Personal computing discussed

Moderators: renee, Starfalcon

 
Crayon Shin Chan
Minister of Gerbil Affairs
Topic Author
Posts: 2313
Joined: Fri Sep 06, 2002 11:14 am
Location: Malaysia
Contact:

Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 6:38 am

Recently I let my 1055T go down to stock speeds from 3.5GHz. What a difference! When all cores are working, the temperatures used to be around the low 60s... now it's 46-47C! I didn't even touch the voltages. Have you folks noticed the same huge variation in working temperature just from a 700MHz overclock?

Since I'm now running BOINC, I'm going to keep it at the slower speeds, because the difference in temperature is just too great, and it's warmer nowadays too.
Mothership: FX-8350, 12GB DDR3, M5A99X EVO, MSI GTX 1070 Sea Hawk, Crucial MX500 500GB
Supply ship: [email protected], 12GB DDR3, M4A88TD-V EVO/USB3
Corsair: Thinkpad X230
 
cheesyking
Minister of Gerbil Affairs
Posts: 2756
Joined: Sun Jan 25, 2004 7:52 am
Location: That London (or so I'm told)
Contact:

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 7:01 am

Crayon Shin Chan wrote:
just from a 700MHz overclock?

I wouldn't think of 25% as a small overclock, certainly not on a 6 core cpu. The next speed grade up only has a 100MHz faster clock and the top of the range only 500MHz.
Fernando!
Your mother ate my dog!
 
Crayon Shin Chan
Minister of Gerbil Affairs
Topic Author
Posts: 2313
Joined: Fri Sep 06, 2002 11:14 am
Location: Malaysia
Contact:

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 7:58 am

Actually I think it's quite conservative. Most people get around 3.8-4.0GHz., and then they have to raise the voltage a bit.
Mothership: FX-8350, 12GB DDR3, M5A99X EVO, MSI GTX 1070 Sea Hawk, Crucial MX500 500GB
Supply ship: [email protected], 12GB DDR3, M4A88TD-V EVO/USB3
Corsair: Thinkpad X230
 
DPete27
Grand Gerbil Poohbah
Posts: 3776
Joined: Wed Jan 26, 2011 12:50 pm
Location: Wisconsin, USA

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 9:11 am

13C difference seems a bit high to me if you left the voltage at stock values? Did you manually set voltage to the stock ~1.35V or are you allowing the mobo to apply "auto" voltage in which it will typically automatically apply extra voltage if you overclock. Might want to overclock it back up to 3.5GHz and check CPUz running Prime95 or Intel Burn Test to see what the voltage is reading under load. At stock voltage, temps shouldnt raise too much. But when you start pumping more voltage into the CPU, you could see a 10C jump. Here is a 1055T overclock at stock voltage

Also, what type of cooler are you using? 6 fully loaded and overclocked 125W TDP cores is a lot for a stock heatsink to cool. You might have exceeded the cooling capacity of your CPU cooler in your overclock, causing temps to go high.
Main: i5-3570K, ASRock Z77 Pro4-M, MSI RX480 8G, 500GB Crucial BX100, 2 TB Samsung EcoGreen F4, 16GB 1600MHz G.Skill @1.25V, EVGA 550-G2, Silverstone PS07B
HTPC: A8-5600K, MSI FM2-A75IA-E53, 4TB Seagate SSHD, 8GB 1866MHz G.Skill, Crosley D-25 Case Mod
 
Crayon Shin Chan
Minister of Gerbil Affairs
Topic Author
Posts: 2313
Joined: Fri Sep 06, 2002 11:14 am
Location: Malaysia
Contact:

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 9:17 am

I figured it out! I had left CPU load line calibration on "auto", and thus when it was under load the mainboard (M5A99X EVO) automatically cranked up the VCore to around 1.5V. That was what had caused the incredible temperature jump! Now that I've switched it to "regular", which means no load line calibration, the voltage stays at stock.

