Gigabyte GA-EP45 Extreme rev1.0 Review

04-09-2008, 02:58 AM #25 (**permalink**)

hey CrypticK

Don't forget mem clocks will vary on mem dividers as 2:3 is one of the easiest dividers to clock memory on. Unfortunately, no access to the Asus derivative boards heh

21-10-2008, 08:34 AM #26 (**permalink**)

Hi eva.. How did you read the n,b/mch readings as on my ga ex38 board or everest ultimate i cant read what the voltage is for this setting.
Is this shown in the extreme mobo?
thanks mate

21-10-2008, 02:57 PM #27 (**permalink**)

no way to read the mch/nb volts even on extreme one unless you find the digital multimeter read points and probe the board with a multimeter

01:24 AM

Hey eva, I have been doing a little testing on my MIIF after looking over your benches on the first page of this thread. Our boards use different naming conventions for the various MCH settings but I'm sure you're familiar with the terms Asus use.

I have been doing some testing with Static Read Control (SRC) & Dram Read Training (DRT) enabled, and Clock Twister (CT) on Strong or Stronger. With Clock Twister set to strong/stronger it takes quite a bit of vNB to stabilize it, I was around 1.37 - 1.45 vNB for the various tests. The use of SRC & DRT and setting strong or stronger gives a definite boost to bandwidth across the board and reduces latency significantly. At 534 FSB 5:6 Divider I had to drop CT to Strong, as even with 1.49 vNB I could not get into windows (I'm not prepared to use 1.5 - 1.6 vNB just yet). Also notice the latency is almost identical between 500 FSB 5:6 CT Stronger and 534 FSB 5:6 CT Strong.

At 500 FSB 5:6 and 534 FSB 5:6 I can out-do the Gigabyte in bandwidth but not by much, and at 534 FSB 5:6 the Gigabyte has slightly lower latency. It seems on my board that Clock Twister lowers latency significantly and has a noticeable but not large effect on bandwidth, and tRD (PL) has a great effect on bandwidth and a moderate effect on latency.

Very nice benching on your behalf, it was incredibly comprehensive and easy to follow.

445 FSB 5:6 Divider, tRD (PL) 8, SRC & DRT Enabled, CT Stronger (my 24/7 settings):

450 FSB 3:4 Divider, tRD (PL) 8, SRC & DRT Enabled, CT Stronger:

450 FSB 3:4 Divider, tRD (PL) 7, SRC & DRT Enabled, CT Stronger:

500 FSB 5:6 Divider, tRD (PL) 8, SRC & DRT Enabled, CT Stronger:

534 FSB 5:6 Divider, tRD (PL) 8, SRC & DRT Enabled, CT Strong:

27-10-2008, 01:34 AM #29 (**permalink**)

Thanks for sharing your results.. yeah Gigabyte slightly slower on latency, but it might be due to the Crucial Ballistix PC2-8500 modules/Team PC2-10400 modules used as I have moved onto DFI DK X48-T2RBS Plus for testing and using Team PC2-10400 vs Crucial Ballistix PC2-8500 Tracer, at same 7.5x533FSB 5:6 639Mhz 5-5-5-9 timings, there's a noticeable difference in everest latency - Team ~46.8ns vs Crucial ~46.1ns.

I'll have those results in a DFI DK X48-T2RSB Plus thread later this week

27-10-2008, 01:51 AM #30 (**permalink**)

Interesting that there is a latency difference between different ram kits, perhaps it has to do with the timings set by SPD that we do not have bios control over? The tests I just posted were done using Team PC-8500 modules. I should compare them to my CellShock PC-8000 CL4 modules to see if there is a difference between them. I'll keep an eye out for the DK-X48 results, I was almost going to buy that board but decided not to due to people struggling to achieve high FSB speeds with any stability. I'd like to see what you can do with one, if it is just a retail board and not 'hand picked' by DFI for testing/promo purposes. I really want a good X48 board, the Rampage Formula just doesn't quite do it for me, although I'm still considering one.

27-10-2008, 02:31 AM #31 (**permalink**)

Yup differ memory have different SPDs which can account for the differences

I posted my findings in DFI DK X48-T2RSB Plus thread I just started DFI Lanparty DK X48-T2RBS Plus bios, info, photos & overclocking tips

DFI DK X48-T2RSB Plus much better on fsb than prior DFI X38/X48 boards which for me stopped around 500-510FSB, DK X48-T2RBS ~563FSB for 32M pi and 571FSB for validation. Praz over at DFIClub forums got to around 550FSB as well.

26-12-2008, 01:15 PM #32 (**permalink**)

From now having 3 gigabyte p45 boards, i have found that the memory read, write, latency is in general about 200-400 points lower on everest than asus or dfi boards, dont know why, in some cases i have seen up to 700 points lower in write scores with same setting, thats why i wont own a gigabyte board anymore, i wrote gigabyte and addressed the issue with them, and they gave me some stupid responce.