I've solved the problem, can a mod please amend the thread title?
Significant find inside.
(edited to remove rant factor, sorry guys - Gaz)
Printable View
I've solved the problem, can a mod please amend the thread title?
Significant find inside.
(edited to remove rant factor, sorry guys - Gaz)
Maybe you got a bad stick? Does memtest give you errors at stock?
Sorry guys.
Is this ram in beta ?
Sorry for ranting.
cactus can you list your setup.
I am curious becuase your BH-5s were incompatible also
try not to be hasty in making calls , I have had pretty good luck with what playing I have done
Cactus, Iam running the same stuff on the DFI NF2 Ultra without problems up to ~255 MHz.
Can you list your latency settings, frequencies and voltages as well as the complete system specs?
I am sure we can get to the bottom of this pretty soon.
for starters, I would think you should try with a single DIMM only in slot #3 with changing the DIMMs to make sure that it is not one single DIMM that screws you up.
Hello Cactus Jack. Seems you are having issues. What motherboard are you running?
I have spent many hours testing the EB RAM and I assure you that if it was $hit we would not have released it. All of the benches and post I made were the real deal.
Email me off my sig for some one on one, I'll give you my IM and we will get you running. Otherwise give me some more system info and we will do it here.
Lets get it ON!!
hope it was a bad stick i was goin try this stuff out guess ill wait.
Well ,
I did find a guy whos having good luck on his Nforce 2
over at DSL reports forums http://www.dslreports.com/forum/rema...8764~mode=flat
Also Megahurtz , rember cactus had issues with Bh-5s also
Ryan I was just going to post Podunks results. He actually got them after seeing most of my results:)
Cactus lets not be hasty. Possibly you got a bad stick, and I am SURE the OCZ guy's will work with you.
so it should be okay with intel? which would benefit me the most at 250fsb? 3200eb, 3500eb, or th3700eb i want the tightest timings. voltage not aprob. ive had my 3.3 rail to 3.90 w/out probs
I don't think there will problem on any platform. Cactus more than likely has a bad stick.
As far as what's best, prolly the pc3500 or pc3700 if you can get it. For 250 the 3700 might be best. Ryan??
I had great luck with the Asus P4C
at 3.3 volts 500mhz I would guess 3-2-2 is a sure bet but you might want to stick a fan on the ram
My guess is that cactus since he had problems with bh-5 also
probably has some other issues .
I've ben testing the bejeesus out of NF2 boards and had great luck. The really cool part is that the 512mb modules work great and overclock just fine. I'll get stickers out of Cactus' behind when he comes on tomorrow.
so which ones 3200,3500,3700 would you recommend? im about to order but would like your opinion?
megahurtz-oc,
for 250+ on the 875P, try the 3500EB (or 3700EB) and shoot for 250, 2.5-2-2-5. Let us know!
C
I am doubtfull on 2.5-2-2 @ 500 , unless you guys have seen people doing it
I see stability issues @ 460and above 2.5-2-2 on 875 boards at normal voltages ( 2.85 )
3-2-2 seems more realistic
trying to be helpfull
also keep in mind , while I was able to run 500+ pretty easily that even the 3700 is only speced at 3700 ;-)
My opinion on 3500 vs 3700 , is definetly get the 3500 and save yourself some money
:toast:
well, the EB stuff really works great at low latencies, that is the main thing for me.
http://www.lostcircuits.com/memory/sanded/all.gif
thanks ryan just ordered some 3500. ill keep ya guys posted. hopefully ill get this stuff at same time as my new 3.4ee.....
nice...keep us up to date.Quote:
Originally posted by megahurtz-oc
thanks ryan just ordered some 3500. ill keep ya guys posted. hopefully ill get this stuff at same time as my new 3.4ee.....
C
I'll push it a bit more tomorrow... time to hit the sack but I am running rock-solid at 2.5:2:2 and 220 MHz, what else do you want?
255MHz is no problem either but I have to relax the latencies then.
Mega -
When you get it , drop me a pm , I would love to get involved in helping you take it to the max :-)
Hmm..this EB ram gets hot enough that it needs a fan right on the mem to get higher? Seems like even with 3.3v my hyperx doesn't get hot enough to need a fan, and even with a fan I don't think the difference would be that dramatic.
a fan does help. I have seen a few ° drop with the fan on than without.
Trust me, anything over 3.0v without sufficient airflow is really pushing it.Quote:
Originally posted by rc7
Hmm..this EB ram gets hot enough that it needs a fan right on the mem to get higher? Seems like even with 3.3v my hyperx doesn't get hot enough to need a fan, and even with a fan I don't think the difference would be that dramatic.
Good expample is the lan I am at right now.. ~30c ambient temps. With my side fans off and my ram running 2.9v I cant hold my finger to the bare chips.... let alone the vdimm mosfet :eek:
The extra temps have been wreaking havoc all over since the lan started.. :stick:
While benching my KHX3200 at 2.85v 224mhz they do not get more than just "warm"...and there isn't a fan pointed directly at them. Mind you that is low voltage but that isn't far from your 2.9v.
thanks man i may take you up on that.Quote:
Originally posted by ryanpgroovy
Mega -
When you get it , drop me a pm , I would love to get involved in helping you take it to the max :-)
LOL
I mean with the crazy voltage and oc people mean to throw at it here. Keep in mind people are wanting to see 500Mhz plus out of 3200/3500
most people in these forums that I know of are trying to get there maximum overclock and even with your kingston a fan might incresase the speed
above 2.9-3.0 volts I would not run any ram
without a nice 80mm sitting on top. personaly that is
The EB RAM does get warm at higher voltages. Always have a fan blowing in modules when running high voltages. A fan will do a good job of keeping them cool. The copper heastspreaders are drawing the heat off the IC's.
