Why would I need a 2T command rate after upgrade to X2?

[5t3v0]

After replacing my sngle core Opteron with a X2 4200+, I have been getting BSODs & lockups all over the place. PCMark05 memory write test can guarantee a crash even at stock speed when at 1T, but after lowering to 2T, it passes! That cannot be right?

Strangely, I completed a full pass of memtest86 at 2.6GHz without a single memory error.

For info, I have replaced processor without reintalling Windows and I have installed the AMD driver, optimizer & MS Hotfix.

RAM is same as in sig. Tried 2.6v and 2.7v vdimm at 1T but no joy. Will try 2.8v but more volts on this RAM did not help stability on old setup. Relaxed timings to 3-4-4-8 defaults to no effect. Could this be a memory controller issue rather than ram?

 

[RadiclDreamer]

Does it really matter? The performance increase from 2-1T are so insignificant that there is no way that you would know the difference unless an app told you

 

[Mucker]

You may need a vCore increase. 2.6 GHz may be near (or past) your upper limit. Did you loosen up your timings also besides setting the command rate?

My X2 3800+ Manchester (E4) will pass SP2004 @ 1.45v @ 2500 MHz but needs 1.60v to pass 2620 MHz. A whole 0.15v increase for just 120 MHz. I am totally stable in both situations so I'm going with 2620. Conroe will replace it in a year anyaway......

 

[orangat]

Is it 4 sticks of ram? If so, 2T is the standard and theres nothing you can do about it.

 

[5t3v0]

I'm not really bothered from the performance perspective (just now at least anyway). My concern is that there might be a fault with the new proc.

I have raised vcore all the way up to 1.5v to see if its power which delivers 1.45 under load (this board is hopeless at delivering a constant vcore) but no joy with that. Temps get up to 60C at this vcore but nothing too alarming. Perhaps the psu is under-rated for the X2?

 

[5t3v0]

2 modules of DDR500. And problem occurs at stock 2.2GHz.

 

[Mucker]

Are you using the X2 on your Epox 9NPAJ Sli? I highly recommend this program for checking your core temps: http://www.thecoolest.zerobrains.com/forums/viewtopic.php?t=137

I am using my X2 on 9NPA+ Sli and the temps never exceed 57C on both cores (CoreTemp) using 1.60v. I also use a DFI Ultra-D and the Epox seemed very whacked with cpu temps and voltages. I believe the load temps are 10 to 15C off on the Epox. I had to disable the thermal shutdown temp in the Epox bios because even a 70C temp kickout limit was not high enough (errant onboard cpu sensor). The voltage regulation on the Epox is also poor, my guess is that the voltage is off too in a range of 0.05 to 0.10v. I've always needed to add 0.1v more than the cpu stock voltage to make things stable.

With all those negatives exposed, there are some positives. The Epox board is very stable. 2620 MHz is treating me very well, I just pay alot of attention to stability and very little attention to voltage and temperature.....works for me

 

[myocardia]

Originally posted by: orangat
Is it 4 sticks of ram? If so, 2T is the standard and theres nothing you can do about it.

 

[5t3v0]

Cheers for the link Mucker, I'll give that a try. My temps never went over 57C either on my Opty, but have hit 62C on this X2. I've noticed that the heatsink on the AC Freezer doesnt sit centrally on the cpu so one side is more exposed that the other, its only a couple of mm so not sure if significant.

The voltage delivery/sensing is definately off. I have the exact same problem as you in that I've always needed +0.1 volts to run stable on stock. In fact, if the BIOS is set to "Auto" or "Off" the reading is 1.45v at idle. Under load vcore can drop by as much as 0.5v according to the sensor. I'm sure that's where the instability problem comes from & why I have to bump up the volts to start with.

Which BIOS are you using? I installed the latest 9sli6424 BIOS just before swapping the cpus so have no idea how this performs on a known good system. Are there any reported issues with this BIOS? Should I try reverting to the previous version?

 

[akshayt]

Isn't 2T to 1T big difference?

 

[myocardia]

Originally posted by: akshayt
Isn't 2T to 1T big difference?

Actually, it makes quite a difference. It makes more of a difference than using a RAM divider makes.

