Web server
| |
All Disabled |
TB Enabled |
Gain, % |
HT Enabled |
Gain, % |
All Enabled |
Gain, % |
| PHP Calculator |
211 |
222 |
5% |
254 |
20% |
267 |
27% |
| PHPSpeed |
|
| Synthetic PHP |
3601 |
3790 |
5% |
3498 |
-3% |
3693 |
3% |
| Synthetic MySQL |
3156 |
3336 |
6% |
2984 |
-5% |
3204 |
2% |
| Synthetic Read/Write |
2339 |
2454 |
5% |
2248 |
-4% |
2413 |
3% |
| Real World PHP |
6360 |
6681 |
5% |
6176 |
-3% |
6513 |
2% |
| Real World PHP & MySQL |
3010 |
3151 |
5% |
2908 |
-3% |
3115 |
3% |
| Server |
2477 |
2589 |
5% |
2340 |
-6% |
2536 |
2% |
| Group Score |
126 |
132 |
5% |
125 |
-1% |
133 |
6% |
Significant performance gain from Hyper-Threading is demonstrated in the only truly SMP-optimized PHP Calculator. Hyper-Threading only slows down "single-processing" PHPSpeed -- this problem is likely to be inherited by this technology from its first incarnation in NetBurst architecture: automatic "halving" of some processor buffers, when it's enabled.
Archiving
| |
All Disabled |
TB Enabled |
Gain, % |
HT Enabled |
Gain, % |
All Enabled |
Gain, % |
| 7-Zip |
0:02:01 |
0:01:58 |
3% |
0:02:01 |
0% |
0:01:57 |
3% |
| WinRAR |
0:01:01 |
0:00:58 |
5% |
0:01:07 |
-9% |
0:01:04 |
-5% |
| Ultimate ZIP |
0:01:54 |
0:01:44 |
10% |
0:01:54 |
0% |
0:01:44 |
10% |
| Group Score |
123 |
130 |
6% |
119 |
-3% |
126 |
3% |
Highly optimistic about Turbo Boost on one hand and absolutely indifferent (with two instances of dislike) to Hyper-Threading -- that's an illustrative example of one peculiarity of modern archivers: even four cores are too many for them (note a significant performance gain, when Turbo Boost is enabled), to say nothing of eight cores (with reduced buffers at that).
Encoding
| |
All Disabled |
TB Enabled |
Gain, % |
HT Enabled |
Gain, % |
All Enabled |
Gain, % |
| FLAC |
0:01:05 |
0:01:00 |
8% |
0:01:05 |
0% |
0:01:00 |
8% |
| LAME |
0:01:33 |
0:01:29 |
4% |
0:01:33 |
0% |
0:01:29 |
4% |
| Musepack |
0:01:50 |
0:01:41 |
9% |
0:01:51 |
-1% |
0:01:41 |
9% |
| Vorbis |
0:03:14 |
0:03:05 |
5% |
0:03:15 |
-1% |
0:03:13 |
1% |
| Canopus ProCoder |
0:04:04 |
0:03:51 |
6% |
0:04:17 |
-5% |
0:04:07 |
-1% |
| DivX |
0:00:58 |
0:00:54 |
7% |
0:00:57 |
2% |
0:00:52 |
12% |
| x264 |
0:02:10 |
0:02:04 |
5% |
0:01:47 |
21% |
0:01:42 |
27% |
| XviD |
0:06:04 |
0:05:31 |
10% |
0:06:05 |
0% |
0:05:32 |
10% |
| Group Score |
124 |
133 |
7% |
127 |
2% |
135 |
8% |
Two audio codecs (one of them features SMP-optimizations) and one video codec (Canopus) were not very optimistic about virtualization of cores. However, SMP-optimized x264 repeats the record of Maya renderer (21% performance gain from HT). This fact is full of hope: this technology can really yield performance gains!
Games
| |
All Disabled |
TB Enabled |
Gain, % |
HT Enabled |
Gain, % |
All Enabled |
Gain, % |
| Call of Duty 4 |
105 |
105 |
0% |
105 |
0% |
105 |
0% |
| Company of Heroes |
57 |
57 |
0% |
57 |
0% |
57 |
0% |
| Call of Juarez |
47 |
50 |
6% |
47 |
0% |
47 |
0% |
| Crysis |
24.63 |
24.77 |
1% |
24.83 |
1% |
24.83 |
1% |
| S.T.A.L.K.E.R. |
122 |
122 |
0% |
121 |
-1% |
122 |
0% |
| Unreal Tournament 3 |
125 |
128 |
2% |
114 |
-9% |
119 |
-5% |
| World in Conflict |
59 |
61 |
3% |
59 |
0% |
60 |
2% |
| Group Score |
105 |
107 |
2% |
103 |
-1% |
104 |
0% |
Back in the first article about Core i7 920 we mentioned that the new architecture from Intel was not quite good at games yet. This article is another proof of this point. That's strange: judging by several tests, games can respond to higher memory bandwidth quite positively.