Do you REALLY believe heatspreaders give cooler module surface temps compared to naked modules????Quote:
Originally posted by andyOCZ
The EB RAM does get warm at higher voltages. Always have a fan blowing in modules when running high voltages. A fan will do a good job of keeping them cool. The copper heastspreaders are drawing the heat off the IC's.
:stick:
C
Well, I've toned down my rant as I am now a more calm, collected person. I was just hacked off at the time and today I'll start giving individual sticks a go.
anywayz, my specs:
2500XP-m (capable of 2.8Ghz, ran no higher than 2.5 whilst testing)
DFI Infinity (4/29 & 1/21 bios used)
Raddy 9800pro (kept at stock speeds for testing)
Antec TC550 (+3.3v @ 3.7v)
Audigy 2.
Entire system cooled by airconditioning! Nbridge cooler replaced with Vantec Iceberq.
CPU Voltage: ran no higher than 1.9 volts whilst testing.
CPU temp: went no higher than 18 degrees at any point.
Board voltage: tried all from Auto to 1.9
Board temp: Went no higher than 20 degrees at any point.
Vagp: Tried 'em all.
I've lumped the whole lot above together because they did not seem to help the situation for better or worse, I was also using my CPU and AGP well within (my) spec to eliminate their effect.
for most test I used a multiplier of 10, occasionally I used 10.5 and 11 just to check it wasn't the 10 multi causing problems as I never use it.
Other stuff:
All onboard stuff turned off (with the exception of IDE/floppy)
CPU interface: aggressive
Ram speeds: tried turbo, aggressive, optimal and expert.
I found that the Ram lasted longest at the "optimal" timings, "turbo" would boot but would not last long
Raising vdimm seemed to allow the ram to last a little bit longer but did not seem like a solution. The inside of my case is extremely cold and the heat jackets do not even get warm (Edit: infact, they are cold) to the touch. I've tried voltages upto 3.1v so far. I will try 3.2 or 3.3 if I get an okay on that.
I tried some Expert timings too:
I tried 11-2-2-2-2.5-13-15-*e-a-f*-CPC on
I tried 8-2-2-2-2.5 etc...
I tried 11-3-3-2-2.5 etc...
I tried 8-3-3-2-2.5 etc....
I also tried the extended beta timings at 2 -(all the threes)- 4 and Auto
Edit: I also tried all the above at Cas 3
At stock speed and turbo settings it seems fairly stable in windows and can run 3d for about 10-20 minutes before locking up or falling back to desktop. At higher speeds, system errors become more apparent and further still give hardlocks or re-starts. At 233fsb (x10) with 8-2-2-2-2.5 (3.0v iirc) I even got a crash while running the Sandra memory benchmark.
Up until now I've been running in slots 2 & 3, I'll try 1 & 3 today.
I have moved recently and was worried that my rig was throwing a wobbler at being disturbed and moved across country. So I threw my known good sticks of CH-5 3500 EL in and BAM! everything is fine. I know that I had issues with BH-5 LE 3500 on my Dfi, but didn't a few other people? :p:
again, sorry for ranting last night. I'd had a pretty tough day and I was kind of drained after the anticipation of this EBs performance left me disappointed. I'm going to try some more stuff today and see how I get on.
Edit: Just did some more testing and found a slight improvement using slots 1 & 3 (strange - the EL only likes 2 & 3) and found a bigger improvement using single channel.
I tried running prime in DC and got a rounding error almost immediatley.
What I did find though is that using a multi of 6 seemed to give a big improvement. This is odd as I can produce stable benches at 233 x 12 with the EL repeatedley. To doubly make sure my house is in order I am going to reseat my heatsink with some fresh AS5, just in case It got disturbed on the ride home. While I'm at it I'll do the 200fsb L12 mod as RGone at Amdmb found this brought stability with high bus speeds and multipliers.
Can someone move this to the correct forum? I posted it in the wrong one in my frustration last night.
hope to see all go well cactus.
funny but I had the reverse problem last night. Put together a system with an A64 and K8Npro and none of my 2x512's of ram would let me overclock until I put the EB in. Must be something with the 512 bh-5 sticks maybe. Anyway works like a charm now with the 2x512 of pc3200 EB!
Cactus, what worries me a bit are some of the issues you are seeing with respect to a single DIMM being more stable and the position of the DIMMs making a big difference as well as that you are getting better results with the lower multiplier.
To me this sounds a bit like a controller issue, either hardware related or else cooling / voltage. I know that if I increase the chipset or AGP voltage on my box, it gets very unstable immediately.