 

[akshayt]

does making a ram divider make a difference?
isn't DDR 400 and DDR 400 using 5:6 ratio and oved to 240mhz fsb making ram 400, equal in performance?

Also, somebody said that his comp is faster at 250 2T than 200 1T

 

[Viditor]

If it's double sided ram, then you may have a problem...

 

[corsa]

...theres actually very little difference, heres a quote from this link.

I know a few of you will argue about my timings, or my settings, or something (some of you just like to argue). Most of you I hope will take a look at the chart, the differences, and realize that just like dividers etc in the original thread, there's very very little difference to be had between 1T and 2T.
so anyone wishing to go from 2x512 or 2x1024 to 4x512 or 4x1024, you can expect a little hit dropping to 2T, but what everyone seems to forget, is that dropping to 2T can sometimes (usually more noticeable when only running 2x dimms instead of 4x dimms as 4x dimms makes the memory controller really work harder) give you a nice big fat increase in overall memory speed ability, which can easily translate into cpu mhz

example

1T i might be limited to 275x9 @ 2475Mhz

but if i drop to 2T, i might be able to get 302x9 @ 2718Mhz

the increase in cpu mhz is going to negate ANY penalty incurred from dropping to 2T

 

[myocardia]

Wow, thank you, Corsa. You proved my point for me. It slows down the system so much, that you have to overclock even higher, just to make up for it. Of course, when I say something, I always have proof of it, since it seems that noone on the internet tells the truth (and no, I am not calling you a liar).

Here's my proof, of exactly what I said in my first post of this thread:
1) 2.7 Ghz SuperPi 1.5 Mod @ 3-3-3-7 1T-----33.891 sec.
2) 2.7 Ghz SuperPi 1.5 Mod @ 3-3-3-7 2T------35.281 sec.

So, at exactly the same processor speed, and exactly the same RAM timings, a 2T command rate is 1.39 seconds slower in SuperPi 1M. Okay, so let's see if I was right about it being roughly the same as having to use a RAM divider. Okay, no longer there, give me a couple of minutes. Okay, I'm back. Here's what happens when you run at 2.7 Ghz, with way, way lower (faster) memory timings, but using a divider: 2.7 Ghz & RAM @ 180 Mhz@2-3-2-5-----35.543 sec.

Hmm, why that's only .26 seconds different from having used 2T @ 225. I guess I did know what I was talking about. /sarcasm Now, what happens if you do both? That's right, it gets even slower.

 

[Rangoric]

Originally posted by: myocardia
Wow, thank you, Corsa. You proved my point for me. It slows down the system so much, that you have to overclock even higher, just to make up for it. Of course, when I say something, I always have proof of it, since it seems that noone on the internet tells the truth (and no, I am not calling you a liar).

Here's my proof, of exactly what I said in my first post of this thread:
1) 2.7 Ghz SuperPi 1.5 Mod @ 3-3-3-7 1T-----33.891 sec.
2) 2.7 Ghz SuperPi 1.5 Mod @ 3-3-3-7 2T------35.281 sec.

So, at exactly the same processor speed, and exactly the same RAM timings, a 2T command rate is 1.39 seconds slower in SuperPi 1M. Okay, so let's see if I was right about it being roughly the same as having to use a RAM divider. Okay, no longer there, give me a couple of minutes. Okay, I'm back. Here's what happens when you run at 2.7 Ghz, with way, way lower (faster) memory timings, but using a divider: 2.7 Ghz & RAM @ 180 Mhz@2-3-2-5-----35.543 sec.

Hmm, why that's only .26 seconds different from having used 2T @ 225. I guess I did know what I was talking about. /sarcasm Now, what happens if you do both? That's right, it gets even slower.

You do realize thats approx 3% right?
And that in Corsa's example the drop to 2T allowed the higher overclock?

Besides, AM2 doesn't support 1T right now anyways, so for some people it doesn't even matter. (Unless they put out the new revision already)

 

[GuitarDaddy]

2T does make a serious hit on ram bandwidth. But X2 chips aren't starved for bandwith, about the only place you will see a significant difference is in SuperPI and Sandra bandwidth scores. Just about anything else can't utilize the extra bandwith anyway.