Non-professional photo processing
| |
All Disabled |
TB Enabled |
Gain, % |
HT Enabled |
Gain, % |
All Enabled |
Gain, % |
| ACDSee |
0:09:00 |
0:08:20 |
8% |
0:08:50 |
2% |
0:08:20 |
8% |
| IrfanView |
0:15:13 |
0:13:54 |
9% |
0:15:14 |
0% |
0:13:55 |
9% |
| Paint.NET |
20514 |
19430 |
6% |
16317 |
26% |
15626 |
31% |
| xat.com Image Optimizer |
0:29:45 |
0:27:31 |
8% |
0:29:46 |
0% |
0:27:34 |
8% |
| XnView |
0:15:23 |
0:14:02 |
10% |
0:15:21 |
0% |
0:14:06 |
9% |
| Group Score |
122 |
132 |
8% |
128 |
5% |
138 |
13% |
Pay attention to Paint.NET. This applications is based on Microsoft.NET, and it puts up the brightest performance in this test. Why? Probably owing to the platform specifics: it's no secret that .NET applications are not honest executable files in a way, they depend on an operating system. On one hand, it's their advantage: they use system resources to the maximum, we can see it in results. On the other hand, they are rather slow -- that's why they respond so well to various performance boosting technologies.
Conclusions
| |
All Disabled |
TB Enabled |
Gain, % |
HT Enabled |
Gain, % |
All Enabled |
Gain, % |
| PRO SCORE |
120 |
127 |
6% |
119 |
-1% |
127 |
6% |
| HOME SCORE |
119 |
125 |
6% |
119 |
1% |
126 |
6% |
| OVERALL SCORE |
119 |
126 |
6% |
119 |
0% |
126 |
6% |
Turbo Boost technology demonstrates impressive performance gains practically everywhere. On the one hand, it's not bad: "common overclocking" provides performance gains to all users without exception, no consequences to system stability (at least according to the manufacturer). On the other hand, overclockers lose their semi-legal ways to get higher performance for less money, as manufacturers take advantage of their favorite tricks. However, it was easy to predict the appearance of something like Turbo Boost: any popular semi-legal practice always attracts manufacturers who want to benefit even by such tricks.
Despite its new reincarnation, Hyper-Threading is still as controversial as in times of NetBurst. This technology yields zero performance gains according to our overall score (rounded), even though some tests demonstrate significant performance gains or drops. The reasons are lying on the surface: it's not easy to optimize software for virtual multiprocessing. What concerns Intel, this company abandoned this field, having rolled out truly multi-core processors -- and software developers immediately followed suit. To all appearances, Hyper-Threading is being revived now. However, developers already accumulated negative experience -- HT had been actively advertised at first, and then it was forgotten because of more popular market tendencies. So those developers, who have already mastered this technology in its first reincarnation, will hardly welcome the revival of HT.
Our today's analysis once again proved the main conclusion from our Intel Core i7 review: it's a "no-nonsense" processor. Candid performance drops of Core i7 caused by the reanimated Hyper-Threading are smoothed over by Turbo Boost in most cases. As a result, the overall impression from its performance in the nominal mode is moderately positive (few users want to know details, like our today's tests). Time will pass, and they may get used to Hyper-Threading and learn to program software that at least won't suffer from this technology, or even will benefit from it -- we already have positive examples. Intel seems to have reached perfection in broadcasting positive thinking attitude to potential users: it does not grab and haul them into its concept of radiant future (farewell, Rambus), it just pushes them mildly in the right direction.
One thing is clear: having analyzed its previous (mostly negative) experience in promoting the "progressive" NetBurst architecture, Intel learned its lesson well and does not risk promoting its architectural concepts in the might-makes-right way anymore. Now the company is more gentle, offering its users not only two birds in the bush tomorrow, but also a bird in the hand today. We've already appreciated Turbo Boost. Thank you. Hyper-Threading is still in the bush. We'll wait and see.
Write a comment below. No registration needed!
|
|
 |
|
|
|
