Sign in with
Sign up | Sign in
Your question

Can memtest86+ detect CPU L1/L2 cache mem problems?

Last response: in CPUs
Share
July 23, 2008 10:20:05 AM

Hi everyone,

I have built a new system and when I ran memtest86+ on Q9450, it comes up with this:

error: mem location: 0000000, 0.0MB
from any test, seems like random. pops up at least 5 errors in 10 minutes of testing.

My question is this: Can the problem be the CPU's L1 or L2's cache?

The guy in computer store INSISTED memtest86+ ONLY test out the RAM.

SO this is what I did - I have swapped my 2x2GB A-Data CL4 memory to a completely new set (since at first I suspected that was the problem), tried using only one 2GB in different slot, the error is still present.

My system is running at stock speeds.

The only reason that lead me to this test was - when I used prime95 to put all 4 cores to 100%, the mouse lags/stutters. I have tried new installs of windows on different new harddrive also.

Could memtest indicate other problem such as:
BIOS (latest update already)
Ati 4850 graphics card/ or its memory?
Motherboard faulty?

My system:
Q9450 stock
Thermalright Ultra-120 extreme push pull /w Arctic-cooling PST fans
ASUS P5Q deluxe
A-Data 2x2GB CL4
Gigabyte Ati 4850
Cooler Master M850W power
Cooler Master 690 case
3x Noctua S12 fans & 3x Noctua P12 fans for exhast & case cooling.

Temp: CoreTemp, Idle - Load:
1. 53 - 63
2. 49 - 55
3. 48 - 55
4. 51 - 61

July 23, 2008 10:38:28 AM

I just need someone with experience to let me know it is likely to be the CPU.

The store I got it from won't help saying it could be anything in the system, and would not replace or help (they got a 7 days replacement policy too!)

He put the CPU on in the store - once he got win xp up and the mouse is moving, he said "looks fine to us"

...

RMA Intel? or really it could be anything in the system? (I already swapped new RAM..)
Related resources
July 23, 2008 12:01:27 PM

My question really was about can CPU L1&L2 cache memory can have problems, that can be detected by tools like memtest.

I do notice my temp are little higher than usual - especially with my push pull heatsink. However I do think they are in the range of normal working conditions.

Is it really WAY too high?

I use arctic silver 5 for thermal paste and have seated reseatted the heatsink and I am quite confident it is seatted properly with the right amount of paste applied.

If the temp is not right I guess that would be the problem?
July 23, 2008 12:13:31 PM

Yes they are WAY too high, you should be at around 32c with that cooler. The Ultra 120 is one of the best value coolers around. Anything sub 30c and you've got a great chip.

I use coretemp to measure my temps, what does look good though is that your sensors don't seem to be stuck.
July 23, 2008 12:54:44 PM

I woudn't go as far as to say that Ultra extreme blows. Actually, the temps should be about 20C lower (idle) and about 10C lower (load) with that cooler. Easily. Particularly at stock. Make sure it's installed correctly.
a c 126 à CPUs
July 23, 2008 1:36:24 PM

Quote:
Quote:
Yes they are WAY too high, you should be at around 32c with that cooler. The Ultra 120 is one of the best value coolers around. Anything sub 30c and you've got a great chip.


How can you make such a statement?? You dont even know the ambient temperature of his room! LOL
So your saying if his ambient room temp is 84f (like mine) that his cpu should run at 32c? ROFL
My ambient room temps is almost that high, how can you make such an ignorant statement???
My cpu idles at 50c with the case side on. 43c with it off. I have a warm computer room. I also have 2 95cfm thermaltake 120mm fans blowing out the back and one smaller one blowing in. Try taking the side of the case off during the summer, I bet your temps go way down.

The Ultra 120 is a joke. This cooler blows it away and its cheaper. TOP RATED COOLER BY THG.
http://www.newegg.com/Product/Product.aspx?Item=N82E168...

Make sure you get the retention bracket thats sold separately ($6)

Its extremely unlikely that there is anything wrong with your cpu. They are tested for cache errors before shipping.

Heres my suggestion. Make sure in your bios that your memory is running at the voltage that your memory manufacturer suggests. Sometimes they default to 1.8v and your memory requires 2.1 or 2.2v. If its undervolted you can have problems such as the one your having.


How about this? I live in AZ where it hits about 100+F and is normally 85f+ in my house at all times, probably hotter where I have 2 PCs and my Q6600 OCed runs at 32c idle and 50c load.

As for the OP, those temps are very high for a Q9450 at stock and that HSF. I would check and make sure its seated correctly. As for it lagging while you are doing Prime95, I never had the mouse lag but when you are stressing all 4 cores at 100% you normally can't do much more except let that run.

