Cache, consumption, overclocking

The impact of cache

To start with, we wanted to compare the performances of the Conroe 4 MB, Penryn 3 MB and Penryn 6 MB in our test applications and this at equivalent frequencies in order to concentrate on the impact of cache and architectural improvements.

More specifically, we compared a Q6600, Q9300 and QX9650, all at the same frequency of 7x333, or in other words, 2.33 GHz. The Q6600 was attributed a 100% in the performance index in order to facilitate the comparison of scores :


Overall, between the Conroe 4 MB and Penryn 3 MB, performances are similar as the difference is +/- 2%. The exception is found in video encoding where the benefit of SSE4 is felt, notably in TMPGEnc as well as with DiVX.

We can see that the transition from 3 to 6 MB on the Penryn has a real impact on performances : 2.4% in 3ds, 1.7% in Maya, 6.3% in WinRAR, 3.4% in TMPGEnc, 3.1% in DiVX, 0.7% in Nuendo, 3.5% in Crysis and 6.1% in World In Conflict.

Power consumption

We measure power consumption of the processor’s power supply stage with the help of an ammetric clip on the ATX12V connection line which it uses exclusively. This allows us to better isolate CPU power use instead of obtaining overall consumption. The only thing we should keep in mind is that the CPU power supply stage has an efficiency of between 80 and 90%.


The figures obtained by Intel 45nm processors are impressive to say the least. First of all, at rest in the worst case scenario we are at 11.2W with the QX9650, while the E7200 consumes almost nothing! In load, results are also excellent because the Q9300 is close to 45W versus more than 70W for a Q6600 G0! As for the E7200, it’s content to run with a little less than 21W, an impressive number for a desktop CPU.

Overclocking

Starting with a relatively low VID of 1.05v, of course we wanted to know to what extent the E7200 and Q9300 are suited to overclocking. This involves stable overclocking, validated by as many sessions of Prime95 as there are cores for 30mn.

We started with the E7200 set at 9.5x266 (2.53 GHz) by default. The first objective was to attain 3 GHz, the frequency reached by E8x00s with a voltage of 1.05V. In this case, it didn’t even boot and we had to move up to 1.25V in order to reach our target. Increasing voltage to 1.35V, we weren’t able to attain a stable 3.1 GHz, although apparently some have obtained better scores with their E7200s. Unless somehow limited in this area, overclocking values of this CPU thus seem to be random.

And the Q9300? The first thing to point out is that it functions starting in 7.5x333 (2.5 GHz). We therefore had to push the FSB rather far to obtain high frequencies; however, not all motherboards are capable of this, especially with a quad core. In actuality, we were able to attain 3 GHz (7.5x400) at 1.2V and then 3.22 GHz (7.5x430) at 1.25V.


Beyond this value, the system was very unstable and it proved to be an FSB limit related to the CPU. We couldn’t surpass 430 MHz by lowering the coefficient while with a Q6600 G0 the motherboard we used (a P5E based on an X38) functioned easily with an FSB of 450 MHz. Otherwise, the overclocked Q9300 consumed 61.2W versus 44.9W with standard settings.

In short, while the results of the E7200 and Q9300 are far from being catastrophic, we could have expected better. For comparison, the latest E8500 that we received was at 3.4 GHz for 1.2V and 3.8 GHz for 1.3V.

The test

We now move on to the comparison of these processors with other dual and quad cores with our test suite. The following configurations were used:

- GeForce 8800 GTX / ForceWare 169.01
- 2 x 1024 Mo DDR2-800 4-4-4
- 2 x Raptor 74 GB
- Windows XP SP2 French
- Socket 775 : ASUSTeK P5E Deluxe
- Socket AM2+ : ASUSTEK M3A32-MVP Deluxe

3ds Max 9 and Maya 8

3ds Max 9 and Maya 8

In this test, we used two test scenes from Maya and 3dsmax developed by Yann Dupont of 3DVF (whom we thank for their use) and which employ the MentalRay rendering engine. This isn’t a random choice as this engine is now a standard part of both programs and is the most commonly used in production.

- the scene with 3dsmax is very heavy in terms of polygons and the number of objects. The objective was to test processor capacity and manipulate a heavy flow of data.

- Maya's scene is much lighter, but uses MentalRay's advanced lighting algorithms and employs the processors’ raw power in terms of mathematical calculations.


In both situations, the Core2 Q9300 is faster than a Q6600 and Phenom X4 9850. For its part, the E7200 fulfills its role as it surpasses, for example, the ‘’good ole’’ E6600, although this is only normal given the frequencies (2.4 vs. 2.53 GHz). Finally, the A64 6400+ is equivalent to the E8200 in terms of performances.

TMPGEnc 4.0 XPress and DiVX 6.8

TMPGEnc 4.0 XPress

Version 4 of this video encoder integrates SSE4 optimizations in its latest patch. The test consists of the encoding of a 10 minute 16 second DV file in MPEG-2 format in 720x576 with an average bitrate of 4500 kbits /s and in 2 paths. The video preview display is activated during this test.


