Sign-in / Sign-up
Your question
Solved

Corsair on Memtest86.. how many errors is bad?

Tags:
  • Computers
  • memtest
  • RAM
  • memtest86
  • Corsair
  • Memory
Last response: in Memory
January 4, 2014 11:23:13 PM

So I built this computer a couple months ago and even though I did pretty damn good on benchmarks the computer has never ran very good. I have had problems with booting, gaming, and video rendering/authoring blue screens.

I have tried upgrading, downgrading drivers and software for everything on this computer and still I have problems.

So finally I ran memtest86. I ran it just now with all 8 sticks of my Corsair DDR3 2133 Dominator Platinum and did the first 8 tests. It took nearly an hour and I came back after 8 tests with 41 errors. I have no idea how good or bad that is? Is that bad enough to cause my computer issues I'm having?

I'm using 8 sticks of 4gig each for 32g total on a Rampage IV Extreme with a 4960x and quad Titans. I also turned all my overclocking back to stock for testing. From here i'm going to test 1 stick at a time to find out which one(s) are bad. Let's assume 1 stick is bad. Can I run with 7 while I RMA that 1 bad stick or should I go to an even number "6" or down to 4?

thanks for any help

More about : corsair memtest86 errors bad

a b } Memory
January 4, 2014 11:32:07 PM

one error is bad. there should be none. memtest is the FIRST thing to run with a new build that blue screens.
m
0
l
January 4, 2014 11:41:33 PM

Any error is bad. Each error is the same as Windows throwing a BSOD. So if you reduce the number of errors that's a good thing, but ultimately you want a full pass without any.

What is the exact model of the kit you are running?
m
0
l
Related resources
January 5, 2014 12:31:17 AM

cuecuemore said:
Any error is bad. Each error is the same as Windows throwing a BSOD. So if you reduce the number of errors that's a good thing, but ultimately you want a full pass without any.

What is the exact model of the kit you are running?


thx for the reply

Update: I just ran all sticks individually for the first 8 tests on memtest86 and they all came back with ZERO errors. This is odd.
So would my next run of tests be to now test each of the 8 dimms? I'm guessing that at this point it looks like a bad dimm?

Is only running the first 8 tests on memtest good enough? When running all 8 sticks at once for the first 8 testsand I got 41 errors and they came in a few at a time
in each test. So being that I got ZERO on the individual tests scares me.

The motherboard would be the worst cast scenario as it's all watercooled and I would basically have to rebuild the whole thing :( 

My ram is Corsair Dominator Platinum 16GB 4x4 DDR3 2133MHz PC3 17000 C9 Kit CMD16GX3m4A2133C9
and I have two of these kits installed
m
0
l
January 5, 2014 1:06:15 AM

cuecuemore said:
Any error is bad. Each error is the same as Windows throwing a BSOD. So if you reduce the number of errors that's a good thing, but ultimately you want a full pass without any.

What is the exact model of the kit you are running?


not good, just ran all 8 at the same time again and got 91 errors after just 8 minutes and then it rebooted... do you suggest I test 1 stick in each dimm now to find out
if there is a bad dimm? Is there a possible cause besides a bad dimm at this point? Lack of power maybe? I have all the extra power going to the board slots for the dimms
and such.
m
0
l
January 5, 2014 1:19:10 AM

Here's the deal. Nothing is defective. The kit is only rated to run at 2133 9-11-10-30 with a *single* kit installed. Try dropping it to 1866 and stock timings with both kits installed and see if it passes.

Also, what is your RAM voltage at?
m
0
l
a b } Memory
January 5, 2014 2:14:03 AM

It's common for memory to work perfectly with just one stick in stalled but fail with multiple sticks since bus loading (capacitance) increases and BIOSes may not adequately compensate for that with slower timings. Also common is substandard memory, identifiable by gaudy heat sinking that costs less than first-rate chips.
m
0
l
January 5, 2014 2:51:03 AM

cuecuemore said:
Here's the deal. Nothing is defective. The kit is only rated to run at 2133 9-11-10-30 with a *single* kit installed. Try dropping it to 1866 and stock timings with both kits installed and see if it passes.

Also, what is your RAM voltage at?


I was running without XMP at 1333 stock and getting the same blue screens though. I didn't realize that you couldn't use two kits together like that.