There is one difference between the CH5 and the other chips, which is the tRCD that you can only run at 3 on those (you can actually set the write tRCD to 2). tRCD is a bit of a funky latency in that the controller sometimes does not differentiate between the command latency and the RCD and throws both of them in the same pot (with respect to when the chipset expects the data back from the memory). Another issue could be the BIOS version you are running.
They have messed around with the BIOS so much that nobody really knows what is happening anymore. One example is the reduced command rate (CPC) that also impacts the arbitration between the memory channels and I've been talking to Oscar quite a bit on these issues and what other latencies they had to increase internally to get it to work ... Among those was the RCD (again)
In any case and FYI, I am playing with mine again and what can I say (everything at auto voltage except for memory at 2.8V)
http://www.lostcircuits.com/memory/sanded/cpuz.gif
OCZ I AM REALLY SORRY FOR DISSING YOU
I have just made a significant discovery with my hardware combination and this RAM.
This afternoon I resat my sink with some fresh paste and while I was at it, I did the 200fsb wire trick.
As I type (on my Laptop), my system is looping 3Dmark at 11x233 with 8-2-2-2-2.5-13-15 timings with CPC on in DC.
It seems that the L12 trick is crucial for the mobile processors with an NF2 and this new ram which is bizzare considering I was able to get my CH-5 so high with the unmodded chip.
I sincerely apologise for jumping to conclusions, I've been under a fair amount of stress lately with moving home and changing jobs etc. I souldn't have blown off like I did last night. I do however think that I've made a significant discovery with respect to this hardware combination and I hope it helps all you OCZguys and DFI/mobilers/NF2 owners in the future.
Now to go break some records!:banana:
:thumbsup:
it's amazing what a little 3cm piece of phone wire in your cpu socket will do :D
I've had the same problem with my mobile. Until I did the L12 mod I couldn't run higher than 225fsb. Now I am trying to find the ceiling. Seems a little conductive paint works wonders.
interesting.
Hmmm.... this is weard.. Do you think L12 helps on Dfi LanParty with mobile cpu?. I have not seen that problem with my mobile xp2400+ 35W. I runing @ 258.9fsb 24/7without problems and i have runed @ 280fsb+ without problem and i have not L12 modded my cpu...Quote:
Originally posted by Cactus Jack
OCZ I AM REALLY SORRY FOR DISSING YOU
I have just made a significant discovery with my hardware combination and this RAM.
This afternoon I resat my sink with some fresh paste and while I was at it, I did the 200fsb wire trick.
As I type (on my Laptop), my system is looping 3Dmark at 11x233 with 8-2-2-2-2.5-13-15 timings with CPC on in DC.
It seems that the L12 trick is crucial for the mobile processors with an NF2 and this new ram which is bizzare considering I was able to get my CH-5 so high with the unmodded chip.
I sincerely apologise for jumping to conclusions, I've been under a fair amount of stress lately with moving home and changing jobs etc. I souldn't have blown off like I did last night. I do however think that I've made a significant discovery with respect to this hardware combination and I hope it helps all you OCZguys and DFI/mobilers/NF2 owners in the future.
Now to go break some records!:banana:
One more thing do you have 2x512 PC3200 EB or?, Use this setting in bios (Cpu Interface:Optimal) if your usb are looking,sound and more. And try 3.3vdimm and see how high you can go. And try 8-2-2-2-2.5-9-12 and 11-2-2-2-2.5-9-12
All is good Cactus Jack. Thanks for the info. I will buy a mobile barton and check this out first hand.
Tha main thing is that you are working and working well. I used memtest86+ to set up my RAM timings. Watch the memory bandwidth results. Change your tmings to 2.5-2-2-10 / 12-14 and see if your bandwidth is higher than wfat you have. Bigtoe and I found different settings that worked best on the DFI NFII B, so experiment with 8,9,10,11 and 11-13, 12-14, 13-15 (always 2 numbers apart), You will find a combo that will have "fine tuned" your EB RAM to it's peak.
Have fun dude!
Ah yes, keep a fan on the RAM, with mucho volts the heat spreaders really draw off the heat.
LV, you aren't using EB (mine is 3500 btw). When I was using EL3500 I was able to go upto 233 without any mods. It was only when I tried the EB and got totally frustrated that I ended up doing the wire trick which solved my problems.Quote:
Originally posted by Lastviking
Hmmm.... this is weard.. Do you think L12 helps on Dfi LanParty with mobile cpu?. I have not seen that problem with my mobile xp2400+ 35W. I runing @ 258.9fsb 24/7without problems and i have runed @ 280fsb+ without problem and i have not L12 modded my cpu...
One more thing do you have 2x512 PC3200 EB or?, Use this setting in bios (Cpu Interface:Optimal) if your usb are looking,sound and more. And try 3.3vdimm and see how high you can go. And try 8-2-2-2-2.5-9-12 and 11-2-2-2-2.5-9-12
I tried doing some runs at 243, but it didn't quite hold together. Does this ram benefit from burning in?
Andy, the heat jackets are cold inside my case, it helps a lot when the air temperature is at 4 degrees C inside my case.;)
I'm going to try a couple of different bioses (using the 4/29 right now and see what that gets me.
