Heat probs reported for Intel E8400s

[ANP !!!]

Is the problem only with the E8400s ??

[mnewxcv]

Is the problem only with the E8400s ??

and the new 45nm Q. ps, can u pm me a link to a bigger version of your avatar?

[MuffinFlavored]

I agree with you. So how do we go about getting accurate on die temp readings? maybe they could leave a hole in the IHS so we could use our own thermal probes.

Would this cause problems with mounting?

[Ugly n Grey]

I was thinking about the SIDE of the IHS not the top lol

[MuffinFlavored]

I was thinking about the SIDE of the IHS not the top lol

That would be wise.

I once purchased a 5.25" bay fsn controller that came with thermal probes, and I saw a great increase in temperature by putting a probe between the heatsink and processor.

Duh?

[FischOderAal]

That would be wise.

I once purchased a 5.25" bay fsn controller that came with thermal probes, and I saw a great increase in temperature by putting a probe between the heatsink and processor.

Duh?

no wonder. did the heatsink have contact to the HS at all? in generall the contact-surface is smaller and a greater amount of the heat has to go through the tim to finally reach the heatsink. the best way is always metal -> metal

[mrcape]

I'm concerned about temp readings on these cpu as mine HAS degraded over the space of a week, with 4ghz first requiring 1.3375, and now requires 1.3825, 4250mhz used to require 1.425, now requires 1.4625.


Loss of clocks happened over the space of a day..one day i primed 4ghz for 12.5 hrs, and when done, went up to 4.25ghz, again primed for 12 hrs.

Left the systems on idle, came back, it wasn't stable. Prime would not run for more than 1 minute.

I spent some time verifying the board was ok, that ram was not the cause, all parts proven stable using other components.

So, during this, cpu was under fuson block in Maximus Formula SE. I never mounted cpu with aircooler. But temps never budged from 35c CPU and 40c DTS0 and DTS1.

So, on air, temps went over 70c DTS at stock, which seems a bit high for 45nm, but oh well

This is the type of finding that is concerning! You're not the only one, and it seems priming for hours at over 1.4v is damaging the chips even if kept cool.

[Eastcoasthandle]

This proves to me that the sensors are not read correctly. I on the other hand tried to go from a 8-pin 12V to 4-pin 12V and only noticed roughly 2C difference (it fluctuates a lot and hard to get a good screen shot of it). If anything I noticed that Everest and Coretemps didn't always show the exact same temp readings as it did when I had the 8-pin 12V connection.


Both of these pics are stock using a 4-pin 12V connection. As you can see the idle temps are 40C-42C





This pic shows the E8400 stock using the 8-pin 12V connection as you can see the idle temps are roughly 4C-8C higher.


I always knew that using an 8-pin 12V connection may raise temps. However, it doesn't explain why the temps are higher then my E6850 which hovers in the early-mid 30C range.

[JohnZS]

My original 45nm QX9650 had 2 cores overheating and I had it RMA'd
You can read my story here.
To cut a long story short 2 of the cores were running 21C hotter than the other 2 cores. Any amount of load which stressed all 4 caused a system crash.
My replacement CPU does not suffer from this problem....although there is still a bit of a difference between core temperatures, it is nowhere near as bad.
I am going to investigate the 8Pin 4Pin thing Eastcoasthandle as my motherboard as a 4Pin socket but I am using half of my 8Pin adaptor (it splits), perhaps I should use the 4pin adaptor which is also included with my PSU?
John

[Hornet331]

This proves to me that the sensors are not read correctly. I on the other hand tried to go from a 8-pin 12V to 4-pin 12V and only noticed roughly 2C difference (it fluctuates a lot and hard to get a good screen shot of it). If anything I noticed that Everest and Coretemps didn't always show the exact same temp readings as it did when I had the 8-pin 12V connection.

Both of these pics are stock using a 4-pin 12V connection. As you can see the idle temps are 40C-42C


This pic shows the E8400 stock using the 8-pin 12V connection as you can see the idle temps are roughly 4C-8C higher.


I always knew that using an 8-pin 12V connection may raise temps. However, it doesn't explain why the temps are higher then my E6850 which hovers in the early-mid 30C range.

i have to try that and see if the 8pin 12V connector really causes a higher temp reading. Also anyone knows why using the 8pin supply would cause a higher temp?