For all the previous memtest runs I was using XMP 2133. However, memtest says it's "1550MHz (DDR3-3100) and "Timings: CAS 19-15-15-31" which I don't get why it shows those numbers
which is not what I see in the bios. Also, my DRAM voltage is at 1.500

So for all those benchmarks I was doing when I first set this up, was it a waste for me to run 8 sticks at 2133? Would 4 be better or a kit that is made to run 8 sticks ideal?

thanks again for the help!!! I'm going to try and take off XMP and run it lower and see if I get errors in memtest now and get the results back to ya
m
0
l
January 5, 2014 3:12:30 AM

cuecuemore said:
Here's the deal. Nothing is defective. The kit is only rated to run at 2133 9-11-10-30 with a *single* kit installed. Try dropping it to 1866 and stock timings with both kits installed and see if it passes.

Also, what is your RAM voltage at?



Ok, just bumped up the voltage and tried running 2133 but still got errors in memtest. However, I was able to keep it at 1.5v and run 1866 with no errors. I'm guessing that's how I should keep it for daily use.

I'm curious though, for overclocking and running benchmarks what's going to get me the best numbers? Going 4 sticks and hitting 2133 or maybe higher or using 8 sticks at a shaky 2133 or reliable 1866?

Or should I just look into getting a kit that is made to run 2133 for 8 sticks?

thanks again for the help!!!
m
0
l

Best solution

January 5, 2014 3:31:09 AM

I had the same problem but at smaller case
I had 2 stick of rams that did not run at 1600 MHz together so now I am running at 1333Mhz
Personally I would use lower settings, littlebit smaller benchmark numbers but stable machines
at that level of machine, I do not expect to see any kind of real life performance drop because of ram speed :) 
Share
January 5, 2014 9:42:24 AM

Nefos said:
I had the same problem but at smaller case
I had 2 stick of rams that did not run at 1600 MHz together so now I am running at 1333Mhz
Personally I would use lower settings, littlebit smaller benchmark numbers but stable machines
at that level of machine, I do not expect to see any kind of real life performance drop because of ram speed :) 



Thanks, and at this point it's a very easy solution to just run my 8 sticks at 1866. However, If I want to give benchmarking another run ex: 3dMark/Heaven and so on, what's going to give me the absolute best numbers?
A. Running my 8 sticks at 2133 as in the past but could be giving me bad numbers because of errors even without crashes
B. Go down to 4 sticks and run at 2133 or higher.
C. Find RAM that is 2133 that comes in a kit made for 8 sticks
m
0
l
January 5, 2014 10:49:29 AM

I would say the 4x4@2133 version, as benchmarks should not really use more RAM, but a faster maybe
but wait for others opinions as well, I am not really good at benchmarks :/ 
m
0
l
January 5, 2014 6:48:24 PM

I agree with Nefos. If you just want good bench numbers then a single kits at stock settings will probably be best.
m
0
l
January 5, 2014 9:37:29 PM

cuecuemore said:
I agree with Nefos. If you just want good bench numbers then a single kits at stock settings will probably be best.



thanks for the help guys :) 
m
0
l
January 5, 2014 10:43:53 PM

That's all well and good... but might I ask what good is a benchmark if it is so rittled with instability that makes you want to rip your hair out and not useable in real life... Your proving nothing by benchmarking a system that is obviously unstable.
m
0
l
January 6, 2014 8:59:14 AM

Bonecrushrr said:
That's all well and good... but might I ask what good is a benchmark if it is so rittled with instability that makes you want to rip your hair out and not useable in real life... Your proving nothing by benchmarking a system that is obviously unstable.


It's like tuning a dragster or Nascar. There is two different types of overclocks in my mind. One that is pushing the limits to see how far you can take it and the other is a stable OC made for daily use. And to switch from one to the other is a click of a button in the bios.

I built my first computer which is a super computer and I wanted to see what it could do. I ended up 6th overall in the most popular benchmark 3DMark Firestrike and number 1 in others. So for me that was a blast to try and achieve. Now I back it down a little for daily use. I don't see how that's such an issue?

It would be like buying a Ferrari and only driving in your driveway. You gotta take it out a few times and see what it can do and then you cruise it nice and slow all you want. :) 

m
0
l
January 6, 2014 7:40:35 PM

hockeytyme said:
Bonecrushrr said:
That's all well and good... but might I ask what good is a benchmark if it is so rittled with instability that makes you want to rip your hair out and not useable in real life... Your proving nothing by benchmarking a system that is obviously unstable.


It's like tuning a dragster or Nascar. There is two different types of overclocks in my mind. One that is pushing the limits to see how far you can take it and the other is a stable OC made for daily use. And to switch from one to the other is a click of a button in the bios.