 

[996GT2]

Originally posted by: Rangoric
Besides, AM2 doesn't support 1T right now anyways, so for some people it doesn't even matter. (Unless they put out the new revision already)

What do you mean, AM2 doesn't support 1T timings? I have that option in my K9N Platinum's BIOS and CPU-Z also shows my DDR2 running at 1T timings...

 

[Viditor]

Originally posted by: Rangoric

Besides, AM2 doesn't support 1T right now anyways, so for some people it doesn't even matter. (Unless they put out the new revision already)

Ummm...yes, they do.

 

[Mucker]

Originally posted by: 5t3v0

Which BIOS are you using? I installed the latest 9sli6424 BIOS just before swapping the cpus so have no idea how this performs on a known good system. Are there any reported issues with this BIOS? Should I try reverting to the previous version?

I am also using tne 9sli6424 bios, it is the best one for me. I use the G.Skill GBZX so I can't help you memory timings. Did you go over to extreme and look at G.Skill technicians timings for the GBHZ? My bet is that the vCore is too low.

With both an Opteron 146 and an X2 3800, I have needed more vCore on the Epox than the DFI. Could be bad sensors or bad voltage regulation. Although some don't, my X2 really came alive with voltage. Either way, I have had my troubles and joys with both mb's. They are finally both 100% stable......

 

[5t3v0]

Guys, lets keep on topic here please. I accept that there are differences in opinion about whether 1T vs 2T is worth worrying about. This arguement has been had many times over on AT and other forums. What I'm interested to know, is whether this is a problem other people have had with the X2. Why would double sided memory matter? This memory ran quite happily at 1T with the single core Opteron. And this is a 939 chip.

Mucker, how much can you trust one software app over another for voltage readings? The differences between cpu-z and Core Temp are massive. Cpu-z and Epox USDM read 1.5v where as CT reads only 1.35v. In the BIOS, vcore is set to +0.05v which should give about 1.4v should it not? Although reading in the bios is nearer to 1.45v. Temp readings are 3-4C lower on CT as well. This might certainly explain the intstability. I've give raising the vcore to 1.6v a bash. I was always convinced the Epox board was dodgy.

Edit. I noticed that the voltage reading in Core Temp did not change after raising another 0.05v so I guess the program doesnt yet sense it or has a bug. I am using the No Logging version. I will try the other version now.

Edit. Logging version has the same problem. The version I have downloaded is the v0.9.0.91 beta. There dont seem to be older versions on the website. AMD64 TCasemax and GCPUID both show vcore to be static at 1.35v too. Definately something dodgy here.

 

[Mucker]

The voltage that Core Temp is showing is just on-chip information about your x2, it doesn't have anything to do with the actual voltage, that is why it stays at one value. My X2 shows 1.35v, that is the stock design. Temperatures on the other hand, should be very acurate because it reads the DTS embedded in the cpu core.

The Epox utility probably gives the best actual voltage reading, afterall, Epox knows the exact sensor type being used in their board. The question is, how close to actual is it? I don't have a DMM or know where the points to take a vCore measurement are.

If the voltage is indeed correct, it proves without a doubt the superiority of the DFI voltage controller which is already well known. Epox voltage droop is pretty horrendous. I've heard that putting a spare fan blowing in the PWM area of the Epox makes the voltage stabilize a bit, apparently is heats up pretty good. Even with these quirks, the Epox is very stable. I don't care to pump more voltage than needed through a cpu, I also don't want to shell out another $100+ for another Ultra-D. I am only going to use this machine another year, then it's Conroe city......

 

[5t3v0]

I see. I thought it was a sensor reading

I think a year's about the life of this system until I can be bothered with a full replacement. In retrospect I probably should have bought a DFI LP Ultra-D instead of the X2. Might have got that Opty 144 upto 2.7-2.8 instead of the p1ss poor 2.5 it ran at on this board.

So what vcore did you settle on with the X2? What was the comparative vcore to get the same overclock on the DFI?

 

[Mucker]

I settled on 1.60v for Epox (+0.20). 1.525 on the DFI....

 

[5t3v0]

Cheers m8.

I still cant get 1T command rate to be stable even after resetting to Optimized Defaults. I've gone up to 1.55v vcore and still no joy. This setting used to be stable on the single core setup still in my sig. Should I RMA the processor?