It seems quite odd...

[cadaveca]

8-pin provides more available CURRENT to cpu. Depends on board design, but AFAIK, second set of phases(5-8) is driven by additional 4 pins as well. If I am wrong, I'd love some clarification.

On a side note, my cpu has further degraded. 4ghz now only stable 1.41v real(DMM).

also, dts0-1 temp difference of 16c while on air, cpu temp is always offset 5c from DTS0, no matter what cooling.

So, for my cpu, throttle will not happen until max cpu temp as specified by intel has been exceeded by +30c.

In my case, thermal reading problems are gonna kill this cpu under normal use. I won't RMA, tho, because I am too honest, but I won't buy another Intel chip, either, until I hear that there is a new core revision, or something changes, as it's obvious Intel is not concerned about these cpu's dying...even if they knew they would before they left the factory. I'd advise others to do the same, or they'll never be concerned, and I won't build a pc for anyone using 45nm chips, either.

I got bit by INtel and faulty BadAxe boards, I'll not have my name further tarnished for supplying OTHER faulty parts.

[JohnZS]

i have to try that and see if the 8pin 12V connector really causes a higher temp reading. Also anyone knows why using the 8pin supply would cause a higher temp?

It seems quite odd...

My motherboard (P5E3) only has a 4Pin ATX connector. Since my Enermax Infiniti PSU's 8pin can split into 2 4Pins, I am using one of those to power my system.

Am I better of using the proper 4Pin ATX connector which is in the box? or will it not make any difference as using half of the 8Pin adaptor would be the same?

Thanks

cadaveca

That sucks
I only RMA'd mine as I was running at stock (never O/C'd it) and was not stable at stock because it was overheating.
I am sorry to hear your having such bad issues with your CPU....this reminds me of the Northwood sudden death syndrom, but compounded by the fact it occurs in much less time and with hardly any overvolting!

FYI the mainstream Quads out in March are C1 stepping. Intel Errata document talks of a new 45nm C1 stepping

John

[audiofreak]

...On a side note, my cpu has further degraded. 4ghz now only stable 1.41v real(DMM)...but I won't buy another Intel chip, either, until I hear that there is a new core revision, or something changes, as it's obvious Intel is not concerned about these cpu's dying...even if they knew they would before they left the factory. I'd advise others to do the same, or they'll never be concerned, and I won't build a pc for anyone using 45nm chips, either.

I got bit by INtel and faulty BadAxe boards, I'll not have my name further tarnished for supplying OTHER faulty parts.

Aww, I am so sorry for you... NOT!

1. Have you read the electrical specification for 45nm parts?

2. Have you seen the absolute maximum rating of 1.45V for Vcore in there?

3. Have you seen something like this on your processor box?



You knowingly exceeded manufacturer specifications so why are you complaining now? Why such a bitter tone against Intel?

If I were you (honest as you say), I would just warn the others not to make the same mistake you did. If you didn't overvolt it, then it would be a different story, this way you have no right to complain.

[cadaveca]

That's why I mentioned no RMA for me, thanks. So quick to judge...

And, if you read the Intel white papers, maximum voltage is 1.45v, not 1.225.

1.225 is the MAX VID, not maximum voltage, BTW.


And, if you read all my posts on the issue, the degredation initially happened BEFORE i gave the chip more than 1.425v. Now, I must digress, it is possible that voltage spikes carried the voltage over 1.45v, and hence my non-interest in an RMA.

But, you fail to notice that due to the temperature readings, and how these play a role in throttling, the way my cpu was, it wold have damaged itself before EVER beginning to throttle. Maximum temp Intel states via whitepaper explictly states maximum 71.4c, however, due to the bad sensors, my cpu would not begin to throttle until 100c.

This throttling issue is why I hold no trust in Intel's 45nm chips.

Just so we are clear tho...I have a QX9650 that does not have any temp issues as this 8400 does.

This tells me that thier binning process is flawed, or they simply just don't care. I'm not about to risk mine, or other's, money on that chance.

[Eastcoasthandle]

Aww, I am so sorry for you... NOT!

