TwoCables

Is this the same HCi Memtest that others are saying is pretty much the best?

I did indeed download it and I will certainly use it in addition to Memtest86+ and Windows Memory Diagnostic, but I'm only curious if this is the very same one that I think I keep hearing about.

__________________

PizzaMan

That's the one I linked.
http://hcidesign.com/memtest/
Here so you can see the address

__________________

____________________________
.:|Official|Volt|Mod|Squad|Member|:.

GTL testing and tweaking for all 7 series nVidia boards

DeadSkull

x2 on this choice.

I found running 4 instances with 850-900mb of memory found either ram or northbridge problems faster then Prime Large FFT.

I usually do 300% coverage on all instances for insuring basic stability before running extensive smallfft/linx tests and later on 1000%+ coverage for rock solid stability test.

__________________

Q9650 Club

TwoCables

But I said I downloaded it.

Besides, I can see the address of any link by pausing my mouse over it.

Anyway, thanks again!

__________________

-n7-

Firstly, Memtest 86+ is the one you want to use; they are almost always ahead on updates vs. the Memtest 86 (non-plus) crew.

I wouldn't be throwing it away; Memtest 86+ is still good for basic RAM testing.

What i don't find it good for is memory stressing.

The thing is, Memtest bootable only tests portions of your RAM at a time for errors...it's not very hard on a machine at all.

HCI Memtest (always try to run as many instances as you have cores/threads) is extremely good for stressing RAM harder than anything else i've ever tested.

So especially in overclocked scenarios, it will fail when Memtest86+ will pass.

I do agree that overall HCI Memtest is far more picky or, accurate, for finding errors, though it can be somewhat of a different type of errors.

PizzaMan especially,
Something also to note is that especially with large amounts of RAM, HCI puts a lot more stress on the MCH or memory controller, & this will also produce errors even when the RAM is technically fine.

Clockadile Dundee

Two Cables : If it is stable, it is stable, right? If there is no more electromigration than normal (and I am assuming Intel has some "very low but present" threshold for this), then by what process is the lifespan shortened?
(Not a rhetorical question, but not one where I would demand the answer from you either)

Are you not using Linpack? It seems to find everyone's instabilities very quickly.

If there are no more tweaks, which is hard to conceive of since using a modern platform for the first time a few months ago, you may need to either improve cooling, or add vCore, or both.

The resistance of a conductor, is determined by it's cross sectional area (width basically), length, and temperature ( assuming it's copper atoms don't go all willy-nilly with their electron valances :| ).

Since the traces and transistors are conducting electricity, and you certainly can't change the other physical factors, the only way to practically reduce resistance, is either to reduce temps, or reduce voltage, which also reduces resistance.

I think this is why you could theoretically put a LN2 pot on a cpu, and run it at say, stock clocks, at an abnormally low voltage, stably.

There is a relationship between voltage, amperage (aka current), resistance (unit: Ohm), and power (unit: watt).

Ohm's law shows that:
V=IR *OR* I=V/R *OR* R=V/I

You can figure out the voltage, just by knowing the resistance (AKA impedance), and the amperage, etc.

(I just thought a little background was necessary.)

I have only just recently tried to go this deep with PCs, so I will learn more, and then I will post what I find out.
---------------------------
I have "heard" that:

HCI Memtest
Memtest86+

One is ideal for stability testing, and one is better for errors, right? Did you know this? Which is which?

Edit: The thing about clock skews and period stability: 150ps cpu skew, and 50ps NB skew has made previously unstable OCs stable, or stable at a lower voltage, for example OCN user GigaByte tried it, and got stable with some "secondary" voltage also being lower, iirc.

I recommend everyone who is on a "quest" for seeing what is apparently magically lurking inside of that box, the force of a thousand suns, slumbering in there, try water cooling, I will help you figure it out cheaply if need be.