I built my first computer which is a super computer and I wanted to see what it could do. I ended up 6th overall in the most popular benchmark 3DMark Firestrike and number 1 in others. So for me that was a blast to try and achieve. Now I back it down a little for daily use. I don't see how that's such an issue?

It would be like buying a Ferrari and only driving in your driveway. You gotta take it out a few times and see what it can do and then you cruise it nice and slow all you want. :) 



It is very hard to manage a 1/4 mile dragster pass if you blow a connecting rod through the side of the engine block 100 yards into each race? Different strokes for different folks I guess, In my mind an overclock is and always has been invalid unless you can manage a of minimum 24 hours Prime 95 stable similar to how a dragster has to cross the finish line in order for a pass to matter..
m
0
l
January 6, 2014 8:21:50 PM

Bonecrushrr said:
hockeytyme said:
Bonecrushrr said:
That's all well and good... but might I ask what good is a benchmark if it is so rittled with instability that makes you want to rip your hair out and not useable in real life... Your proving nothing by benchmarking a system that is obviously unstable.


It's like tuning a dragster or Nascar. There is two different types of overclocks in my mind. One that is pushing the limits to see how far you can take it and the other is a stable OC made for daily use. And to switch from one to the other is a click of a button in the bios.

I built my first computer which is a super computer and I wanted to see what it could do. I ended up 6th overall in the most popular benchmark 3DMark Firestrike and number 1 in others. So for me that was a blast to try and achieve. Now I back it down a little for daily use. I don't see how that's such an issue?

It would be like buying a Ferrari and only driving in your driveway. You gotta take it out a few times and see what it can do and then you cruise it nice and slow all you want. :) 



It is very hard to manage a 1/4 mile dragster pass if you blow a connecting rod through the side of the engine block 100 yards into each race? Different strokes for different folks I guess, In my mind an overclock is and always has been invalid unless you can manage a of minimum 24 hours Prime 95 stable similar to how a dragster has to cross the finish line in order for a pass to matter..



Haha I hear ya. I was lazy for my OC stability testing. When ur making tweaks to try and get every little edge, 24 hours of prime95 is too long. I can run a 5 min stress test on other programs. Now, once I backed it down for daily use I did as you said

m
0
l
January 6, 2014 9:49:40 PM

hockeytyme said:
Bonecrushrr said:
hockeytyme said:
Bonecrushrr said:
That's all well and good... but might I ask what good is a benchmark if it is so rittled with instability that makes you want to rip your hair out and not useable in real life... Your proving nothing by benchmarking a system that is obviously unstable.


It's like tuning a dragster or Nascar. There is two different types of overclocks in my mind. One that is pushing the limits to see how far you can take it and the other is a stable OC made for daily use. And to switch from one to the other is a click of a button in the bios.

I built my first computer which is a super computer and I wanted to see what it could do. I ended up 6th overall in the most popular benchmark 3DMark Firestrike and number 1 in others. So for me that was a blast to try and achieve. Now I back it down a little for daily use. I don't see how that's such an issue?

It would be like buying a Ferrari and only driving in your driveway. You gotta take it out a few times and see what it can do and then you cruise it nice and slow all you want. :) 



It is very hard to manage a 1/4 mile dragster pass if you blow a connecting rod through the side of the engine block 100 yards into each race? Different strokes for different folks I guess, In my mind an overclock is and always has been invalid unless you can manage a of minimum 24 hours Prime 95 stable similar to how a dragster has to cross the finish line in order for a pass to matter..



Haha I hear ya. I was lazy for my OC stability testing. When ur making tweaks to try and get every little edge, 24 hours of prime95 is too long. I can run a 5 min stress test on other programs. Now, once I backed it down for daily use I did as you said



Believe me it didn't start at 24 hours. To begin with more like 20 mins per adjustment and as I got closer and started bumping into stability issues requiring more volts etc the tests grew to an hour then 2 then 6 then 12 then 24 for the final tweaking. I have not tried OCing my i7 4770k haswell yet however my old i7 920 bloomfield made it to 3.76ghz on water cooling and never broke 50 degrees full load @ 8 thread prime 95. My goal was 3.8ghz but core 3 just wouldn't do it.
m
0
l
January 6, 2014 11:28:48 PM

Bonecrushrr said:
hockeytyme said:
Bonecrushrr said:
hockeytyme said:
Bonecrushrr said:
That's all well and good... but might I ask what good is a benchmark if it is so rittled with instability that makes you want to rip your hair out and not useable in real life... Your proving nothing by benchmarking a system that is obviously unstable.


