Test Setup
Test System |
---|
CPU: | AMD Athlon64 3200+ Winchester (multi lowered to 9) |
---|
Motherboard: | DFI LanParty NF4 UT (Bios 623-3) |
---|
Memory: | 2x 1024 MB G.SKILL F1-4000USU2-2GBHZ PC4000 |
---|
Video Card: | MSI NX6600GT TDI128E |
---|
Harddisk: | Hitachi T7k250 160GB |
---|
Power Supply: | TSP 420W |
---|
Software: | Windows XP SP2, Forceware 81.85 official |
---|
We will be testing this memory at 1T timing only, because that is what enthusiasts are using to get maximum performance out of their memory.
Alpha Timing settings in Bios:
First we tried to determine the fastest timings for 200 MHz at 2.6V, the result was 2.5-3-3-8. After this we increased the voltage and tried to lower the timings again - no go. So it is not possible to run CAS Latency 2 at 200 MHz - not the best result for 200 MHz.
Then, we tested how far we could overclock the memory at 2.6V with tightest timings possible. We reached 228 MHz. When using the more relaxed timings of 2.5-4-4-8, our overclock dropped four MHz down to 224 MHz, which is strange. We did not run any benchmarks at these setting, since the 228 MHz @ 2.5-3-3-8 will yield better performance in all cases.
Now we changed to 3-3-3-8 and got 242 MHz, then 263 MHz at 3-4-3-8. When running at the manufacturer's official specification of 3-4-4-8 we could overclock from 250 MHz to 275 MHz. That's a nice result for a 2 GB kit. Generally, 2 GB kits do not overclock as well as 1 GB kits.
Now we increased voltage in several steps to 2.8V, 3.1V and 3.3V. We saw absolutely no increase in maximum clock speeds.
The last test "JEDEC DDR-400A" is for comparison with a generic DDR module running at JEDEC standard timings.
G.SKILL F1-4000USU2-2GBHZ PC4000 |
---|
CPU Clock & Memory Ratio | Memory Speed | Memory Timings | Everest Read | Everest Write | Everest Latency | Quake 3 Timedemo | 3DMark 2001SE | SuperPi Mod 1M |
---|
9 x 200 1:1 | 200 MHz | 2.5-3-3-8 2.6V | 5789 MB/s | 2200 MB/s | 49.3 ns | 431.1 | 18554 | 42.12 s |
---|
9 x 224 1:1 | 224 MHz | 2.5-4-4-8 2.6V | - | - | - | - | - | - |
---|
9 x 228 1:1 | 228 MHz | 2.5-3-3-8 2.6V | 6445 MB/s | 2394 MB/s | 45.7 ns | 446.7 fps | 19032 | 40.98 s |
---|
9 x 242 1:1 | 242 MHz | 3-3-3-8 2.6V | 6616 MB/s | 2445 MB/s | 44.6 ns | 468.4 fps | 19671 | 38.24 s |
---|
9 x 250 1:1 | 250 MHz | 3-4-4-8 2.6V | 6951 MB/s | 2524 MB/s | 45.0 ns | 476.6 fps | 19700 | 37.52 s |
---|
9 x 263 1:1 | 263 MHz | 3-4-3-8 2.6V | 7442 MB/s | 2649 MB/s | 44.1 ns | 503.1 fps | 20155 | 35.52 s |
---|
9 x 275 1:1 | 275 MHz | 3-4-4-8 2.6V | 7729 MB/s | 2790 MB/s | 40.6 ns | 520.9 fps | 20817 | 33.75 s |
---|
JEDEC DDR-400A | 200 MHz | 2.5-3-3-8 2.6V | 5789 MB/s | 2200 MB/s | 49.3 ns | 431.1 fps | 18554 | 42.12 s |
---|
If you are going to run this memory under the specification of 250 MHz, there might be better alternatives with tighter timings like G.SKILL's F1-3200BIU2-2GBHX. But if you want to overclock to high memory speeds, and need 2 GB of memory, you will love these modules.
SuperPi - 275 MHz; 3-4-4-8; 2.6V and SysTool Memory Timing screenshot
Benchmarks
For easier comparison with other modules, we set the maximum voltage required for the best result (here 2.6V; max 3.1V) and tested until we found the highest clock frequency and fastest timings for this memory. The benchmarks Everest Read, Everest Write and Quake 3 were run. We then calculated the performance increase in percent compared to some standard DDR-400 memory running at JEDEC standard timings (2.5-3-3-8). The average percentage of the three benchmarks is listed in following table: