G.Skill with Opteron... ZX or HZ ?

nikebee

nikebee

nikebee

Soldato
Joined
9 Nov 2004
Posts
13,984
Location
Pembrokeshire
'lo all
just a quick question, hoping to move from 1GB to 2GB soon once my Opteron gets back from RMA (which... could be some time with them being so hard to come by).
i'm currently using 2x512mb GeIL value sticks, they've suited me fine in my P4 rig but i feel the need for more, and better ram constantly growing.

so i'm stuck between these two at the moment.

G.Skill 2GB DDR ZX PC3200 (2x1GB) CAS2 Dual Channel Kit
and
G.Skill 2GB DDR HZ PC4000 (2x1GB) CAS3 Dual Channel Kit

they'll be fitting in nicely with the following once its back up and running
Opteron 146 (original could hit around 2.8GHz with the geil, but wasn't stable even at stock speeds {and below} so had to be sent back... hoping the new one when it arrives will be up to around the same)
7800GT
DFI LanParty SLI-D motherboard
and a host of other stuff as well...

with the HZ being PC4000, it should be easier to clock with the opteron?
but, the ZX kit has lower latencies which is supposedly better for performance...

so i'd like peoples opinions on each kits, whether they're good/bad etc...etc...
don't particularly want the HS kit as they don't show any advantages over the ZX or HZ sets.

cheers for the advice/help :)
 
cheers for the link, wasn't aware a load of people were split between ZX and HZ... that'll teach me not for searching.

and the opteron... a bit of an odd problem really. i stuck it in at stock speeds first time round, with a completely new build... installed windows half way onto a HDD before the computer shutdown.

a reinstall later it appears to be fine, i try and install something and... pop... shutdown.
i ran 32m superpi, shutdown... 16m... shutdown.... 8m... shutdown... 4m i think it completed.

lots and lots of testing and component swapping later...
i'd tried 2 motherboards, 2 psu's, 2 sets of ram, reseating everything, different HDDs.... etc..etc...

whenever the Opteron went under heavy load for extended period of time... it just shutdown the computer.

annoying as hell really... the irony that an Opteron... a very stable CPU and beast of an overclocker... wasn't even stable at stock speeds :(

edit:
sent it back today, the guy who replied to my RMA request said that he thinks OcUK should be getting more Opterons in stock soon. he didn't say 144, 146, 148 etc. just that he thinks some more are coming in.

i'm hoping its not too long of a wait before i get a new CPU back... if i haven't heard anything in about a weeks time, i think i'll send them a webnote asking for my money back and i'll buy a 3700+
 
Last edited:
Personally I'd go with the HZ since you will be running high FSB. I've done a lot of testing in recent weeks and found that FSB can be just as important as timings, it's about finding the right balance. Get the HZ down to CAS2.5 and with a high FSB it will likely perform as well as if not better than low latency RAM clocked down in the low 200s.
 
sounds like a plan HangTime...
any links for a readup on your findings....or are they yet to be posted?
can the HZ be pushed to low latencies easily?
 
nikebee said:
sounds like a plan HangTime...
any links for a readup on your findings....or are they yet to be posted?
can the HZ be pushed to low latencies easily?
Click to expand...
No. kimandsally very selflessly did some testing for me and he said the lowest he got was about 2.5-3-3-6 which falls in line with some of the review sites I have read on the HZ (they got it to 2.5-3-3-8 1T @ 200mhz). And it is VERY difficult to get CAS2.5 @ 250mhz on the HZ unless you get amazing sticks. Good sticks will hit about 3-3-3-5 1T :).
 
smids said:
No. kimandsally very selflessly did some testing for me and he said the lowest he got was about 2.5-3-3-6 which falls in line with some of the review sites I have read on the HZ (they got it to 2.5-3-3-8 1T @ 200mhz). And it is VERY difficult to get CAS2.5 @ 250mhz on the HZ unless you get amazing sticks. Good sticks will hit about 3-3-3-5 1T :).
Click to expand...

