TCCD with higher HTT vs BH-5 with tighter timings?

[[XC] serlv]

Does anybody know how this works with Rosetta?

Using a DFI Ultra-D, would TCCD @ 280 (+) ,and obviously with looser timings, bench as high, or higher, as BH-5 @ 250-255, with tighter timings?

[Rodzilla]

Does anybody know how this works with Rosetta?

Using a DFI Ultra-D, would TCCD @ 280 (+) ,and obviously with looser timings, bench as high, or higher, as BH-5 @ 250-255, with tighter timings?
I would think that 280 would give you more BW... so my guess would be TCCD

[Fr3ak]

I am not so sure about that. Tight timings seem to make t hell of a change.
I think tight timings at 255 is faster than 280 tccd.
I dont have TCCD, otherwise I would test it :/

[Entity_Razer]

a shame I can't push these xeons any harder (crap psu and cooling) otherways I got some RAM here (cheap twinmoss ram) that can do 235 with 6-2-2-2.5 timings.... I know its not much but for xeons it is :D

shame :)

[Rodzilla]

yea... if I had it I could test it.

Try it for us Serlv!

[[XC] serlv]

Trying to get a hold of some...

[ex2cib]

i can run a bandwidth test for you of really high clocked TCCD, its not stable there, but i can run a quick test if you just want to know numbers, a64 3000+.

i can do either 9x300 2.5-4-3-7 or 9x295 2.5-3-3-6.
ill rerun the bench once i do it so im not cheating, im just offering to do this if anyone wants to know the rosetta bandwidth values for higher clocked TCCD.

just find someone else that can get about 2700 with bh5 type ram in the 250-260 range i guess. venice CPU if possible.


actually, as soon as i finish this current work im on, ill go run it and get the values for you

[ex2cib]

well, so much for that, i loaded up settings that were one time 32m stable, blue screens every single time, my processor has gone downhill lately for some unknown reason. i still think TCCD would have the edge over bh5 in this. ive hit some crazy numbers before with just pushing my processor

[[XC] serlv]

I'll be testing in a few days, have some things on the way.

[Rodzilla]

It will be good to know whether timing or bandwidth make more of a diff...

[Haltech]

Well i got some scorching TCCD stuff here, OCZ PC5000 and some OCZ VX stuff.. The VX is good for 260 @ 2,2,2,5.

did you guys just want a benchmark within Rosetta?

I can do same clock mhz and multiplier for both ram. I can also do dual channel NF4 and single channel NF3 since a lot of you use semprons.

[ex2cib]

try with the TCCD in the 280-300 range and the VX 250-260

[[XC] serlv]

yes, within Rosetta. thanks.

[Haltech]

I thought i was going to be able to get this done in the next hour. I just got called in for work early, so it wont be happening tonight.

[[XC] serlv]

I thought i was going to be able to get this done in the next hour. I just got called in for work early, so it wont be happening tonight.

no prob. we'll get it eventually. you, me, someone else, maybe some reduncy to verify. but we'll get it.

[Rodzilla]

no prob. we'll get it eventually. you, me, someone else, maybe some reduncy to verify. but we'll get it.
He's looking for intelligence for XPC!!! Run him down!!!

[mongoled]

Well here are the numbers for the rig in my sig

4005 double precision MIPS
12284 integer MIPS

this is with all my background programs running, please let me know if you want me to run the benchmark with no programs open/clean boot

[Fr3ak]

I have 4020/12300 with my Opteron 144 at 2925Mhz and 2x512MB Twinmos UTT running at 244Mhz 2 2 2 5 , also with a ultra-d

[Aphex_Tom_9]

so according to mongoled and fr3ak 300 2.5-4-3-7 is a little under 244 2-2-2-5.
innnnteresting...

[[XC] moddolicous]

Yea, right now it looks like tight timings are a little better, although 2.5-3-3-7 might be able to tie it. Question is price: What is cheaper: 2x512mb that could run 240 2-2-2-5 or 2x512 280 2.5-3-3-7. Anyone got some 300 or higher?? Only one way to really test the bandwith issue. 250 2-2-2-5 definitely doesnt have as much badwith as 300 2.5-3-3-7.

