LDBHE 3800 x2/Sapphire RD 580 results tonight

chew* · 03-13-2006, 07:07 PM

Originally Posted by situman

Did this thread fall off the face of the Earth?

No its been held up for certain reason at this time. Namely I will not bang up any results till I am 100% sure its 100% stable. Working on it as fast as i can. I would rather give no new info than innacurate info.

**dinos22** · 03-13-2006, 07:23 PM

nice CPU

i picked up a week 27 BH 3800+ X2...should be OK

**dogsx2** · 03-13-2006, 07:32 PM

Originally Posted by chew*

No its been held up for certain reason at this time. Namely I will not bang up any results till I am 100% sure its 100% stable. Working on it as fast as i can. I would rather give no new info than innacurate info.

Thanks for the update Chew. Check in from time to time until the fine wine is ready.

Is that red wine or white wine?

chew* · 03-13-2006, 07:53 PM

I'm fond of white wine

**Kaiser_Sose** · 03-13-2006, 08:10 PM

Can you please dl cinebench and do a cpu render test with different LDTs and see if there is in fact any performance gain with higher LDTs

chew* · 03-13-2006, 08:40 PM

Originally Posted by Kaiser_Sose

Can you please dl cinebench and do a cpu render test with different LDTs and see if there is in fact any performance gain with higher LDTs

I can and will but for now it will have to be with a 2 gig set of OCZ plats 2-3-2-5 ddr 400. How about 9x300 with 5x 4x and 3x LDT?

**Kaiser_Sose** · 03-13-2006, 09:20 PM

Originally Posted by chew*

I can and will but for now it will have to be with a 2 gig set of OCZ plats 2-3-2-5 ddr 400. How about 9x300 with 5x 4x and 3x LDT?

Sure, that be ok ...

I am more interested in cpu performance than memory performance anyways ... hopefully we some kind of benefit with a 4 or 5 LDT. Currently I get just under 800 on cinebenches 2 cpu render test with a 165 at 2.8 Ghz, 3 LDT, and 280 MEMCLOCK FREQUENCY

I need a board for my other cpu 165 UPMW

Oh, do you have any idea what the going rate will be for these boards more or less

chew* · 03-13-2006, 11:50 PM

Ok quick test with some speed binned Corsair pc 3200 speed binned UCCC. I adjusted Dqs drive strength to 50% reduced and set Beats to 8, other than that this is untweaked or as i like to call it Raw and Uncut. I will push higher tighter later.

*edit* P/N drive settings were run at default of 6/6. Also some of you may have noticed i swapped cpu's. This cpu has a known good mem controller and x2's have to many chances of having a bad 1/1 mem clock due to just a single bad mem controller.

chew* · 03-14-2006, 01:42 AM

OCZ plats 2x1 gig pc 3200 2-3-2-5 kit. I didnt bother trying to OC these as its preety much a waste of time. Anyone that has them know they max out around ddr 434ish. Just wanted to show that even when oced which increases the bandwidth some they are still stable and wanted to make sure i was stable at 300+ x 5xLDT before running cinebench.

chew* · 03-14-2006, 02:32 AM

Cinebench scores. Hope this is what you wanted. Never ran this before. I honeslty don't think this is a good bench for taxing the chipsets bus.........Need something alot more demanding IMHO that taxes the bus as the NB has quite a few lanes.

**s7e9h3n** · 03-14-2006, 03:20 AM

Cinebench isn't a very good test to compare the different LDT multis. Just for future reference, spi8m should be sufficient enough to show the performance gains from running a higher LDT

**Kaiser_Sose** · 03-14-2006, 04:01 AM

Thanks

...

hmm, no gains whats so ever ...

I hope there is some gain for using a full LDT

I find cinebench pretty good to test stability. It might not be the best but it works well. There have been times when I passed super pi mod and cinebench couldnt ... it taxes the cpu pretty darn good

**s7e9h3n** · 03-14-2006, 04:09 AM

Originally Posted by s7e9h3n

Cinebench isn't a very good test to compare the different LDT multis. Just for future reference, spi8m should be sufficient enough to show the performance gains from running a higher LDT

.

**Revv23** · 03-14-2006, 06:48 AM

yeah i think superpi or something along those lines would be better.

chew* · 03-14-2006, 10:50 AM

Ok so decided to find max Htt before it got unstable. Note I have not removed heatsinks, repasted, or done any modifications to keep my results as close to what the end user should see in a NIB retail version. Seems 1675 Htt is doable.

After 330 I started to see some hard drive corruption with 5x LDT so keep this in mind and drop the multi down after 325-330. Upping SB volts did nothing for it. 340 and The pci E slot lost video. Once again dropping multi to 4 x resolves all.

**dqniel** · 03-14-2006, 11:13 AM

325x5 is still impressive as hell

Keep the thorough results coming man

**thinkingbear** · 03-14-2006, 11:32 AM

Originally Posted by dqniel

325x5 is still impressive as hell

Keep the thorough results coming man

Agreed. Is there any measurable performance gain at 5x or 4x versus 3x? Great thread.

chew* · 03-14-2006, 11:34 AM

Originally Posted by thinkingbear

Agreed. Is there any measurable performance gain at 5x or 4x versus 3x? Great thread.

Im looking into that now. One user requested cinebench however i feel it is the wrong bench to try and measure if there are any gains.

**s7e9h3n** · 03-14-2006, 11:37 AM

Originally Posted by thinkingbear

Agreed. Is there any measurable performance gain at 5x or 4x versus 3x? Great thread.

There is some measureable gains by running a higher LDT, but the difference is so insignificant that there's almost no point in doing so. That being said, if you're trying to squeeze out every last second in a bench such as SuperPi, a higher LDT will give you a faster result (assuming you're stable enough to finish the bench)

chew* · 03-14-2006, 12:07 PM

Originally Posted by s7e9h3n

Cinebench isn't a very good test to compare the different LDT multis. Just for future reference, spi8m should be sufficient enough to show the performance gains from running a higher LDT

As per your suggestion. Note the last test is done twice once at stock volts and second with a small .08 bump to NB and HT. I did this on a hunch after seeing results from all 3. I will retest test 2 300x4 later with some voltage.

**s7e9h3n** · 03-14-2006, 12:28 PM

Nice work

I also want ot mention that Spi is a very voltage sensitive bench. You'll notice BIG swings in results from using too little voltage -> too much voltage. The key with Spi benching is finding the OPTIMAL stable settings @ a certain speed and that will result in the fastest times

**xgman** · 03-14-2006, 01:24 PM

How far off are these boards from full production?

**s e t h** · 03-14-2006, 01:31 PM

2 weeks according to the sapphire man i spoke to today.

**Revv23** · 03-14-2006, 05:47 PM

also chew keep in mind that your first run is never fastest, to try and eliminate this run in realtime mode and run 16k a few times until it stops getting faster, after that, then run a few 256k's until it stopps, then after a good 256k go to whatever you are benching, 8m in this case.

chew* · 03-14-2006, 06:17 PM

Originally Posted by Revv23

also chew keep in mind that your first run is never fastest, to try and eliminate this run in realtime mode and run 16k a few times until it stops getting faster, after that, then run a few 256k's until it stopps, then after a good 256k go to whatever you are benching, 8m in this case.

Yep took that in mind. Each one of those benches represents a fresh reboot in order to make all things equal.

Thread: LDBHE 3800 x2/Sapphire RD 580 results tonight

Thread Tools

Search Thread

Rate This Thread

Display

Bookmarks

Bookmarks

Posting Permissions