1. Have you read the electrical specification for 45nm parts?

2. Have you seen the absolute maximum rating of 1.45V for Vcore in there?

3. Have you seen something like this on your processor box?



You knowingly exceeded manufacturer specifications so why are you complaining now? Why such a bitter tone against Intel?

If I were you (honest as you say), I would just warn the others not to make the same mistake you did. If you didn't overvolt it, then it would be a different story, this way you have no right to complain.

VID is from 0.85V – 1.3625V for the E8400. But you can read what's on that label as well as core temps to get VID (from my understanding).

The max Core voltage is 1.45V (page 17, someone correct me if I am wrong). E6000 series CPUs are able to go as high as 1.55V (page 20)

[soundx98]

hmmm, Corsair HX-620 PSU (separate 4-pin and 8-pin leads), E8400 and DFI LT LP P35 T2R.

There was no difference in Idle temps using either lead/connector.
Loaded Optimized defaults, killed all unnecessary programs and services and let it rest for 5 minutes that checked with CoreTemp and Everest.
Tried twice for each connector about 8 minutes apart.

Same results (no change between 4 and 8 pin) at 4200MHz with 1.4v CPU using same methodology.
CPU VTT was set at 1.2v for the 4200 tests.

1st Core is 3-4 degrees hotter with Optimized defaults and 5-6 degrees hotter with 1.4v.

While I love the high OCs, I too am very nervous about the "wonky" temp readings.

Anyone else tried 4 vs 8 pin connections?

[KTE]

IDK about heat problems, we can't know that even with IHS measurements which are always plus >15C cooler than the hottest internal core temp, but the QC on these new chips is very poor and it looks certainly like a rushed launched whilst knowing of these problems. The chips have a large oc variation, some are very good and some not, QX is noticebley better than all and lower voltages kill/degrade the chips than before. AFAIK, the DTS are known to be faulty on all desktop/mobile Penryns, including why the mobile half was retracted and delayed from launch since that is where temps will count immensely. These desktop chips weren't and that looks more like market play to not lose major profits if they had recalled them. Coretemp and most software we've seen is reading the DTS reported temps correctly but the DTS are themselves faulty, most of the time they are stuck and at many other times very high/very low readings on all cores or more frequently, just two cores. We had to return our batches of E8200 and E8400 at work this last week because they were faulty in the same way and my boss won't tolerate that since it's a very crucial parameter in any business environment. You have CPU fanspeed control, oc headroom, stability and throttling which is all decided on the reported temps. My own E8400 was stuck at 6C no matter what but I should have a newer batch soon. I hope they get these problems sorted including short supplies which means our prices have shot up.

[mrcape]

I wonder what Intel is using as a QC test platform. Current boards? P45?

The errata document that JohnZS posted a link to is very interesting. I wonder how many of those error workarounds will get into upcoming bioses.

[JohnZS]

I think I figured out what is going on here. I am not 100% sure and I am by no means the brightest bulb in the light shop, but I think our motherboards (or in my case the ASUS P5E3) might be sending high voltage for FSB Termination.
I had a read of this QX9650 Datasheet

Now, I don't quite understand the whole electrical technical data, but it looks to me like the FSB Termination Voltage should be ~1.15V Max?
Now the minimum I can set FSB Termination Voltage on my P5E3 is 1.20V
CPU PLL Voltage it reads like the maximum Intel suggest should be no more than 1.55V?
The minimum my ASUS board will let me select is 1.50V
I was naive thinking that FSB Termination of 1.38 and SPU PLL of 1.56 were safe, but not anymore.
John

[Eastcoasthandle]

I think I figured out what is going on here. I am not 100% sure and I am by no means the brightest bulb in the light shop, but I think our motherboards (or in my case the ASUS P5E3) might be sending high voltage for FSB Termination.
I had a read of this QX9650 Datasheet

Now, I don't quite understand the whole electrical technical data, but it looks to me like the FSB Termination Voltage should be ~1.15V Max?
Now the minimum I can set FSB Termination Voltage on my P5E3 is 1.20V
CPU PLL Voltage it reads like the maximum Intel suggest should be no more than 1.55V?
The minimum my ASUS board will let me select is 1.50V
I was naive thinking that FSB Termination of 1.38 and SPU PLL of 1.56 were safe, but not anymore.
John

Would enabling loadline calibration help?