Now the temperatures under load are 53C, which is a reasonable increase from 43-46C before the overclock. I'm leaving LLC to off, I don't like it automatically increasing the voltage without my knowledge.
Mothership: FX-8350, 12GB DDR3, M5A99X EVO, MSI GTX 1070 Sea Hawk, Crucial MX500 500GB
Supply ship: [email protected], 12GB DDR3, M4A88TD-V EVO/USB3
Corsair: Thinkpad X230
 
Yeats
Gerbil XP
Posts: 349
Joined: Wed Oct 19, 2011 8:53 pm
Location: NJ, USA

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 9:27 am

Crayon Shin Chan wrote:
Actually I think it's quite conservative. Most people get around 3.8-4.0GHz., and then they have to raise the voltage a bit.


Most people you read about get up to 3.8-4.0, I'd bet most folks get less, especially for 24/7 usage.

Congrats on solving your problem, LLC definitely needs to be used carefully or not at all.
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 9:50 am

In theory, power dissipation of a CPU should scale roughly linearly with frequency, and with the square of the VCore. Furthermore, the temperature delta above ambient should be proportional to the power dissipated (yes, I'm assuming the fan speeds don't change... details, details).

Assuming an ambient of around 23C, all of these numbers make sense now (plus or minus a few degrees). It is nice when theory and practice agree! :D

Oh, and from what I've seen, low 60s was probably nothing to worry about; AMD's CPUs are tough little beasties! (My Phenom 9550 ran for an unknown number of days at 100C after the CPU fan died, and suffered no apparent ill effects... it still runs my web server to this day. Though I do wonder why there wasn't an automatic thermal shutdown.) But no sense consuming excess power and generating extra heat needlessly; glad you got it sorted!
Nostalgia isn't what it used to be.
 
Forge
Lord High Gerbil
Posts: 8253
Joined: Wed Dec 26, 2001 7:00 pm
Location: Gone

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 10:41 am

Maybe it's Intel myopia, but in my world LLC doesn't work like that at all.

What you saw sounds more like the Voltage: Auto issues.
Please don't edit my signature for me. Thanks.
 
DPete27
Grand Gerbil Poohbah
Posts: 3776
Joined: Wed Jan 26, 2011 12:50 pm
Location: Wisconsin, USA

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 11:40 am

I agree with Forge. LLC is meant to prevent voltage drop when the CPU gets loaded up. My understanding is that: normally LLC shouldnt increase voltage, just keep it from dropping. Increasing voltage is nornally left to an "Auto" VCore voltage setting. But hey, if adjusting LLC fixed the problem, I can't really argue that.
Main: i5-3570K, ASRock Z77 Pro4-M, MSI RX480 8G, 500GB Crucial BX100, 2 TB Samsung EcoGreen F4, 16GB 1600MHz G.Skill @1.25V, EVGA 550-G2, Silverstone PS07B
HTPC: A8-5600K, MSI FM2-A75IA-E53, 4TB Seagate SSHD, 8GB 1866MHz G.Skill, Crosley D-25 Case Mod
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 11:59 am

The whole point of a voltage regulator is to prevent voltage from varying with load. So what the heck is LLC anyway? The "we fixed our voltage regulator" feature?
Nostalgia isn't what it used to be.
 
Crayon Shin Chan
Minister of Gerbil Affairs
Topic Author
Posts: 2313
Joined: Fri Sep 06, 2002 11:14 am
Location: Malaysia
Contact:

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 12:10 pm

Well when I enabled LLC on my other board (M4A88TD-V EVO/USB3), it always overcompensated and ended up giving more voltage to the CPU when it was under load than when it was idle. Specifically, around 1.488V when idle, 1.525V loaded.

The M5A99X EVO provides an option to adjust the compensation, but I had left it on "auto", so it's understandable that it ended up overcompensating again too.