Yes, I said it, a thousand suns. :|

__________________

TwoCables

This is precisely why I said that I will utilize both HCi MemTest and Memtest86+ (as well as Windows Memory Diagnostic). The more thorough I am, the better off my overclock's stability is.
From what I have been taught here on OCN, all overclocks are unstable by nature. That's why it's a bad idea to overclock a server - especially a commercial/high-traffic server. That's also why it's a bad idea for a high-profile software developer to be developing on an overclocked system. That's also why Pixar, Industrial Light & Magic, and other places like that don't use overclocked systems, but rather really high-end pre-builts.

So, regardless of how stable we make our overclock, it's still unstable. And due to the very nature of what an overclock is, the lifespan of our stuff is shortened a little. After all, it's beyond specifications. But that's a part of what makes us enthusiasts!
Yep. As I mentioned earlier, it passed 100 runs using 3400 MB of memory last night (and through today). And as I've said many times in the past, I highly recommend including linpack testing in everyone's stress-testing toolboxes.
My motherboard was manufactured sometime in the year 2006. I first started using in on March 12th, 2008. It's an EVGA 680i SLI. I'm currently looking at one more "bump" before the Northbridge voltage can't go any higher in the BIOS, and I'm looking at 3 more bumps for the Southbridge voltage. For the FSB, I'm looking at one more bump. For the memory voltage: several more, but there's no way I'm increasing it anymore since it already got really hot during the LinX runs. I think I'm at the maximum level of heat. When I touched my memory during the LinX testing last night, I don't think I could have withstood keeping my finger on there any longer than 10 seconds. I think it would have hurt. That's how hot it was. So, as for not having any more tweaks, I think that's what I mean. I still have tweaks I can make, however my voltages are all really high right now even though everything is running cool.
Willy-nilly? hehehe
True. And I think I'm pretty much at this point right now. But meh.. maybe not. I don't know for sure. But that's why I'm re-testing using a slightly more-extreme version of the stress testing that I always recommend.
Absolutely. The colder it is, the more possible it is to meet some extremes.
I feel like I just tried reading Chinese.
True. Memtest86 and Memtest86+ (yes, same thing, I know) are for checking errors. Other programs, like HCi's MemTest, is apparently for stress testing as well as error-checking.

So yeah, I've always known the difference. But you're right: not everyone does. Some people seem to think that Memtest86+ is for stressing the memory.

__________________

Clockadile Dundee

I remember now that I had read about the 100 linpack runs, but I was looking at the little list of p95 tests, which had apparently wiped my memory banks.

Forget the equations, electricity works like water in a pipe:

Voltage = pressure of water

Current = flow of water

Resistance = diameter, length of pipe

The consistency of nature is curious..

Now, the thing about how OCs are inherently unstable: What causes instability if not electromigration? I am not convinced that they are not the same thing, in a sense (in the sense of signal quality, as opposed to a lack of period stability, a mis-timing of signals on the mainboard, if I understnd correctly).

I was genuinely asking about those memory programs, I have not used either yet, because I haven't had a working rig for long enough to try everything I wanted to. I have used computers for a couple decades, but I just got into OCing, and all this deeper stuff, 4 months ago; my last rig was a skt 939 emachines, and it has been broken for years.

I am a noob! I swear, I wont ask rhetorical questions, because I have too many real questions.

And now I have been having a string of RMAs; it is either the mobo, or the 920's integrated memory controller. Two of 4 boards have dropped the primary memory channel. I have a P6T7 on the way from Newegg, for "testing"...muhahaha

This is the most rewarding, enthralling, stimulating, educational, and fulfilling thing I have done, as far as getting all those things from one activity.

__________________

PizzaMan

This is way I'm testing the system tonight with a fresh pair of OCZ plats. If these 1066 plats fail at 800 then I know it's not the ram. Though, it's most liekly the ram as I have this problem with D9s all the time. They have a really high failure rate. The heat just gets to them after a while. A round in the freezer normally takes good card of them though.

__________________

____________________________
.:|Official|Volt|Mod|Squad|Member|:.

GTL testing and tweaking for all 7 series nVidia boards

-n7-

Make sure you're at stock when you're testing this.

If you're @ your sig's OC, 450 FSB could easily be slightly unstable in the NB area causing issues that look like RAM.