iXBT Labs - Computer Hardware in Detail






How CPU Features Affect CPU Performance, Part 4

<< Previous page

     Next page >>

As promised, this short article will be devoted to the analysis of probably one of the most interesting technologies used in Intel Core i7 processors -- Hyper-Threading. We were interested in a simple issue: when there are four physical cores, each split into two logical cores by HT, totalling in 8 processors per system, even an operating system may get lost in its cores. Especially as the issue of interaction between Windows Vista x64 Ultimate and eight processors hasn't been studied well enough. It's reasonable to assume that Microsoft Vista hasn't been prepared for that stuff. That's why the previous article provided some food for thought, but it didn't satisfy our curiosity about HT completely: we had to cover too much -- physical cores, logical cores, technologies.

How can we create the most favorable conditions for Hyper-Threading? We decided to use the simplest method. Indeed, two cores are not eight, most programs handle two cores very well. So when Hyper-Threading emulates only two cores instead of eight, pros and cons of this technology should reveal themselves in the most pronounced way. So this time we shall not try to reproduce a realistic situation, we'll do on the contrary: we'll provide perfect conditions for Hyper-Threading to reveal its true worth. That's what we've done: first of all we disabled in BIOS all cores but one in the Core i7, we disabled Turbo Boost (to fix the operating frequency). After that we measured system performance without Hyper-Threading (in this case the operating system detected our processor as a single-core model) and with Hyper-Threading enabled (in this case the operating system detected two cores, but it worked with one core anyway). Then we enabled two cores in BIOS, disabled Turbo Boost and Hyper-Threading, and reran our tests. Thus, we got an opportunity to analyze how useful Hyper-Threading is in a practically perfect situation from two point of view: comparison with the honest single-core system (without HT) will show us how useful Hyper-Threading is. And comparison of a HT system with the honest dual-core system will demonstrate how worse virtual multiple cores are versus real multiple cores.

Besides, in this case we test not only Hyper-Threading, but also software to some degree: is it ready to use virtual cores? Can it utilize their features? At at least not slow down?


Testbed configuration

  • CPU: Intel Core i7 950
  • Cooler: ASUS Triton 81
  • Motherboard: ASUS P6T SE (Intel X58)
  • Memory: 3 x 2GB Corsair DDR3-1800 in DDR3-1600 mode
  • Graphics card: Palit GeForce GTX 275
  • PSU: Cooler Master Real Power M1000

We traditionally publish a link to a Microsoft Excel spreadsheet for curious readers, which contains all test results in the most detailed form. Besides, it includes an additional tab "Compare" to facilitate their analysis. Just like the tables in the article, it compares the situations percentagewise. Tables are colored in the traditional way: bright blue background marks outstanding positive results (in this case we marked performance gains above 10%), red background signals problems: performance drops where we should have performance gains (or at least no changes). Performance gain percent is always calculated relative to the leftmost system (single core, no HT).

We'll publish only short comments to the diagrams, because in the previous article we wrote much about the effect of Hyper-Threading on performance. So we are going to pay attention only to interesting results that differ from the previous article.

3D visualization

  1 core, HT OFF 1 core, HT ON 2 core, HT OFF
3ds max ↑* 16.36 16.93 3% 16.45 1%
Lightwave ↓ 14.52 13.7 6% 12.85 13%
Maya ↑ 3.59 4.17 16% 4.08 14%
SolidWorks ↓ 54.39 53.81 1% 52.64 3%
Pro/ENGINEER ↓ 1065 1056 1% 998 7%
UGS NX ↑ 2.99 3.14 5% 2.84 -5%
Group Score ↑ 132 139 5% 139 5%

* The up arrow (↑) marks tests, where the highest results are the best, the down arrow (↓) marks tests, where the best results are the lowest.

Maya and Lightwave were champions in the previous article (but in the other order), and they shine here as well. SolidWorks responded to HT by slowing down on a quad-core system. But in this case it's more disposed to this technology. By the way, we'll see similar situations many times.

3D rendering

  1 core, HT OFF 1 core, HT ON 2 core, HT OFF
3ds max ↑ 4.71 6.02 28% 9.17 95%
Lightwave ↓ 332.32 275.89 20% 182.23 82%
Maya ↑ 10:35 08:25 26% 05:22 97%
Group Score ↑ 40 50 25% 77 93%

The average performance gain from HT on a quad-core system in this group amounted to 21%. In this case (single core) it's 25%. It's formally bigger, but frankly speaking we expected higher contrast.

Scientific and engineering analysis

  1 core, HT OFF 1 core, HT ON 2 core, HT OFF
Maya ↑ 7.1 7.44 5% 8.89 25%
SolidWorks ↓ 40.06 40.2 0% 35.39 13%
Pro/ENGINEER ↓ 1632 1622 1% 1526 7%
UGS NX ↑ 5.47 5.66 3% 5.47 0%
MAPLE ↑ 0.2185 0.2193 0% 0.2194 0%
Mathematica ↑ 2.5295 2.5555 1% 2.9533 17%
MATLAB ↓ 0.071984 0.061414 17% 0.049676 45%
Group Score ↑ 117 121 3% 134 15%

There is a comic situation in this group of scientific and engineering applications: three programs out of seven (almost half) rejected Hyper-Threading in our previous article -- they dropped performance when HT was enabled. We can see nothing of the sort here, on the contrary: performance drop in MATLAB turned into a noticeable gain.

Raster graphics

  1 core, HT OFF 1 core, HT ON 2 core, HT OFF
ACDSee ↓ 06:53 05:34 24% 04:42 46%
Paint.NET ↓ 01:06 00:54 22% 00:34 94%
PaintShop Pro ↓ 09:20 09:14 1% 09:05 3%
PhotoImpact ↓ 06:53 06:29 6% 06:20 9%
Photoshop ↓ 10:49 09:48 10% 07:13 50%
Group Score ↑ 87 97 11% 118 36%

It's a similar situation again: most packages responded negatively to HT last time (4 out of 5). This situation on a single-core system is cardinally different: three programs out of five demonstrate performance gains.

Data compression

  1 core, HT OFF 1 core, HT ON 2 core, HT OFF
7-Zip ↓ 05:01 04:16 18% 03:49 31%
WinRAR ↓ 01:42 01:29 15% 01:17 32%
Group Score ↑ 105 122 16% 139 32%

Performance gains are much higher here: in general, 16% versus 3% in the previous article. However, it's easy to explain: archivers can use two cores effectively, so we could expect advantage of the HT system over an honest single-core computer.

Write a comment below. No registration needed!

Next page >>

Article navigation:

Page 1: Introduction, tests, part 1

Page 2: Tests, part 2, conclusions

blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.