I know what memmory i´m using lol.. But i cant belive why the L12 modd make a difrents on sticks to sticks "bh5 to EB". But maybe we will figure it out know :)Quote:
Originally posted by Cactus Jack
LV, you aren't using EB (mine is 3500 btw). When I was using EL3500 I was able to go upto 233 without any mods. It was only when I tried the EB and got totally frustrated that I ended up doing the wire trick which solved my problems.
I tried doing some runs at 243, but it didn't quite hold together. Does this ram benefit from burning in?
Andy, the heat jackets are cold inside my case, it helps a lot when the air temperature is at 4 degrees C inside my case.;)
I'm going to try a couple of different bioses (using the 4/29 right now and see what that gets me.
I don't know how it worked either, but it did. ;)
It is weird how i didn't have this issue with my CH-5.
Ok Jack, glad to see it's working better today, Yes 4c in your case is a good thing.
The bios issues are manifold as you know. Some of the bios' Oskar created won't run test 5 in memtest86 at all. I would try the 12/31 beta. I had the best luck with that one overall. I reviewed my NFII B using that one and never had one issue. I have not tested the EB yet with that bios. I am not at the facility with my DFI right now, but will be by next weekend. Let us know how the bios search goes, and what works best for you.
You might have said, but is this a 1gb kit of EB? Also is your CH5 a 512mb kit?
Let us know what sizes they are, as this might help all of us out.
Thanks
The EL 3500 is a 2 x 512 kit, that clocks quite nice for CH-5. I initially bought a 2 x 256 kit of LE 3500 for benching purposes when I ran into the well documented compatability issues of late BH-5 and late NF2s. Tony sorted me out with an RMA for a 2 x 256 kit of EB and I guess you can follow the rest, lol. I'll give the 12/31 a try in due course. I'll need to try another bios as there is still some instability at above 243 fsb, unless I resort to looser timings. I'll get some more testing done tomorrow, did any of you OCZ blokeys try burning this stuff in for any gains?Quote:
Originally posted by andyOCZ
Ok Jack, glad to see it's working better today, Yes 4c in your case is a good thing.
The bios issues are manifold as you know. Some of the bios' Oskar created won't run test 5 in memtest86 at all. I would try the 12/31 beta. I had the best luck with that one overall. I reviewed my NFII B using that one and never had one issue. I have not tested the EB yet with that bios. I am not at the facility with my DFI right now, but will be by next weekend. Let us know how the bios search goes, and what works best for you.
You might have said, but is this a 1gb kit of EB? Also is your CH5 a 512mb kit?
Let us know what sizes they are, as this might help all of us out.
Thanks
I'd like this Ram to get close to 250fsb while still holding reasonable timings. I'd need to get that high to see any benefit.
Hey pal :DQuote:
Originally posted by Lastviking
I know what memmory i´m using lol.. But i cant belive why the L12 modd make a difrents on sticks to sticks "bh5 to EB". But maybe we will figure it out know :)
The L12 mod gained me stability with my 1700+ tbred too... at least at 9x262 :D
I had random reboots without it and sometimes could not even boot into windows...
Update:
I'm still trying to find my stable ram overclock, but at least this time I can see when things are failing rather than just having to deal with a totally ballsed up system.
I managed a run at 11.5 x 243 at 10-2-2-2-2.5-12-14 on the 12/31 bios but it seems to kill off some of the 3d performance. I was going to flash to the 1/21 bios and then try the 2/15 but I think my hard drive has got corrupted, so I'm doing a repair right now and will try some more later.
I managed to get to desktop at 250 using 10-2-2-2-2.5-12-14 (using *cough* *cough* volts) and clockgend to 253. I managed a couple of test of 3Dmark but unfortunatley it froze.
I'll be glad when I find the sweet spot, but it'll require a fsb/multi combo that will allow me to take advantage of the 2.8Ghz my moby can run at. So untill then it's all juggling apples and oranges til then.
p.s. I started a thread informing peeps of what I found at Amdmb in the DFI forum, there has already been one person come forward experiencing the symptoms I described.
cactus, i too am experiencing those same problems..
but im using a normal XP2500, NF7, and 2x512
217 2.5-2-2-8 3v reboots during the second run of sandra mem
217 2.5-3-2-8 3v failed prime blend after 5mins
i am hoping there is an initial period where these sticks like to just stuff up completely, then once this has passed they perform like they should
I don't think that was the case with me. Before I did the 200fsb wire trick, these sticks were unworkable at any reasonable speed or any reasonable multi. After the wire mod I was looping 3Dmark at 233x11 for well over an hour.
I can only hypothesise that the patented "EB" system must alter the way that the sticks communicate with the memory controller and we all know that Nforce2 chipsets have some quirks. I'm sure that the OCZ blokeys will be able to give us a more complete answer in due course.
I have done extensive testing as I have stated on the DFI NFII B and The NF7-s 2.0 and have a few insights. I have been using a 1gb, 2 x 512mb kit of PC3200 EB RAM exclusively. All of the below tests were at 2.5-2-2-10 or 11 timings. Your timings and speeds may vary due to your board and RAM.
1) DFI: Using a standard unlocked Barton 2500+, I was able to get to 240 FSB and higher stable in Windows running Prime 95 and 3dMark 2001 in a loop for 7 hours each. Due to a known 2-17 Alpha bios bug, I was unable to complete memtest86+ test 5. It just froze up every time after about 15 seconds. Supposedly the newest bios fixes that. With the latest official bios I hit 237 fsb stable with great performance. To hit over 240fsb can require in excess of 3.0v, so make sure you have a fan blowing on the modules as they will get hot.