If you reseat the HSF and the temps don't come down I would try RMAing the CPU. As for the error, have you checked your memory setting in the BIOS? Make sure the timings and voltages are correct.
July 23, 2008 2:06:51 PM

I don't think memtest tests cache. However PC checks CPU test runs cache checks. Try upping up RAM voltage as MrsBytch suggested if it is not already 2.1V and You should be fine.
July 23, 2008 2:27:36 PM

Wouldn't a CPU test be more applicable since the L1/L2 cache is housed inside the CPU?
a c 126 à CPUs
July 23, 2008 2:31:36 PM

^Bingo.
July 23, 2008 2:37:04 PM

Quote:
Russki, how can you say what his cpu temp should be when you dont even know the ambient temp of the room he's in????????
So your saying if his computer is in a room thats 33c, his cpu should run cooler than the room temp??? Thats the most ignorant statement I've ever heard.


Well, the assumption is that the ambient temps are reasonable (around 75 F). I would feel really sorry for anyone in a 33C ambient environment.

By the way, I take it back. Unless he reset CoreTemp to use 95 tjmax his temps are 10C too high, and things look much more reasonable (Edit: still, high, but not OMG it's crazy high).

But to answer your question, MrsBytch (the name now seems a little more appropriate, by the way), with my ambient at about 80F, my set-up - exact same as the OP's with respect to the processor / cooler config (different fans, but that's probably not that important) and it idles at 30-32C between the cores at stock speed.

By the way, you are the ignorant one. Ultra Extreme is a different cooler from Ultra. 4 heatpipes vs. 6. Also, Thermalright IFX-14 was the top rated cooler by THG (unless my eyes deceive me; check for yourself: http://www.tomshardware.com/reviews/cpu-cooler-charts,1...). And Ultra Extreme in push pull was shown to (gasp) beat IFX-14 in push pull (Google is your friend; I am lazy at the moment).

Xigma is a decent cooler, but it has a big retention bracket and is quite loud with stock fan.

So do yourself a favor, do some research / thinking before you post, least you look mighty stupid.
July 23, 2008 3:08:21 PM

The TRUE is a fine heatsink, but the 1283 does put up better numbers. Has been tested and verified in about a dozen reviews. The heatsink market is so competitive right now that basically boils down to buy whatever you want. Most of the decent ones, TRUE and 1283 included, are within a degree of each other.

MrsBytch is right. His ambient temp could be really high in the computer room. However those temps are a touch high.

Check your memory voltage as others have suggested. Increase if needed.

It is unlikely that his CPU is bad. Intel does have solid quality control. You may need to swap some components around to see if you are having a problem. Try different memory for starters. If you can get your hands on a different processor, try that. Sometimes the only way to trouble shoot stuff like this is to use known good components, and then make a decision from those tests.

Best of luck.
a c 309 à CPUs
July 23, 2008 3:24:46 PM

Memtest will test ram. Of course, the cpu has to be functional for it to run at all. It is unknown which a-data modules the OP has, but at cl4, I suspect that it is high performance ram that needs higher voltages.

As to the temperatures, 53c. does seem a bit high. A properly mounted TRUE cooler should be able to keep the idle CPU at <15c. over ambient. If the ambient is 38c.(100f), then the cooler is doing it's job, otherwise it is probably not mounted properly, or the case does not have sufficient airflow. To test the latter, run with the side cover off, and a housefan directed at the innards.
July 23, 2008 3:47:49 PM

my friends laptop had an nvidia video card that had bad ram, hes trying to get rid of it.
July 23, 2008 3:57:44 PM

MrsBytch:
"So your saying if his ambient room temp is 84f (like mine) that his cpu should run at 32c?"

i have to agree with the ambient temp argument. Yes. smaller room with less air circulation does contribute a small rise in ambient temp. check this out, when i bring my tower to the living room the ambient temp went down by approx. 3-5c while the cores went down by 2-3c consistently.

maybe there is another factor you should look into or at least double check for. it may sound dumb but check and make sure all your fans are pointing in the direction... they supposed to be... usually the side of the fan with the guards is where the air is pushing towards... so if one of your intake fans is point in the wrong direction that may mean warm/hot air inside taking twice as long to get out... possibly.


gl
a b à CPUs
July 23, 2008 4:24:32 PM

Quote:
Quote:
How about this? I live in AZ where it hits about 100+F and is normally 85f+ in my house at all times, probably hotter where I have 2 PCs and my Q6600 OCed runs at 32c idle and 50c load.


Well its 100 here too and 85f in my computer room too. Its impossible for your cpu to run cooler than the room its in, unless your cooling with a refrigerated water cooling system. So either your a liar or your temp program is lying to you bro.

Quote:
Well, the assumption is that the ambient temps are reasonable (around 75 F). I would feel really sorry for anyone in a 33C ambient environment.