[Vapor]

Where would 275 2-2-2-3 go then? :D

[[XC] moddolicous]

Who runs 275 2-2-2-3 on a dedicated rig?! Maybe when skt 939 becomes like skt A, we'll see 275 for a dedicated cruncher, until then, no. BTW, can anyone confirm if the PC-DL can go above 240fsb? I think I remember ddtung saying it doesnt go higher than something.

[Vapor]

Dedicated rig? No one I think. Daily rig? I certainly can (I don't because my CPU has problems holding the 10x multi for more than a day--don't ask). Never the less, I think comparing at the extremes of daily stability would also be useful (well, not really, but definitely INTERESTING). 272-275 2-2-2 vs. 300-306 2.5-3-3 vs. 330-340 x-y-z (probably 3-4-4)

[[XC] moddolicous]

Yea, thats true, I forgot about daily rigs. I think in 275 2-2-2-5 vs 322 2.5-3-3-x (I think thats the 32m record for 2.5-3-3-7) would be something to see.

[Vapor]

Probably with that is the discrepency with CPU clocks at those speeds ;)

Anyway, the tests should also be at daily CPU speeds (so like 2.5GHz to 3.4GHz would be acceptable).

EDIT: also, would having higher bandwidth on dual core be more beneficial than on single core?

[mongoled]

Well to be a real comparisson we need to have a baseline. So whts it to be?

1/ Clean boot
2/ Use msconfig to disable all start up services
3/ Set baseline CPU speed
4/ Benchmark

Wht baseline CPU speed should we use?

[Fr3ak]

I just booted, started Boinc and benched. I had smartguardian, razersoftware and the usual stuff running in the background.
Unfortunately my cpu doesnt like to run at 3000Mhz.
My raptors dont run in Raid 0 either, but I guess Hdd performance aint important for that benchmark.
I will try to run it closer to 3ghz in a bit.

[Vapor]

Well to be a real comparisson we need to have a baseline. So whts it to be?

1/ Clean boot
2/ Use msconfig to disable all start up services
3/ Set baseline CPU speed
4/ Benchmark

Wht baseline CPU speed should we use?No baseline....just get it running as fast as you can (without bugging it), similar CPU speeds are logical (i.e., 322MHz 2.5-3-3 is a useless comparison to 275 2-2-2 as their respective CPU speeds will never be equivalent in the logically used MHz range). For example: 315x8 = 280x9. 315 2.5-3-3 isn't as impressive as 280 2-2-2, IMO, but they'd be a great comparo.

If anyone has TCCD that can do 306 2.5-3-3 and a 2.75GHz 1MB A64 or X2, I'd be happy to run my UTT against them in the benchmark at 275x10 vs 306x9. If both your TCCD and CPU can go higher to say 2.8GHz and 311MHz (not in that order, LOL), I'm sure I could try to match you, but that would start to be pulling away from the dedicated cruncher typical speed.

[Fr3ak]

Cant get anything above 2940 stable right now, so I ran it again with my previous settings.
No windows tweaks or anything else, just rebootet, started the BOINC manager and run the benchmark.

With a lower ram divider, Ram at 212Mhz 2 2 2 5 and cpu at ~2940Mhz I got 39xx/11990.

Edit: same settings as in the screenshot, but 2 3 3 7 instead of 2 2 2 5 timings scored 3996/12137.

Edit2: I tried 2.5 4 4 8, but I cant boot with those timings...

And I ran my defaults again and got 4083/12284.

Looks like clockspeed is the most important thing. After that tighter timings seem to be favoured.

[mongoled]

Well looking at your results it does seem tht the program favours tighter timings. I think im going to have to reboot the system to check (uptime has been more than 30hrs).

But really you should run with everything disabled, use msconfig, then we can have even baseline at least in terms of apps not running in the background. My computer is running quite alot of services and im sure they effect the result (2xSamurize clients, SQL server client, AV, Firewall, IIS, atitool etc etc).

So whts the procedure to 'stop' this program, just close it and reboot, will anything be lost?

-EDIT-

Well i just shut it down, disabled everything with msconfig and ran test, here are the results

4118 double precision MIPS (Whetstone) per CPU
12324 integer MIPS (Dhrystone) per CPU