NF7-S 2.0: Make sure to replace the "cooler" on the northbrigdge chip. Mine failed after 2 hours of use on a new board. Before replacing it with a standard chipset HS and fan, I wa stuck at 218mhz with the PC3200 EB RAM. After the replacement and fan addition I was able to hit 223fsb. I was using the latest bios and will experiment more with the Tic-Tac and possibly another modded bios. This was on a completely stock board, other than the NB cooler.
3) DFI Ultra Infinity and Mobile Barton CPU: I just ordered this setup and expect it by Tuesday. I will be testing to prove or disprove Catus Jacks experience with the "wire mod" the changes the CPU to a 166fsb part. This is commonly used on the ABIT boards, to aid in overclocking and is generally not needed for the DFI's. Stay tuned......
Andy, cut right to the chase on the NF7-S..... use the original D10 mod'd by Equito/MikeyPage to have the 4247 Sata bios in it.
This bios I cant hit 250Mhz 11,2,2,2 @ 3.3Vdimm on CH5 and be 2D stable, and 240Mhz 11,2,3,2 @ 3.3Vdimm and be 3D stable. All with 2600+Mobile.
Anyone going for high OC on NF7-S will likely be using D10 at this point too, so please add it to your test regime. Im looking forward to your results, our hunt for Red BH5 has not worked out yet, so the 3200EB may be the solution after all. Thanks!
Quote:
Originally posted by andyOCZ
I have done extensive testing as I have stated on the DFI NFII B and The NF7-s 2.0 and have a few insights. I have been using a 1gb, 2 x 512mb kit of PC3200 EB RAM exclusively. All of the below tests were at 2.5-2-2-10 or 11 timings. Your timings and speeds may vary due to your board and RAM.
1) DFI: Using a standard unlocked Barton 2500+, I was able to get to 240 FSB and higher stable in Windows running Prime 95 and 3dMark 2001 in a loop for 7 hours each. Due to a known 2-17 Alpha bios bug, I was unable to complete memtest86+ test 5. It just froze up every time after about 15 seconds. Supposedly the newest bios fixes that. With the latest official bios I hit 237 fsb stable with great performance. To hit over 240fsb can require in excess of 3.0v, so make sure you have a fan blowing on the modules as they will get hot.
NF7-S 2.0: Make sure to replace the "cooler" on the northbrigdge chip. Mine failed after 2 hours of use on a new board. Before replacing it with a standard chipset HS and fan, I wa stuck at
218mhz with the PC3200 EB RAM. After the replacement and fan addition I was able to hit 223fsb. I was using the latest bios and will experiment more with the Tic-Tac and possibly another modded bios. This was on a completely stock board, other than the NB cooler.
3) DFI Ultra Infinity and Mobile Barton CPU: I just ordered this setup and expect it by Tuesday. I will be testing to prove or disprove Catus Jacks experience with the "wire mod" the changes the CPU to a 166fsb part. This is commonly used on the ABIT boards, to aid in overclocking and is generally not needed for the DFI's. Stay tuned......
Andy, what timings where you running and what slots were you using? I seem to have taken two steps back, now I can't run stable beyond 236. This was while trying to find my highest stable fsb while using a multiplier of 9.5 and stock clocks on my vid card. At 236, I was getting falls back to desktop when running 3D and I needed 3.0volts to run 233 @ 8-2-2-2-2.5-10-12 CPC on. Running again with the memory timings defaulted to "Turbo" did not help. It is still better than before I did the 200fsb mod, though.
I think I may need to download memtest. Is there a particular version you recommend?
Cactus Jack, I spoke to bigtoe and he says when you flash the newer bios be cautious and use a BIOS Savior or have a spare bios chip handy as many have said the 4/29 is a bios killer. He also says you need to udate the bootblock area of your bios. He feels there may be an issue with flashing and not updating the entire bios. Clear you CMOS and load setup defaults and then flash, using the switch "awdflash xxxxx.bin /WB /CC"
I was using slots 1 and 3 and as I as you and I discussed by IM the bios version is critical. The bios issue will take some experimenting as there are a dozen or so.
I orderd a DFI Ultra Infinity and a 2600+ Mobile last night, so I should have them by tomorrow. I wanted to duplicae your setup (less the Xtreme cooling).
i just passed 8hours of memtest test 5 @ 217 2.5-3-2-8 2.85v
btw im using tictac's 22 bios, with it ive gotten mushkin 2x256 3500 to 234 2.5-3-3-7 3v
with my stick of kingston 256 BH5 memtest did std test for 12 hours @ 230 2-2-2-6 3v, i couldn't get prime blend stable @ 223 2-2-2-6 3v, but 223 2.5-3-3-7 was fine
is it possible for the mobo to not like tight timings?
try 2.5-2-2
HAYUP, are you running our EB RAM in a 2 x 512mb 1gb kit? Or a 2 x 256mb 512mb kit? I have been doing some tseting with the Tic Tac 22 Turbo and official bios 23. I was able to hit 223mhz with the bios 23 on an unmodded boatrd with the PC3200 EB 1 gb kit, and I was back to about 218mhz with 22 Turbo or Tic Tac 23. Try some better cooling on your northbridge chip. I have a feeling the tighter timings and 1 gb of memory create more heat.
