Performance tests

Testbed configuration

• Motherboards:
• ASUS M3N-HT Deluxe BIOS 0901 based on NVIDIA nForce 780a
• Gigabyte MA78GPM-DS2H rev. 1.0 BIOS F4 based on AMD 780G
• ASUS M3N78-EMH HDMI BIOS 0404 based on NVIDIA GeForce 8200
• ASUS M3N-H/HDMI BIOS 0707 based on NVIDIA GeForce 8300
• ASUS P5E-V HDMI BIOS 0505 based on Intel G35
• Processors: AMD Athlon 64 X2 4000+ and Intel Pentium E2160
• Memory: 2 GB DDR2-800 Corsair XMS2-6400 (DDR2-800, 5-5-5-15-2T)
• Discrete graphics: ATI Radeon X1900XTX, 512 MB GDDR3
• HDD: Seagate Barracuda 7200.10 (SATA-II, 7200 rpm)
• Power supply unit: Chieftec CFT-560-A12C

Software

• OS and drivers:
• Windows XP Professional SP2, Windows Vista Ultimate SP1 (32-bit versions)
• DirectX 9.0c, DirectX 10.1
• ATI Catalyst 8.6
• NVIDIA ForceWare 174.74, ForceWare 175.19
• Intel Chipset Drivers 15.9.0.1472
• Test applications:
• 7-Zip 4.10b
• XviD 1.0.2 (29.08.2004)
• Doom 3 (v1.0.1282)
• FarCry (v1.1.3.1337)
• Unreal Tournament 2004 (v3339)
• Futuremark 3DMark06 (1024x768)
• Serious Sam II 2.07 (Shield Generator28 demo, HDR ON, available here)
• F.E.A.R. (v1.04, Maximum system, Middle graphics)
• S.T.A.L.K.E.R. (v1.0005, ixbt3 demo)
• Company of Heroes (v1.2)
• World in Conflict (v1.0, Low Quality)
• CyberLink PowerDVD 7.3.2911 (we only used VC1 and H.264 decoders), Media Player Classic (Home Cinema)

Test results

As always, we'll start with computational tests in order to find out how the integrated graphics core affects memory performance.

When nForce 780a was designed, maximum attention was paid to this aspect, because motherboards on this chipset are intended for active users. General performance drops, when the system switches to the integrated graphics core, must be minimized for them. Indeed, it demonstrates better results than GeForce 8200, although it's still outperformed by AMD 780G. And GeForce 8300 shows even worse results in RMMA test than GeForce 8200.

However, these deviations are minimal in real applications, even if we compare the best and the worst results. These results are predictable. In RMMA we also register a noticeable difference only in tests for maximum bandwidth, while average read and write rates are closer to real conditions. We are surprised even where this difference comes from, as the memory controller is built into a processor. But we have an answer to this question: BIOS code, which is usually used by motherboard manufacturers without modifications, is written by a chipset maker. It determines how a memory controller is configured (what values are loaded into internal registers of the memory controller for given memory modules -- along with main timings we have lots of auxiliary data).

The graphics core in nForce 780a is apparently similar to GeForce 8200 (in this case, to within the difference in driver versions). And GeForce 8300 indeed performs better, because its shader units operate at a higher clock rate. What concerns a comparison with AMD 780G, GeForce 8300 would have probably almost caught up with AMD 780G operating at the stock frequencies, without an on-board video buffer. But as there appeared motherboards with a video buffer already, it's more correct to compare GeForce 8300 with such competitors (in this case we've taken results of Gigabyte MA78GPM-DS2H). It turns out that AMD 780G has a little advantage again here. Besides, motherboards on the AMD chipset, unlike NVIDIA, practically all support GPU overclocking. As we can see in test results, this feature provides an imposing performance gain.

What concerns the comparison with the competing platform, no changes here. The system based on Intel G35 is outperformed 2--4-fold. It even makes no sense to evaluate price differences between CPU-motherboard combos, which widens the gap. What concerns G45, formally launched in early June, no one tested it so far. We've got information from motherboard manufacturers that it won't set the Thames on fire, that is it will hardly reach the level of even GeForce 8200.

Write a comment below. No registration needed!