It's reassuring that it can withstand high temperatures, but still I wouldn't like the temps to go that high, otherwise the fans kick into overdrive and the computer ends up heating my small dorm room which is the last thing I need now that winter's over. Besides, I always feel like "it can do better" when I look at the temperatures.
Mothership: FX-8350, 12GB DDR3, M5A99X EVO, MSI GTX 1070 Sea Hawk, Crucial MX500 500GB
Supply ship: [email protected], 12GB DDR3, M4A88TD-V EVO/USB3
Corsair: Thinkpad X230
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 12:24 pm

It sounds like LLC is a kludge to avoid having to implement a feedback mechanism (from the power pins at the CPU socket) to control the VCore regulator on the motherboard. IOW they're guesstimating how much the voltage will droop due to electrical resistance of the power plane and CPU socket, and boosting the output by a corresponding amount. They're trying to fix something in the BIOS firmware which really ought to be handled by a physical change to the design of the PCB and VRMs...

Re LLC overcompensating... it probably isn't, or at least isn't overcompensating as much as you think. Unless the sensor you're getting that readout from is on the CPU it will be measuring the boosted voltage, which will be lower by the time it goes through the resistance of the power plane and the CPU socket contacts.

When you're dealing with low voltages and currents approaching 100 amps, resistance of the power plane and the socket becomes very significant!
Nostalgia isn't what it used to be.
 
Yeats
Gerbil XP
Posts: 349
Joined: Wed Oct 19, 2011 8:53 pm
Location: NJ, USA

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 12:49 pm

From my limited experience with multiple Asus mobos, I've noticed LLC over-boosting voltage more with AMD Phenoms than with Intel C2Q's when overclocking. In higher-end motherboards, you get to select the amount of LLC, in some you get settings (off, low, medium, high, extreme) and in low-end boards LLC is on/off.

I'm guessing the reason for LLC is to ease the burden on the VRMs during heavy overclocking, without having to burden the user with the granularity of setting voltage levels for each CPU power state.
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 1:01 pm

Yeats wrote:
I'm guessing the reason for LLC is to ease the burden on the VRMs during heavy overclocking, without having to burden the user with the granularity of setting voltage levels for each CPU power state.

It isn't going to "ease the burden" on the VRMs. If anything it will increase it, since they're being asked to pump out more power than they would without LLC.
Nostalgia isn't what it used to be.
 
DPete27
Grand Gerbil Poohbah
Posts: 3776
Joined: Wed Jan 26, 2011 12:50 pm
Location: Wisconsin, USA

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 1:01 pm

just brew it! wrote:
It sounds like LLC is a kludge to avoid having to implement a feedback mechanism (from the power pins at the CPU socket) to control the VCore regulator on the motherboard. IOW they're guesstimating how much the voltage will droop due to electrical resistance of the power plane and CPU socket, and boosting the output by a corresponding amount. They're trying to fix something in the BIOS firmware which really ought to be handled by a physical change to the design of the PCB and VRMs...


Yeah, and isn't "digital" VRM supposed to deliver smoother power delivery anyway? IMO that should almost negate the need for LLC altogether. Sounds like ASUS (and probably others) are using LLC nowadays for things it wasn't originally designed for.
Main: i5-3570K, ASRock Z77 Pro4-M, MSI RX480 8G, 500GB Crucial BX100, 2 TB Samsung EcoGreen F4, 16GB 1600MHz G.Skill @1.25V, EVGA 550-G2, Silverstone PS07B
HTPC: A8-5600K, MSI FM2-A75IA-E53, 4TB Seagate SSHD, 8GB 1866MHz G.Skill, Crosley D-25 Case Mod
 
Yeats
Gerbil XP
Posts: 349
Joined: Wed Oct 19, 2011 8:53 pm
Location: NJ, USA

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 1:14 pm

just brew it! wrote:
Yeats wrote:
I'm guessing the reason for LLC is to ease the burden on the VRMs during heavy overclocking, without having to burden the user with the granularity of setting voltage levels for each CPU power state.