You might try a little more voltage, but have a fan on the RAM as it will get warm.
im using 2x512 3500EB
changing NB cooling might be a prob cos my heatpipe kinda hangs over the NB
OK, well the 1gb kits runs slower in this board, I don't have a 512mb kit of EB to try ATM. I experimented with cooling the NB chip more than I had been and it helped very little (anything is better than the stock fan that froze up). I did have better results with the official ABIT bios 23.
This memory does work very well given that 512mb modules usually don't run in this board worth a hoot. :)
Cactus - Tried 3-2-2-2-x or 3-3-2-2-x or 3-2-3-2-x or 2.5-x-x-2-x since the wiremod? :D
Yeah, I tried Cas 3 and it got me as high as 250 fsb. But I still wasn't stable. I'm going to be trying some other stuff at some point today. I had a conversation with Andy and Tony last night which gave me a few things to look into.Quote:
Originally posted by STEvil
Cactus - Tried 3-2-2-2-x or 3-3-2-2-x or 3-2-3-2-x or 2.5-x-x-2-x since the wiremod? :D
I'm using the official NF7 d23 bios now, @ 217 2.5-2-2-11 3v i was able to complete 3 consecutive sandra mem runs, which had my hopes up
but prime blend still failed after 5 seconds :(
The NF7S isn't to good it seems, i know Andy has hit 220ish but not much more.Same ram on the DFI is awesome so its not the modules.
Also i had a play with oskars latest 4/29 bios and i must say its damn fast..well worth the flash.Also Oskar has OCZ ram now so he may be fine tuning a little..(I hope so :))
That's good to know :).Quote:
Originally posted by bigtoe
.....Also Oskar has OCZ ram now so he may be fine tuning a little..(I hope so :))
I went a little bit drastic today. My system is still not playing ball, though I did get upto 243 again though by no means 24/7 stable. So I've done a full stripdown, lapped the Nbridge slightly (perfectly flat now), sinked everything, put a nice big Zalman Nbridge sink on the Sb and should be recieving a nice new SP-97, a Neo Backplate for my CPU and a NB1 to stick on the Nbridge.
The down side is that I've found a tiny, tiny, tiny nick on my CPU :(. Now I don't know how long it's been there, but I do know my CPU was working when I shut it down. I hope I haven't knackered it because I was pretty damn lucky to have a 2.8Ghz chip.
Update tomorrow after my rebuild.
Just got a knock on my door. I know have my 2600+ Mobile and a Tawain DFI Ultra Infinty for testing. Gonna load bios 4/29 and "have a go" with the PC3200 EB. I'll keep the thread updated
I am running 2x512 of pc3200 EB in my K8Npro right now at 253x10 2.5-2-2-11. I am at 3.3v vddr but only because my vdimm is modded to my +3.3v rail. This ram feels like it would do this much lower. Gonna drop the multi and see how high we can go.
i got desperate last night and tried 220 2.5-3-2-8 3v in single channel... it failed prime blend after 12mins
I just finished some extensive ABIT NF7-S 2.0 testing. The brand new board I have will not run 2 x 512mb BH5 RAM any faster than it will run the PC3200 EB. I feel that the chipset and/or bios on these boards is holding back the overclocking with the 1gb (2 x 512mb) kits, no matter what RAM type they are. I was able to hit 217fsb in dual channel mode with ZERO errors. This was using the Tic Tac D22 and Official D23 bios.
The ABIT board is being held back by the 2.9v DDR and 1.7v Chipset maximum settings. The Nforce 2 chipset is very agressive and needs the voltages and a bios that will allow the higher FSB settings.
The same RAM in a DFI Ultra Infinity NF2 board with the latest official bios will run 237fsb stable with ZERO errors in memtest using the advanced voltage settings.
Each board will be different as the NF7's seem to vary quite a bit. The good news is that the 1gb kit will work and overclock better on some boards than others.
With a Barton Mobile CPU the 166fsb "wire mod" IS required to be stable at 200fsb or over.
I hope this info helps.
have you tried prime blend on the NF7?
Has anyone tried an Epox 8rda+/3+ yet?
Hayup, I am using memtest86+ to verify memory and CPU stability. As far as Windows goes you may have issues there with your install (don't we all at some time). Using memtest and getting zero errors in a loop of test 5 and then running the standard test is the best indicator of memory stability.
Many customers have "memory" issues that miraculously clear up after a fresh Windows install.
What's an Epox? I had my shots.
I no longer have access to an Epox, but the one I had, the 8rda+ would not EVER go over 215fsb with any RAM. I suspect that with higher voltages available on the Epox boards the results would be as good or better than the ABIT. So far all NF2 boards are liking the EB RAM.
Did you try it to see if it made any improvement at all (maybe move the highs higher?):D
Thanks MikeE
....apart from mine, that is, j/k.Quote:
Originally posted by andyOCZ
What's an Epox? I had my shots.