Well you must feel sorry for us then. Its summer and in FL and Arizona, it is that hot inside, WITH the A/C on buddy.

33C = 92F. So, 32C idle in an 85F room is perfectly reasonable, with a really good cooler and ventilation.
July 23, 2008 4:36:13 PM

Quote:

It is technically and physically impossible for the temp of your cpu cores to be cooler than the temp of the room the computer is in, period!!


Crap. I know it's not smart to nit-pick with someone named MrsBytch... but here goes.

http://en.wikipedia.org/wiki/Peltier_cooler

A Peltier cooler with a heat sink or waterblock can cool a chip to well below ambient temperature.
July 23, 2008 5:24:14 PM

Yeah, MrsBytch, if you're sitting in a 90 degree room with air conditioning - sorry.

Kaldor - one thing the TRUE likes is airflow. It reacts extremely well to push pull. Most reviews do not test that specific arrangement, which, IIRC, gains you about 2-3C, so that would put it in front on Xigma. Again, Xigma is a good cooler. The drawback is the loudness with the stock fan. But honestly, it is all just a big pissing contest. Any of those sinks should cool it below 42-ish at idle (adjusting for 10C difference because of the CoreTemp issue with 45nm chips).

Heck, mine right now is running 47-ish 100% load under OCCT at 3.2 (20% over stock, higher VCore, obviously).

One thing to do to help determine if the airflow is all screwed up like gaiden and geofelt suggested, is to take off the side panels (and potentially do the housefan thing per geofelt).

If that drops the temps dramatically, then a) there's not enough airflow; b) airflow direction is not good.

Also, really, OP should isolate the culprit. Relax the multi on the processor, set loose timings on the RAM, give it the minimum spec voltage, and test it at the default FSB (333). If that fails - northbridge issue.

Then tighten up memory timings to spec, test it - if there are errors now where they were not before - it's probably RAM.

Then bring the CPU to spec, test again - if you're getting errors now - it's the processor.
July 24, 2008 12:07:53 PM

"How can you make such a statement?? You dont even know the ambient temperature of his room! LOL "

Well you do have a point, I don't know what the ambient tempreature of the room is but EXPERIENCE tells me they generally high too high for many parts of the world excluding those areas right on the equator where people sit sweating in front of the computer, dripping sweat on to the keyboard. EXPERIENCE tells me they are too high because I own an overclocked 45nm cpu that is 30c at idle. EXPERIENCE in the forums also tells me that when people in very hot places complain about cpu tempreatures they make a mention of the current conditions.

Hey, I could be wrong, he may in fact be sitting in a pool of sweat out in a blistering desert somewhere, I guess if he replies he might make a mention of it.
a c 126 à CPUs
July 24, 2008 2:38:06 PM

Quote:
Quote:
How about this? I live in AZ where it hits about 100+F and is normally 85f+ in my house at all times, probably hotter where I have 2 PCs and my Q6600 OCed runs at 32c idle and 50c load.


Well its 100 here too and 85f in my computer room too. Its impossible for your cpu to run cooler than the room its in, unless your cooling with a refrigerated water cooling system. So either your a liar or your temp program is lying to you bro.

Quote:
Well, the assumption is that the ambient temps are reasonable (around 75 F). I would feel really sorry for anyone in a 33C ambient environment.


Well you must feel sorry for us then. Its summer and in FL and Arizona, it is that hot inside, WITH the A/C on buddy.


Meh I just have dry heat. You have humidity too. No wait I do too. Damn Monsoosn season.

And if Coretemp, My BIOS, Speedfan are all lying to me I must be crazy. It will get 55c when it is a bit hotter but I have never had it go above that.

Then again I have a Zalman 9700 with 5 80mm Fans (2 intake 3 out) and a 120mm side fan.
July 27, 2008 7:10:50 PM

Thanks for all the suggestions made here, much appreciated.

I have convinced the store to exchange a brand new Q9450 for me. However the memtest problem is still present. I have given up on it - since hours of prime95 testing showed no problems, nor crysis had any trouble running for an hour.

And for your information, I live in Hong Kong (south east asia approx 22N lat) but I have air conditioning running whenever I'm at home. And after turning ALL my fan speeds to full, here are the results:

Ambient Temp: 22 - 23C
ATi HD 4850 card GPU temp: 42C

IDLE (for ~5 mins), Q9450 Vcore: 0.95V
System (motherboard) Temp: 37C

Speed Fan: 44 39 41 41
Real Temp: 39 33 36 36
Core Temp: 49 44 46 46

Prime95 (for ~15 mins), Q9450 Vcore: 1.06V
System (motherboard) Temp: 40C

Speed Fan: 55 51 51 51
Real Temp: 51 46 46 46
Core Temp: 60 56 56 56


How is the temp look?