Yeah the best i got was 2.5 3 3 7 @ 200 and 3 4 3 8 @ 250
 
nikebee said:
sounds like a plan HangTime...
any links for a readup on your findings....or are they yet to be posted?
can the HZ be pushed to low latencies easily?
Click to expand...

I haven't 'published' anything although a search will probably turn up a post or two I've made over the past couple of weeks.... basically I found that on my setup that a 9/10(180) divider at cas2.5 was slightly faster than a 3/4(150) divider at cas2 with the other timings tightened up as well.

Given that people are having difficulty getting HZ to run cas2.5 at high FSB, maybe it isn't such a great choice afterall. I guess the real question is whether there is any reasonably priced ram capable of cas2.5 at >250fsb.
 
If i were buying again i'd probably have gone for the ZX. I suppose though, as i've found out myself today, it's not just about getting the ram with the best speed to timings ratio, it's about getting the ram that has the best speed to timings ratio that meets what the rest of your system can do. My ram can do some great times at very low timings, but i just can't exploit that with my limited divider options (Asus premium)and CPU multiplier (Opt 144). If i had a 146.....oh baby it would be smoking, 9.5x290 = 2.75 on a 5:4 divider with 2-3-2-5 timings. As it is, i need to drop to case 2.5 and go up just 11 on my fsb to get into the 2.7 area.
 
running my opty with HZ @ 8-3-3-3 compared with my old TCCD ram at 7-3-3-2.5 (both at 250mhz) has added an extra 0.1 to my super pi 1mb time.... not exactly painful :P
 
Acording to the DFI guide from dfi streets the CAS latency doesnt make much diffrence on A64 939 Boards.
 
Amps said:
Acording to the DFI guide from dfi streets the CAS latency doesnt make much diffrence on A64 939 Boards.
Click to expand...
No, the guide proved the converse, high FSB makes no difference when compared to CAS. They are true both ways round, however Angry was demostrating that high fsb/loose timings is no better than low fsb/tight timings.
 
smids said:
No, the guide proved the converse, high FSB makes no difference when compared to CAS. They are true both ways round, however Angry was demostrating that high fsb/loose timings is no better than low fsb/tight timings.
Click to expand...
True, but CAS latency itself has a fairly small impact on performance. Other timings such as tRP and tRCD have a larger impact on performance.

eg. the difference between 3-4-4-8 and 2.5-4-4-8 would be negligible, but the difference between 3-4-4-8 and 3-3-3-8 would be much greater.
 
Metalex said:
True, but CAS latency itself has a fairly small impact on performance. Other timings such as tRP and tRCD have a larger impact on performance.

eg. the difference between 3-4-4-8 and 2.5-4-4-8 would be negligible, but the difference between 3-4-4-8 and 3-3-3-8 would be much greater.
Click to expand...
Possibly, I couldn't say for sure without testing. I think it would be pretty tight between 2.5-4-4-8 and 3-3-3-8.
 
A lot depends on wether you can run 1t command rate at higher fsb. If not, then its not probably not worth going for high fsb and keep the speed lower with say 2.5-3-3-7-1t. I've got my generic memory at 210mhz @ 2.5-3-3-7-1t, though I haven't tried higher with slacker timings. Anyone here reckon 3-3-3-8-1t is possible, it roughly the same as ocuk value ram btw.
 
i see the debate is still on going :D

well i bought the ZX set today, can't wait to get my hands on the stuff.

unfortunately, i'm STILL *sigh* waiting for my processor to get back.. shall be on to OcUK if there's no word in the next few days.
 
nikebee said:
i see the debate is still on going :D

well i bought the ZX set today, can't wait to get my hands on the stuff.

unfortunately, i'm STILL *sigh* waiting for my processor to get back.. shall be on to OcUK if there's no word in the next few days.
Click to expand...
Same boat as you!!! Sent my processor back for an RMA last tuesday, still no word (not OcUK though). Going to phone them up now and find out what is going on.
 
You must log in or register to reply here.
Back
Top Bottom