I no longer have access to an Epox, but the one I had, the 8rda+ would not EVER go over 215fsb with any RAM. I suspect that with higher voltages available on the Epox boards the results would be as good or better than the ABIT. So far all NF2 boards are liking the EB RAM.
Getting this ram to work has become quite expensive. I've forked out for several lumps of aluminium and copper and in the process of refitting my system I seem to have killed my 2.8Ghz mobile, so now I'm having the use the 2.7Ghz one instead.
Still, I did manage a run at 240fsb x 11 @ 9-3-3-2-3-11-13 but without the extra performance of my faster CPU, I've lost 1000 marks. (I bought this ram solely for benching)
The 200fsb wire has popped out as well, so now I've got to remove my CPU again!!.
A tip for SP-97/NB-1/DFI users. The Clips of the '97 and the NB-1 will foul. It is better to glue the Nb fan to the NB-1 to prevent fouling.
On the DFI...10-2-2-2-2.5-12-14 seems best with EB.If you are on the 4/29 bios with the 4200EL 8-4-3-3-2.5-12-1-4 is nice with cpc enabled on both.EB to 250 on my board and 4200EL at 263.
Oskar worked some magic with the 4/29 and i strongly suggest you flash it if using the Infinity or LP rev B.
If you EB is not quite as fast as mine 10-3-2-2-2.5-12-14 is pretty good with 10-3-3-3-2-2.5-12-14 being a little slacker.All the other added teaks set to auto.
CPC with DS dimms is a much better option on the DFI now, at 258fsb i did see over 2000MB/s in memtest and would imagine well over 4000 bugffered in sandra.
The NF7-S and mobile combo works well with the EB memory. You mst likely will need the wiremod. With Tic Tac's d23-3Dfire I have been able to get to 233 2.5-2-3-11 and 229 2.5-2-2-11. What is really interesting is that I can get the same results with much less voltage on an NF3 board. This stuff just drinks voltage, the more the better. Just make sure you keep it cool it runs really hot.
I have a DFI on the way I'll be able to test more when it arrives. I'm going to go voltmod my AN50R then I will report back again ;)
It seems dedicated employees and overclockers have really turned OCZ around to the point that I may well try a stick. The previous OCZ was craptacular. I ordered some PC2400 when 2100 was the max in the old KT266A days and they sent me generic 2100 that barely held tight timings at PC2100. Then after a week the RAM died and killed my board. I called OCZ and they gave me a major runaround. I got screwed pretty badly and vowed I would never again own OCZ but you guys on the forum here are changing my mind slowly but surely. I'm glad that OCZ has been able to turn itself around. This post is in no way meant to be inflamatory rather a bit of praise. The fact is OCZ was one of the biggest scams on the net a few years back and now it seems to have really turned around. I don't know why exactly the change but I can say this I'd suspect it has something to do with people like ryanpgroovy and AndyOCZ. Good job guys.
NForce loves chipset voltage Cactus, I'd push at least 1.9v to the NorthBridge. Also 3.1 v to the RAM should give you a bit of stability. I'm going to reiterate that memtestx86 is a must. If you can pass about 7 loops you're rock solid. Memtest has even shown me a faulty power supply before and dipping RAM voltages on some boards. That program is an absolute GEM and a must have for o/c'ing.
Man you guys at OCZ give some kick ass support good job fellas.
Hey Catus man,
Let me see, the RAM cost you a CPU? How do it do dat? I think possibly it was your move, not the RAM. ;) Sorry though that it is muffed.
I'm still testing my 2600+ mobile and the 4/29 bios. The CPU is fast. I have to remove the wire mod now and see the effect that has.
Keep in touch.
Quote:
Originally posted by andyOCZ
Hey Catus man,
Let me see, the RAM cost you a CPU? How do it do dat? I think possibly it was your move, not the RAM. ;) Sorry though that it is muffed.
I'm still testing my 2600+ mobile and the 4/29 bios. The CPU is fast. I have to remove the wire mod now and see the effect that has.
Keep in touch.
No the Ram didn't cost me a CPU directly. However, if you recall I was trying to improve my Chipset cooling and CPU cooling following my conversation with you and Tony and I guess I lucked out by removing my CPU once too often. $hit happens and I'm not going to start screaming OMGWTF OCZ KILLED MY CPU. It was just a consequence of making sure my house was in order first.
I'm going to have to admit defeat with this Ram, though. After getting a successful bench at 240 with 11-3-3-2-3-13-15 timings I tried to raise the fsb to 243. When I rebooted I was greeted with the memory failure siren and had to clear CMOS, which means removing my graphics card on my system because the jumper is obscured by my Sbridge cooler and my Vidcard cooler.
This happened a third time, so this time I switched off, pulled the EB out and stuck my EL in. Without clearing the Cmos, it booted no problem, set the timings back to 11-2-2-2-2-13-15 and then booted into windows, not 3Dstable obviously. So I set it back to 233 with the same timings (pity I can no longer use my 12 multiplier at that speed) Ran 3Dmark as well as it has ever done.
Then I set the timings to 11-3-3-3-2.5-13-15 and booted up at 245 and ran 3Dmark again (got a lower score due to the slacker timings but my EL is quite nice at 233). So I don't think my chipset is at fault, either. It looks like I just lucked out and got some EB that will do exactly what it says on the box but not much more. I guess I'll have to cut my losses and see if I can sell this Ram on, I need to save up for a couple of mobile processors now to see if I can find another good one:( .