[JohnZS]

Would enabling loadline calibration help?

Thanks for the suggestion
As far as I have been able to figure out, the Load Line Calibration minimises VDroop under load. so therefore you require less voltage to be set into the VCore.

I have re-read my post and I didn't really make myself clear, what I meant to say was that the lowest value which are availible n the options are 1.20V for FSB Termination and 1.50 for CPU PLL....you cannot set any lower as those values are not availible.

Now, my calucations might be off, or I may not have read the datasheet for the QX9650 correctly...but it looks like the ASUS P5E3 is overvolting FSB Termination and is near the Electrical Specification Maximum for the CPU PLL.

Of course... I might be wrong....but if I am right... or even nearly right... it is quite worrying that a board with everything @ stock... overvolts!

and in turn it could explain the crazy temperature readings and why there are reports of chips dying.

John

[Eastcoasthandle]

Thanks for the suggestion
As far as I have been able to figure out, the Load Line Calibration minimises VDroop under load. so therefore you require less voltage to be set into the VCore.

I have re-read my post and I didn't really make myself clear, what I meant to say was that the lowest value which are availible n the options are 1.20V for FSB Termination and 1.50 for CPU PLL....you cannot set any lower as those values are not availible.

Now, my calucations might be off, or I may not have read the datasheet for the QX9650 correctly...but it looks like the ASUS P5E3 is overvolting FSB Termination and is near the Electrical Specification Maximum for the CPU PLL.

Of course... I might be wrong....but if I am right... or even nearly right... it is quite worrying that a board with everything @ stock... overvolts!

and in turn it could explain the crazy temperature readings and why there are reports of chips dying.

John

The Maximus Formula has similar settings as the P5E3. It would be interesting to see if this ever gets resolved.

[audiofreak]

That's why I mentioned no RMA for me, thanks. So quick to judge...

Not judging you, just saying that you should not overvolt it.

And, if you read the Intel white papers, maximum voltage is 1.45v, not 1.225.

I said that haven't I? I only pointed out that the max voltage or VID or whatever you want to call it can be lower for any particular CPU. For example, that particular 8200 says 1.225V on the box and it runs at 1.15V default.

That said, I wouldn't dare pushing the voltage up over 1.200V because 45nm process may be (due to smaller features) more prone to electromigration than the older manufacturing processes especially considering new (and unproven) materials used.

So what I am advising is more caution (yes I know this is XS forum but still).

Now, I must digress, it is possible that voltage spikes carried the voltage over 1.45v, and hence my non-interest in an RMA.

That is quite possible giving that the majority of boards now support Vdroop disabling which is actually a feature put there to protect both the VRD MOSFETS and the CPU.

Clarification: if you disable Vdroop, you get more straight voltage (which seems desirable), but when the CPU consumes lot of power during load (benchmark for example) and then abruptly stops doing that, the excess voltage/current has nowhere to go and forms a positive spike which could cross the defined rating much easier.

As for the temperature reading, I wasn't following the other posts. I can only say that I have noticed (even though I haven't overvolted my E8200) that the core temperatures differ sometimes by several degrees. At the moment I have 39c and 43c idle at 3.6GHz with (somewhat poorly seated due to a weak spring) Zalman 9500. I am not sure whether I should worry about that difference?

And this old article is worth reading:
http://www.hardocp.com/article.html?art=MjMw

[JohnZS]

The Maximus Formula has similar settings as the P5E3. It would be interesting to see if this ever gets resolved.

I wonder what other motherboards have for those settings?
With any luck some sort of clarification will be made soon, and hopefully BIOS updates which resolve this issue as it is scary to think that @ Stock settings the FSB Termination Voltage is 0.05V higher than the MAXIMUM allowed by Intel and that the CPU PLL Voltage is near the maximum tolerance for the CPU.

Is it anywonder there are many people having problems with these 45nm CPU's...and it might not be Intel's fault, but the fault of the motherboard manufacturers for not implementing the correct settings in the BIOS

John

[audiofreak]

Strange...

Not only Core #0 is cooler by 4-5 degrees Celsius:


But it also has lower utilization (DivX 6.8 encode from VirtualDub which I would expect to peg both at 100&#37:


Has anyone else noticed something like that?