Today we shall focus on Intel Core i7 processors, in particular those outperforming Core i7-880. Firstly, we need to retest them with our new test method. Secondly, the new LGA2011 platform is to be announced very soon, and we'll need these test results for further comparison. LGA2011, like the preceding LGA1567, is designed for high-performance multiprocessor systems, and it will also replace the extreme LGA1366 platform that has been out there for nearly three years.
Testbeds
| CPU |
Core i7-860 |
Core i7-880 |
Core i7-2600 |
| Core |
Lynnfield |
Lynnfield |
Sandy Bridge QC |
| Process technology |
45 nm |
45 nm |
32 nm |
| Core clock rate (std/max) |
2.8/3.46 GHz |
3.06/3.73 GHz |
3.4/3.8 GHz |
| Initial multiplier |
21 |
23 |
34 |
| Turbo Boost scheme |
5-4-1-1 |
5-4-2-2 |
4-3-2-1 |
| Cores/threads |
4/8 |
4/8 |
4/8 |
| L1 cache, I/D |
32/32 KB |
32/32 KB |
32/32 KB |
| L2 cache |
4x256 KB |
4x256 KB |
4x256 KB |
| L3 cache |
8 MB |
8 MB |
8 MB |
| Uncore clock rate |
2.4 GHz |
2.4 GHz |
3.4 GHz |
| RAM |
2xDDR3-1333 |
| Graphics core |
- |
- |
GMA HD 2000 |
| Socket |
LGA1156 |
LGA1156 |
LGA1155 |
| TDP |
95 W |
95 W |
95 W |
| CPU |
Core i7-920 |
Core i7-970 |
Core i7-990X |
| Core |
Bloomfield |
Gulftown |
Gulftown |
| Process technology |
45 nm |
32 nm |
32 nm |
| Core clock rate (std/max) |
2.66/2.93 GHz |
3.2/3.47 GHz |
3.47/3.73 GHz |
| Initial multiplier |
20 |
24 |
26 |
| Turbo Boost scheme |
2-1-1-1 |
2-1-1-1-1-1 |
2-1-1-1-1-1 |
| Cores/threads |
4/8 |
6/12 |
6/12 |
| L1 cache, I/D |
32/32 KB |
32/32 KB |
32/32 KB |
| L2 cache |
4x256 KB |
6x256 KB |
6x256 KB |
| L3 cache |
8 MB |
12 MB |
12 MB |
| Uncore clock rate |
2.13 GHz |
2.13 GHz |
2.66 GHz |
| RAM |
3xDDR3-1066 |
| Graphics core |
- |
- |
- |
| Socket |
LGA1366 |
LGA1366 |
LGA1366 |
| TDP |
130 W |
130 W |
130 W |
| |
Motherboard |
RAM |
| LGA1155 |
Biostar TH67XE (H67) |
Corsair Vengeance CMZ8GX3M2A1600C9B (2x1333; 9-9-9-24) |
| LGA1156 |
ASUS P7H55-M Pro (H55) |
Corsair Vengeance CMZ8GX3M2A1600C9B (2x1333; 9-9-9-24) |
| LGA1366 |
Intel DX58SO2 (X58) |
12 GB; 3x1333; 9-9-9-24 / 3x1066; 8-8-8-19 (9x0 / 990X) |
We usually have 8 GB of RAM in our testbeds, but this time we made an exception for LGA1366 as it's the only platform with a triple-channel memory controller. Hence 12 GB. The memory clock rates differ because regular and extreme LGA1366 processors have different uncore clock rates.
Tests
Our new test method is briefly described here. The scores on diagrams are relative to that of our reference testbed that always scores 100 points. It's based on the AMD Athlon II X4 620 CPU, 8GB of RAM and Palit's NVIDIA GeForce GTX 570 1280MB. Detailed (absolute) results are traditionally provided in this summary.
3D modeling
The leadership of Core i7-2600 doesn't need any comments. It's simply the best Sandy Bridge processor. Results of other CPUs lower with clock rate, which, traditionally for this low-threaded group depends on Turbo Boost. And the latter is more aggressive in Lynnfield than in Bloomfield and Gulftown. Core i7-990X is only saved by its initially high clock rates.
Final 3D rendering
Since that's what multicore processors are created for in the first place, the victory of six cores (and twelve threads) is not surprising. However, the efficiency of the new architecture plays its role as well: Core i7-990X manages to outperform Core i7-880 by 1.5 times (which is logical), but the advantage over Core i7-2600 is reduced to the humble 20-25%.
Data compression/decompression
The capacious cache and the ability of 7-Zip to use multiple threads are not enough for Gulftown to win a decisive victory. The extreme Core i7-990X, however, managed to settle at the top, but Core i7-970 loses noticeably to Core i7-2600.
Audio encoding
This benchmark favors multicore processors, and it's clear that six cores are obviously better than four, given that the architecture is the same. But we can also see that sheer performance isn't the only factor as improvements of Sandy Bridge let it catch up nicely.
Compiling
Six cores, twelve threads, 12 MB of L3 cache... The result is obvious. Especially as compilers really do not care much for the innovations of the new architecture, so the performance boost is mostly dictated by the difference in cache and core clock rates.
Mathematical and engineering computations
This reminds of the very first benchmark in our list. Although Core i7-970 looks nice enough, it still cannot outperform Core i7-2600 — that would also require an advantage in clock rate which it hasn't.
Write a comment below. No registration needed!