Cactus if you want a rplacement just PM tony ;-)
I think I will, but tbh. I don't think I want to try EB again. I've seen some people getting good results with the BH-6 and the late revisions of CH-5 seem to do well too. I really don't know what to try now. :(
We will pre test 3500EB on the DFI if you need it. If this pre tested ram doesn't clock the same on your hardware then you need to look at the board etc.
My email is in my sig
well my 2x512 3500EB will not do ras-to-cas 2
i tried memtest86 3.1 test 5 @ 217 2.5-2-2-8 and 11 and it errored out big time on the first run, and this was with 3v
i had high hopes for this ram, but im left very disappointed
i think i might have to try some 3500EL and see how that goes, thats CH5 right?
The 3500 EB is not rated at 217fsb 2.5-2-2-8. It is rated 217fsb 2.5-3-2-8. I know you have an ABIT NF7 of some type. We have had mixed results with the 2 x 512mb kits on the ABIT NF2 boards. They will all hit 200fsb or better. Sorry it didn't work out for you. You could buy a DFI NFIIB or Infinity and make better us of the RAM.
ya can always send them up here to me for some in-depth testing to confirm your findings on my P4/A64 boards hehe :)Quote:
Originally posted by HAYUP
well my 2x512 3500EB will not do ras-to-cas 2
i tried memtest86 3.1 test 5 @ 217 2.5-2-2-8 and 11 and it errored out big time on the first run, and this was with 3v
i had high hopes for this ram, but im left very disappointed
i think i might have to try some 3500EL and see how that goes, thats CH5 right?
actually I'd really like to find out how they go in A64Quote:
Originally posted by eva2000
ya can always send them up here to me for some in-depth testing to confirm your findings on my P4/A64 boards hehe :)
but I'd need some decent ram do use in the mean time ;)
what board you using hayup ?
:D
Eb is new and were all still learning the tricks to get it running the best
my 512MB XMS3200C2v1.1 sticks not appealing in OCAU forsale forum ? :)Quote:
Originally posted by HAYUP
actually I'd really like to find out how they go in A64
but I'd need some decent ram do use in the mean time ;)
I'm using NF7 non-S Rev2
eva, your mushkin sticks are more appealing :)
just did a bit more testing with my 2x512 3500EB...
220 2.5-3-2-8 3v wouldnt prime stable, but passed 18hours of memtest
i tried the sticks separately and i've found out that one of the sticks won't do ras-to-cas 2, but the other stick can, and passes memtest and a lil bit of prime so far
Hayup, try all the memory slots. Sometimes 2 and 3 work better than 1 and 3.
i have 2*256 of 3700 EB in the mail to me.. waht kind of responce should i expect form these? i havtn had time to read the threads.
are they voltage hungery like bh5?
from what i hear they do high speed at low voltage with medeocre timings and perform the same as bh5 as those levels. waht should i expect. im used to bh5 so idunno waht other ram woudl be like.
they love voltage
are they similarly tolerant of high votlage liek BH5?
i have up to 3.7v availible, will this kill them (i wont run 3.7 allteh time)
basically how high is safe?
and waht kind of fsb are most getting at waht voltage?
It looks like 3.2 volts is about all that really helps. I have hit 240fsb completely stable at 3.1v on my unmodded DFI NFII B with the PC3500 EB. Faster on the A64.
have people been getting 265+ on nforce2?
im really hoping for that kinda speed, otherwise ill have to find better bh5 :(
With board differences all other factors, the chances of you hitting 265 fsb reliably with CPC on with decent timings are slim even with BH5. If the EB RAM was able to hit PC4200 speeds at 2.5-3-2-8 timings we would sell it as PC4200 SUPER EL. No don't excpect our PC3700 EB to hit 265fsb in an NF2 system. It does clock well in NF2 and better in an A64 system. We can never promise 265fsb.
Your Best bet is to stick with 2 x 256mb modules for that type of overclocking (benchmarks only). Our 512mb modules run exrtremly well, but the 256mb modules will clock even better.
hmm, well as of now my board(s) are not the bottle neck, but rather my ram is crapping out at 260fsb for benching. 256 for 3dmark. im just hoping this stuff can get me to about 265fsb with equal performance as bh5. that extra bit will get me my goal of 22k 3dmarks in winXP.
as we all know finding ram that does 265fsb is not easy, but compared to bh5, similar chance of this i take it?
heh ill just shove 3.7v up its ass and see waht she gives. ill have it setup tomorow. should i do the L12 wiremod on my 2600m? i read soemthign about that helping alot. wahts the dealio?
thns for the help.
The wire mod won't hurt, but on the DFI boards I did not need it with my 2600+ mobile.
Put about 10 fans on the RAM. You will need it. ;)
thats the first thing i did: tried 1+3, 2+3, swapped the sticks around and tried againQuote:
Originally posted by andyOCZ
Hayup, try all the memory slots. Sometimes 2 and 3 work better than 1 and 3.
even tried single channel
then i tried each stick on its own and it really got interesting... this dual channel kit will do rated settings, but they certainly aren't matched sticks
i'll play around with it a bit more and see what else i can find