Intel HD Graphics 2000Like the other Sandy Bridge processors, the Core i3 2100 has an integrated graphics core. This is the HD 2000 with six execution units. Note that although the Core i3 2100 doesn’t have Turbo mode for its CPU cores, the graphics core does have Turbo mode which allows it to clock up to 1100 MHz (against a base clock of 850 MHz) which is identical on paper to the graphics core in the Core i5 2500. One of the particularities of Sandy Bridge architecture is that the last level cache is shared with the graphics core. Given the fact that it has been reduced in size here (3 MB) we wanted to check the impact on performance. For more details on the Sandy Bridge HD Graphics architecture, we refer you to our previous article.
Energy consumptionWe measure the energy consumption of the graphics core using the internal reading on the Sandy Bridge Power Control Unit via the hwinfo32 software:
No surprises here, at 8.5 watts in load, the HD 2000 is particularly economical. Note that we’re a long way from the TDP announced for the combined graphics and CPU load.
We measured the 3D performance of the Core i3 2100. Although the IGP is identical in its configuration to the solution on the Core i5 2500, we wanted to measure the impact, if there is one, of the reduced size LLC. We also tested the Core i3 2100 together with the Radeon HD 5450 (GDDR3).
Here are the platforms used, with Windows 7 64 bits being the OS used across all platforms:
- Intel Core i5 661, Asus P7H55M, 4 GB DDR3 1333 Crucial
- Intel Core i5 2500/2500K and i7 2600/2600K, Intel DH67GD, 4 GB DDR3 1333 Crucial
- AMD Phenom II X4 975, Gigabyte 890GPA-UD3H, 4 GB DDR3 1333 Crucial
Far Cry 2
For our games tests, we used three levels of different performance:
- 1280 x 720 low
- 1280 x 720 medium
- 1680 x 1050 medium
In Far Cry 2, the graphics modes used correspond to the low and medium modes on offer in the game. These modes use DirectX 9 exclusively.
The results with the Core i3 2100 HD 2000 are more or less similar to those with the i5 2500, except at the lowest quality mode where the processor puts the brakes on slightly. Coupled with the Radeon HD 5450, the trend is confirmed, the processor has an impact on performance in spite of the very modest graphics settings.
The Low and Medium modes correspond to the ‘mainstream’ and ‘gamer’ modes in Crysis Warhead:
Performance levels between the Core i5 and Core i3 are closer here but the same thing is happening in terms of the CPU.
CPU impact vs IGP
In our Sandy Bridge test, we tried to measure if there was any impact resulting from the fact that the last level cache (LLC) was shared between the IGP and the x86 cores.
To try and get a reading for this, we opted to take a reading of the graphics and processor performance at the same time using Cinebench + Tom Clancy’s H.A.W.X. together. The percentage indicates the number of threads used for Cinebench (1, 2, 4) and corresponds to the percentage of threads in comparison to the number of physical cores on the chip (200% corresponds to 4 threads on the Core i3 2100 which has two physical cores + HT). We looked at the development in performance in these three cases:
- Core i3 2100 + Radeon HD 5450
- Core i3 2100
- Core i5 661
While you’d expect to see a slight dip in performance on a dual core as you’re cumulating processor and graphics load (2.9% in the case of the Core i3 2100 + HD 5450 for graphics performance), it should be limited. When the Core i3 2100 is on its own with its IGP, graphics performance drops by 23%.Hold the mouse over the graph to see indexed performance.
We are still waiting for Intel to get back to us to clarify this issue that we had already noted with the Core i7 2600K. To recap, we had noted, other than the dip in performance, a dip in the IGP turbo clock. We then also recorded a drop in the IGP energy consumption using hwinfo 32. In the case of the Core i3 2100, the energy consumption in H.A.W.X alone is around 6 watts on a fixed game scene (moving at altitude). Launching Cinebench then leads to an almost instantaneous dip in the GPU energy consumption, falling to around 5 and then 4 watts and going as low as 2.5 watts in some cases. When we stopped Cinebench there was an immediate increase in performance and clock and the energy consumption rose to 6 watts instantaneously. In a similar scenario with the Core i7 2600K, the dip was from around 10 watts in IGP load to 3 watts in IGP and CPU load.
With an equal number of threads (4), it’s interesting to note that we observed a 22% reduction in graphics perforamnce on the Core i7 2600K, which is almost identical to what we got here. If this isn’t directly due to sharing the LLC, this dip in performance may well be linked to the ring bus which links the LLC to the x86 cores and IGP.