SSE4 bears its fruit and the Q9300 breaks away notably from the Q6600 as well as the Phenom 9850. As for the A64 6400+, it can’t compare to the E7200.

VirtualDub & DiVX 6.8

We now use the version 1.7.8 of VirtualDub and the version 6.8 of DiVX, which has SSE4 optimizations. We encode the same video source as with TMPGEnc in Fast recompress mode and with the DiVX 6.7 codec in one path with an average bitrate of 1500 Kbits /s, the highest quality encoding. Also, Experimental SSE4 full search is activated in SSE4 or SSE2 mode. The video preview mode is activated during this test.


Gains related to SSE4 here are even more significant and this time the Phenom is far from competing as it previously could with the Q6600. It’s the same for the A64 6400+ compared to the E7200, which is ahead of the E6850.

WinRAR 3.7 and Nuendo 3

WinRAR 3.7

Since its version 3.6, WinRAR was given multithread optimizations. We compress in RAR at the highest level a total of 588 MB of files comprised of 493 Word & Excel files (69 MB), 22 e-mail Eudora files (251 MB) and a single wav audio format file (268 MB).


The Core 2s are always ahead of AMD, although only slightly. In itself, the transition from the Q6600 to the Q9300 adds only a very small gain.

Nuendo 3

Something new in our test suite is Nuendo in version 3. This is a solution devoted to production and audio post-production. The test consists of exporting a relatively heavy project to an audio file (thanks to DraCuLaX for the file).


Aided by its frequency, the Q9300 does notably better than a Q6600 here and even surpasses the Phenom 9850. For once, the A64 6400+ proves to be faster than an E7200 while the latter is ahead of our reference E6600.

Crysis and World In Conflict

Crysis and World In Conflict

The latest very popular FPS, Crysis, whose demo was just released is used for its processor test integrated to the game. AMD performances are behind here as a « simple » E6600 is enough to surpass the Phenom as it does the Athlon.


World In Conflict is a recent real time strategy game with a rather resource heavy integrated benchmark. It’s the same situation as in Crysis for AMD processors. As usual, we witnessed the E7200 surpass an E6600 while the Q9300 was slightly faster than a Q6600.

Conclusion

Conclusion

So what do we think of these new Intel 45nm processors intended for smaller budgets in their respective domains? They should be considered separately.
The E7200 is rather convincing because it replaces the E4700 for the same price. Of course, the frequency is slightly lower, 2.6 vs. 2.53 GHz, but there are the advantages related to Penryn architecture, low power consumption due to the 45nm and an extended cache (3 MB vs. 2 MB). It’s unfortunate that overclocking is limited or at least is random as our test results suggest; however, given the price of this CPU, maybe we are being a bit too demanding. In the end, the E7200 is a good processor that perfectly fulfills its role. Those looking for better overclocking should opt for the E8200.


By the way, we wanted to warn overclockers about Intel 45nm CPUs. As the VID is lower and even if these processors seem to easily bear higher voltage, we should be careful about not surpassing certain values beyond which the durability in the long term is not assured. On 65nm versions, Intel set this figure at 1.55V and it falls to 1.45V for 45nm models.

The Q9300 is amazing like all Intel 45nm processors for its low consumption for a quad core. For example, we are below an E6850 while the Phenom 9850 consumes 2.7x more! At equivalent frequencies, it generally has the same performances as 65nm quad cores, except of course when SSE4 is involved. However, the small problem is that while it initially was positioned at the same level as the Q6600 (2.4 GHz), the latter’s price has been lowered to $224 versus $266 for the Q9300 (2.5 GHz) which is now equivalent to the price of the Q6700 (2.67 GHz).

The Q6700 will have higher performances while of course consuming more energy. In addition, we saw that overclocking a Q9300 is not the easiest, because with a coefficient of x7.5 we have to quickly rely on a higher FSB to attain honorable frequencies: 427 MHz for 3.2 GHz and 453 MHz for 3.4 GHz ...or frequencies that aren’t necessarily attainable on all motherboards or even CPUs (FSB wall)!


In the end, it’s clear that these 45nm models are far from being revolutionary. Also, unfortunately AMD isn’t “pushing” Intel to position its products in a more aggressive manner. The E7200 is basically a very good CPU, with high performances and very low consumption, but sadly overclocking is random at best, and in the worst case it has been limited. As for the Q9300, it has trouble setting itself apart from 65nm competitors in terms of performances and overclocking and it’s only distinguished by the low power consumption.

Clearly, Intel’s revision of prices for quad 65nm models wasn’t without motive because in actuality there is an availability problem with 45nm versions. Between a less expensive Q6600 65nm and a pricier Q9450 but with a higher coefficient and more cache, it’s difficult to find a place for the Q9300 ... while waiting for future price reductions!