substantial:
· Clock speed: 3.33GHz (Clarkdale) versus 2.66GHz (Lynnfield)
· Maximum Turbo Frequency: 3.6GHz (Clarkdale) versus 3.2GHz (Lynnfield)
· Dual core with hyper-threading (Clarkdale) versus quad core without hyper-threading (Lynnfield)
· 4MB shared L3 cache (Clarkdale) versus 8MB shared L3 cache (Lynnfield)
· TDP: 87W versus 95W (Lynnfield)
So Clarkdale has a 25% raw clock speed advantage and a 12.5% maximum turbo boost performance. Lynnfield has the edge in cache – while both have a shared L3 cache, the sheer cache size gives Lynnfield an edge over Clarkdale in cache sensitive apps. Lynnfield also has four actual, physical cores, rather than two physical plus two SMT (virtual) cores. Interestingly, Clardale only has a modest TDP advantage over Lynnfield.
With these thoughts in mind, I ran a number of game benchmarks, plus a few other performance tests. Given the clock speed disparity versus cache size and number of cores, I didn’t expect big performance disparities. As it turns out, I was in for a few surprises.
System Configuration
These systems weren’t identically configured, but were similar.
|
Component |
Clarkdale System |
Lynnfield System |
|
CPU |
Core i5 661 @ 3.33GHZ base |
Core i7 750 @ 2.67GHz base |
|
Motherboard |
Asus P7H55-M EVO |
Gigabyte GA-P55M-UD4 |
|
Memory |
2 x 2GB OCZ DDR3-1600 @ 1333 |
2 x 2GB OCZ DDR3-1600 @ 1333 |
|
GPU |
XFX Radeon HD 5850 |
XFX Radeon HD 5850 |
|
Hard Drive |
OCZ 250GB SSD |
WD Caviar Blue 640GB |
|
Optical Drive |
Asus BD-ROM / DVD+/-RW |
Lite-On DVD+/-RW |
|
PSU |
Cooler Master 500W |
Cooler Master 500W |
The key differences in components were motherboards (Asus H55 versus Gigabyte P55) and hard drive (an SSD versus a standard rotating media drive.) None of the tests I ran were particularly storage intensive, and any power advantage due to drive differences were pretty minimal.
Facebook
Twitter
RSS
45 Comments
View All Comments
GeorgeH - Tuesday, May 4, 2010 - link
It’s nice to see you test these systems, but the numbers and analysis you present are completely useless; here are two reasons why:Unigine:
From the article: “When we keep the resolution and detil levels low (tessellation is off in the low resolution test), the CPU differences are noticeable. Once we dial up the graphics pain, though, the difference is negligible.” At both resolutions the difference is about 3%; how can 3% be both noticeable and negligible at the same time?
S.T.A.L.K.E.R.:
How repeatable is 195.15 FPS – can you really report 5 significant figures here? From the article: “The numbers vary slightly, but the pattern doesn’t change.” Varying numbers are incredibly important to establishing the relevance of a pattern; for all I know 195.15 is a high outlier and 143.1 is a low outlier. You might be reporting the mean of multiple runs with significant figures appropriate to their standard deviation, but the overall impression of the article is that you ran the benches once and reported whatever number got spit out, with verification that it indeed should have been “bigger” or “smaller” than the number spit out by the other system by running the benchmarks again.
Combining those two problems, I’m forced to conclude that any numbers and analysis you report here are worthless. That’s unfortunate because they probably aren’t, it’s just the style of the report and failure to indicate more clearly what your testing methodology is that unfairly gives that impression.
tno - Tuesday, May 4, 2010 - link
I agree, completely, and this brings me to the biggest worry I've had over the content on AT for the last year or so. When it was Gary and Anand, there were some hiccups, either with style or content in many of the early articles. But before too long they both got very good at what they do and the result was for a while though throughput wasn't high it was of high quality, consistently. As the staff has grown, however, some of the new blood is really good at analysis and great create models for comparing various pieces of technology but don't write very well. Other staff members, Loyd inclusive, seem to have a great grasp of style and even their analysis sounds good, yet clearly here there are some content errors that lessen the impact of the piece.Anand, you have become a master at providing excellent content with impeccable style. Your new writers are all capable, but I think it might be time for a little writing and statistical analysis boot camp.
tno
dfonseca - Tuesday, May 4, 2010 - link
Seconded.This article by Loyd is a great example of this problem. It is well-worded and pleasant to read, but entirely inconsistent with other AT articles - with the exception of those written by Loyd.
Articles with this kind of content (hardware configuration benchmarking) abound in AT, and they roughly follow the same pattern with regards to what data is gathered, how it's presented, chart formatting, etc. Reading AT would be a better experience if the articles followed on that formula, or improved on it (with a strong focus on continuity).
futrtrubl - Tuesday, May 4, 2010 - link
Speaking of charts I think he grouped his bars badly. He's comparing systems so he should group by test/setting not by CPU so it's easier to compare.Also, he is inconsistent in referring to the chips, sometimes by model number and sometimes by codename. I don't know which is which and while I can look it up I shouldn't have to translate it in my head every time they are mentioned.
anactoraaron - Tuesday, May 4, 2010 - link
But the biggest thing with these numbers is the HDD difference. Having a 250gb SSD vs anything other than another SSD will throw these numbers off... Putting in a SSD for the i5-750 will likely add ~10% more to all numbers across the board for it. I just can't help but think how much better the i5 750 would have been with a SSD. Anand himself no longer does any benchmarking amongst cpu's now without a SSD since it takes away variables a platter HDD may cause.I second the boot camp idea.
alphacheez - Tuesday, May 4, 2010 - link
This shows that the i5-750 seems to have more room to grow as demands will increase in the future.The 750 should be able to keep up with the next generation of games while the 661 might run out of steam and leave you CPU-bound.
I think the colors on the Cinebench and video encoding graphs are/were backwards. The higher Cinebench score should correspond with the multi-CPU test and I'd expect the 1080p wmv to h.264 encode to take longer than the avi to mp4 (iPod) encode.
I'd be interested to hear others experience with Clarkdale-based HTPCs as far as video playback, encoding, power usage, and noise.
The systems examined in this write-up are pretty high-powered compared to what might be in a typical HTPC. A Radeon 5770 should be enough to power games at 1920x1080 (HDTV) resolutions and really put the kibosh on power usage.
jasperjones - Wednesday, May 5, 2010 - link
agreed, the legend in cinebench seems incorrect.Jaguar36 - Tuesday, May 4, 2010 - link
Could you check the power usage without the Radeon in the system? Since I leave my PC on 24/7 I really want to know how much power the card uses at idle compared to the IG.justinegg - Tuesday, May 4, 2010 - link
How much of a difference could PCI-Express and other chipset bottlenecks affected these gaming benchmarks?Would the i5 -750's numbers be closer to the 661's if it were in the H55 board?
jonup - Tuesday, May 4, 2010 - link
I am not an expert, but per my discussions with several reviewers H55 performance is similar if not identical to P55.As for the performance difference between 750 and 661, it could be due to the memory latencies on top of the L3 size.