My full system:

Intel Core 2 Quad Q9450 @ stock speed 2.66GHz
Thermalright Ultra-120 Extreme, Push-Pull 120mm Arctic Cooling PWM PST
Arctic Silver 5 Thermal Compound

Asus P5Q Deluxe
A-Data PC-800+ extreme CL4-4-4-12 (running at its recommended spec of 2.1V)
Western Digital Raptor 150GB
2x Seagate 320GB

Gigabyte ATi HD4850 /w
Arctic Cooling Accelero twin turbo,
Arctic Silver 5 thermal compound

Old Creative audigy 2ZS /w gigaworks S750 THX speakers

Cooler Master M850 moduler 850W power
Cooler Master RC-690 Case
/w 1 x 120x38.5mm Artic cooling (linked PWM to CPU) rear exhaust
/w 5 x Noctua performance & silence 120mm in/out

A shot of my case:
http://picasaweb.google.com/rayleemw/SystemAssembly/pho...
Full album of WIP:
http://picasaweb.google.com/rayleemw/SystemAssembly

July 27, 2008 7:15:35 PM

Just put a winter coat over the PC.... :oops: 
a c 309 à CPUs
July 27, 2008 9:51:37 PM

At this point in time, I think your system is fine.

Memtest was written quite a number of years ago for DOS and was updated with the + version. I suspect that there is something in the new 45nm quad parts that cause a problem. Perhaps someone else has reported it.

One possible solution: On the Corsair forum, they recommend disabling legacy USB support on some Mobo's(mostly ASUS) when running any memory test . http://www.houseofhelp.com/forums/showthread.php?t=7106...

With such a good cooler and case, I think that idle temperatures should be 3-5c above ambient. Assuming that the cooler is mounted well, and is not defective, I would attribute the higher temperatures to some sort of measurement problem.

The load temperatures look good, so I would not worry about it.
July 28, 2008 12:41:10 PM

I think your most reliable core temp is core 1 @ 33c which is ok but not exactly great, not that it matters unless your concerned with maximum overclocks.
September 12, 2008 2:16:14 PM

Hi-

I'm having exactly the same problem with the same CPU -different memory approved for the board (G.SKIL); running at stock speeds

tried limiting the memory in the BIOS; swapping chips; no effect.

I also run microsoft's own memory testing utility (which limited the testing range to 3.3gig but still includes the 0000000 location) overnight and that came out clean

Given that:
a. MEMTEST86+ is old and does not recognize the chipset and CPU correctly,
b. the memory address is rather special i.e. 0000000 and that's the only one that pops up,
c. switching the chips did not change the bad memory address,
d. microsoft's test came out clean,
e. vista seems to run fine,
f. you report the same symptoms,

I am almost certain that the memory is o.k. and it's MEMTEST86+'s inability to deal with the CPU/Chipset that's causing this apparent problem

hope this increases your comfort level (your note increased mine)
September 12, 2008 3:08:12 PM

Hi anew5,

After further testing I have found out the reason behind it.

It is the USB legacy support - which the ASUS motherboard tend to use some of the memory for it or some reason, I'm no tech expert but here is how u can have zero errors.

Go into your motherboards BIOS, and disable the USB legacy support. Re run memtest.

Problem gone ;) 

September 12, 2008 4:08:39 PM

Hi rayleemw,

Yes, actually I started experimenting with that immediately after I wrote my note - I waited to complete 3 clean passes before reporting back to this forum - all clean and the test is still running on that machine.

[I guess this ASUS BIOS option uses (and then blocks out) a portion of the memory that memtest86+ tries to test; so regardless of what memtest86+ tries to write, it still reports back the original value i.e. 0000000 at location 0000000]

Many thanks to geofelt who pointed us both to the right direction, and to you for both reporting it in the first place, and following up with me too.
a b à CPUs
September 12, 2008 5:03:36 PM

For the record:

Realtemp gets the temps right, core temp does not. Intel (finally) released the TjMax values a week or two ago. Coretemp reads 10 degrees high.

Also, if there was a problem with the L2 cache, you would crash well before windows could even load. Hell, I doubt the bios screen would be stable with a busted L2 cache...
September 12, 2008 5:33:01 PM

For the record, the OP got his problem fixed. Another user has the same problem, and the OP help him.

As far as apps that read core temps... There is no such thing as one app is more accurate then the other. If you can adjust the tjmax on the program that you use the read core temps, the better, when you calibrate it correctly.

And besides that, you shouldn't rely on any app that reads core temps for idle temps. It's basically the load temps that matter most.
December 10, 2012 5:17:40 AM

Use Prime95 if you would like to test CPU cache.
!