Show off your Everest Latency time!

D

digitalx0

Gawd
Joined
Jan 4, 2005
Messages
801
I hate to turn this into another cliche "Show off your <insert hardware here>" thread, but here goes nothing!

See my sig below, I got 43.5ns

I'd really like to see what some of your guys are getting with your 2-2-2-5 ram, because im killing every other ram on everest's list and I'm thinking 2-2-2-5 timings would kill me. show me what you got!
 
No overclock here and i ve seen 34.5 too when i was using CL 2.5

Latency.jpg


Cheap TwinMos RAM.
 
Faethon said:
No overclock here and i ve seen 34.5 too when i was using CL 2.5

Latency.jpg


Cheap TwinMos RAM.
Click to expand...

Wow! That ULi combination PCI-e/AGP chipset really shows ultra-low latency even with JEDEC-reference CAS 3-3-3-8 memory!
 
Yeah,i m quite satisfied with the AsRock board till now.Has some weirdnesses at the beginning,but i don't overclock and it's very stable.

Those timings are with 2x1GB TwinMos entry level modules (actually one is 333 sold to me as 400 ,so i had to oc it to 400) and actually while i run the test i was on the internet and in p2p too,so might have cut off something from the performance.
 
E4g1e said:
Wow! That ULi combination PCI-e/AGP chipset really shows ultra-low latency even with JEDEC-reference CAS 3-3-3-8 memory!
Click to expand...
i have a suspicion that's a mis-report, esp cause the chipset has nothing to do with latency on a64's ;)


here's my record for now:

33.8.PNG



though i think we've have one of these threads before ;)
 
ok, I just reconfigured some things. I now I my cpu at 8x300 and my ram running 1:1

so now my ram is running at 300mhz (2.9v 3-3-3-8 1T) with two fans on them for optimal cooling :)

and my new latency is 38ns
 
There might be something wrong with Everest and ULI chipsets,but this is the lowest i ve seen (not mine).

Latency2.jpg
 
yeah, i'm guessing there's a bug that's making it output 1/2 the latency it should
 
I created a report and sent it to Lavalys.Just in case they missed something on this chipset.
 
By the way, the following Everest result of mine is pictured below:

Everest.jpg


Note that I beat an Athlon64 X2 4800+ rig with the same motherboard. (Though granted, that 4800+ rig was using 2.5-3-3-8 timings, while I'm at 2-3-3-6 timings.)
 
serbiaNem said:
I have a theory that the latency test has a lot of glitches.
Click to expand...

definately is flawed, seems like it's like sandra and just outputs "theoretical calculations"
 
(cf)Eclipse said:
i have a suspicion that's a mis-report, esp cause the chipset has nothing to do with latency on a64's ;)


here's my record for now:

http://freewebs.com/cfeclipse/33.8.PNG


though i think we've have one of these threads before ;)
Click to expand...
Yes, and I got <30 with my BH-5. I think I should move to nF4... the BH-5 isn't getting nearly enough volts on this thing here; plus the s939 opterons look nice and promising.

Edit: yes, the test is indeed very glitchy. I think it actually does measure the time it takes to read/write something; but that gets affected by many things because it has to go through your OS to the memory first. At some point, I was getting completely ridiculous scores, like over a 1000 on my 3200+ ClawHammer (the same one that I had <30ns on)...
 
Remember my 56.6ns score @ 2-3-3-x-1T?

Well, I tried a few other settings.

3-4-4-x-2T gave me about 67.5ns of latency.
2.5-3-3-x-2T gave me about 66.6ns.
2.5-3-3-x-1T gave me about 60.5ns.

Thus, the Everest latency test is good at showing relative latency times between the CPUs and memory settings. It does not accurately show absolute latency times. Also, within a given CPU/chipset family, higher-clocked and overclocked platforms will always show a lower latency time than lower-clocked and underclocked platforms at the same timings and command rate.

Thus, you can say that an X2 3800+ with 2-3-3-6-1T memory will have slightly lower overall latency than an X2 4800+ with 2.5-3-3-8-2T memory. But that does not translate into improved application performance; obviously, higher clockspeeds will make a bigger difference.
 
you have a link for this appz i can only find 'Everest corporate edition {evaluation}" and i dont see the bench maeking tool
 
maybe we should be using sciencemark, it has a nice memory benchmark built-in :D
 
you guys do know it doesnt make much sense to compare these numbers across difference versions of everest
 
You must log in or register to reply here.
Back
Top