It's like tuning a dragster or Nascar. There is two different types of overclocks in my mind. One that is pushing the limits to see how far you can take it and the other is a stable OC made for daily use. And to switch from one to the other is a click of a button in the bios.

I built my first computer which is a super computer and I wanted to see what it could do. I ended up 6th overall in the most popular benchmark 3DMark Firestrike and number 1 in others. So for me that was a blast to try and achieve. Now I back it down a little for daily use. I don't see how that's such an issue?

It would be like buying a Ferrari and only driving in your driveway. You gotta take it out a few times and see what it can do and then you cruise it nice and slow all you want. :) 



It is very hard to manage a 1/4 mile dragster pass if you blow a connecting rod through the side of the engine block 100 yards into each race? Different strokes for different folks I guess, In my mind an overclock is and always has been invalid unless you can manage a of minimum 24 hours Prime 95 stable similar to how a dragster has to cross the finish line in order for a pass to matter..



Haha I hear ya. I was lazy for my OC stability testing. When ur making tweaks to try and get every little edge, 24 hours of prime95 is too long. I can run a 5 min stress test on other programs. Now, once I backed it down for daily use I did as you said



Believe me it didn't start at 24 hours. To begin with more like 20 mins per adjustment and as I got closer and started bumping into stability issues requiring more volts etc the tests grew to an hour then 2 then 6 then 12 then 24 for the final tweaking. I have not tried OCing my i7 4770k haswell yet however my old i7 920 bloomfield made it to 3.76ghz on water cooling and never broke 50 degrees full load @ 8 thread prime 95. My goal was 3.8ghz but core 3 just wouldn't do it.


I'll say this..I had a blast building and OC'n my first computer. After awhile I grew tired of the reboots and running Benchmarks over and over to gain just a few points every so often. So I quit. It was a blast to see what I could do.

You can see my setup here... It's way overbuilt but go big or go home I guess.. lol

http://www.overclock.net/g/a/1054916/flatline-build/

I wish I could sell it for cost just so I could build another. It was a lot of fun
m
0
l
January 7, 2014 5:50:07 PM

hockeytyme said:
Bonecrushrr said:
hockeytyme said:
Bonecrushrr said:
hockeytyme said:
Bonecrushrr said:
That's all well and good... but might I ask what good is a benchmark if it is so rittled with instability that makes you want to rip your hair out and not useable in real life... Your proving nothing by benchmarking a system that is obviously unstable.


It's like tuning a dragster or Nascar. There is two different types of overclocks in my mind. One that is pushing the limits to see how far you can take it and the other is a stable OC made for daily use. And to switch from one to the other is a click of a button in the bios.

I built my first computer which is a super computer and I wanted to see what it could do. I ended up 6th overall in the most popular benchmark 3DMark Firestrike and number 1 in others. So for me that was a blast to try and achieve. Now I back it down a little for daily use. I don't see how that's such an issue?

It would be like buying a Ferrari and only driving in your driveway. You gotta take it out a few times and see what it can do and then you cruise it nice and slow all you want. :) 



It is very hard to manage a 1/4 mile dragster pass if you blow a connecting rod through the side of the engine block 100 yards into each race? Different strokes for different folks I guess, In my mind an overclock is and always has been invalid unless you can manage a of minimum 24 hours Prime 95 stable similar to how a dragster has to cross the finish line in order for a pass to matter..



Haha I hear ya. I was lazy for my OC stability testing. When ur making tweaks to try and get every little edge, 24 hours of prime95 is too long. I can run a 5 min stress test on other programs. Now, once I backed it down for daily use I did as you said



Believe me it didn't start at 24 hours. To begin with more like 20 mins per adjustment and as I got closer and started bumping into stability issues requiring more volts etc the tests grew to an hour then 2 then 6 then 12 then 24 for the final tweaking. I have not tried OCing my i7 4770k haswell yet however my old i7 920 bloomfield made it to 3.76ghz on water cooling and never broke 50 degrees full load @ 8 thread prime 95. My goal was 3.8ghz but core 3 just wouldn't do it.


I'll say this..I had a blast building and OC'n my first computer. After awhile I grew tired of the reboots and running Benchmarks over and over to gain just a few points every so often. So I quit. It was a blast to see what I could do.

You can see my setup here... It's way overbuilt but go big or go home I guess.. lol

http://www.overclock.net/g/a/1054916/flatline-build/

I wish I could sell it for cost just so I could build another. It was a lot of fun


WOW!!, That is a fantastic build. Excellent work.
m
0
l