It isn't going to "ease the burden" on the VRMs. If anything it will increase it, since they're being asked to pump out more power than they would without LLC.


Sorry, I should have said "ease the burden" on the reaction times for the VRMs. The circuitry isn't as responsive as the CPU.
 
Yeats
Gerbil XP
Posts: 349
Joined: Wed Oct 19, 2011 8:53 pm
Location: NJ, USA

Re: Frequency vs Temperature on a Phenom II X6

Fri Apr 13, 2012 1:16 pm

DPete27 wrote:
just brew it! wrote:
It sounds like LLC is a kludge to avoid having to implement a feedback mechanism (from the power pins at the CPU socket) to control the VCore regulator on the motherboard. IOW they're guesstimating how much the voltage will droop due to electrical resistance of the power plane and CPU socket, and boosting the output by a corresponding amount. They're trying to fix something in the BIOS firmware which really ought to be handled by a physical change to the design of the PCB and VRMs...


Yeah, and isn't "digital" VRM supposed to deliver smoother power delivery anyway? IMO that should almost negate the need for LLC altogether. Sounds like ASUS (and probably others) are using LLC nowadays for things it wasn't originally designed for.


It's cheaper to implement a hack than a redesign.
 
fff888
Gerbil In Training
Posts: 1
Joined: Mon May 28, 2012 7:21 pm

Re: Frequency vs Temperature on a Phenom II X6

Mon May 28, 2012 7:37 pm

just brew it! wrote:
voltage will droop due to electrical resistance of the power plane and CPU socket

It would be a big step for the computer industry to finally learn how to use superconductivity for the purpose of preventing any and all unwanted voltage drops in CPU sockets. So much money has been invested into such research, does anybody know any positive outcome?
 
JustAnEngineer
Gerbil God
Posts: 19673
Joined: Sat Jan 26, 2002 7:00 pm
Location: The Heart of Dixie

Re: Frequency vs Temperature on a Phenom II X6

Mon May 28, 2012 8:18 pm

fff888 wrote:
So much money has been invested into such research, does anybody know any positive outcome?
Plenty of PhD's have had gainful employment for decades. :)
 
Captain Ned
Global Moderator
Posts: 28704
Joined: Wed Jan 16, 2002 7:00 pm
Location: Vermont, USA

Re: Frequency vs Temperature on a Phenom II X6

Mon May 28, 2012 10:12 pm

JustAnEngineer wrote:
fff888 wrote:
So much money has been invested into such research, does anybody know any positive outcome?
Plenty of PhD's have had gainful employment for decades. :)

What's the price of LH2 these days?
What we have today is way too much pluribus and not enough unum.
 
Forge
Lord High Gerbil
Posts: 8253
Joined: Wed Dec 26, 2001 7:00 pm
Location: Gone

Re: Frequency vs Temperature on a Phenom II X6

Mon May 28, 2012 10:20 pm

Captain Ned wrote:
JustAnEngineer wrote:
fff888 wrote:
So much money has been invested into such research, does anybody know any positive outcome?
Plenty of PhD's have had gainful employment for decades. :)

What's the price of LH2 these days?


Nobody wants LH2 these days, that's rocket fuel. Everybody wants LN2.
I'm thinking about doing some dry ice or LN2 overclocking this year. Who's with me?
Please don't edit my signature for me. Thanks.
 
travbrad
Gerbil XP
Posts: 423
Joined: Mon Dec 08, 2008 5:39 pm

Re: Frequency vs Temperature on a Phenom II X6

Mon May 28, 2012 10:38 pm

Crayon Shin Chan wrote:
I figured it out! I had left CPU load line calibration on "auto", and thus when it was under load the mainboard (M5A99X EVO) automatically cranked up the VCore to around 1.5V. That was what had caused the incredible temperature jump! Now that I've switched it to "regular", which means no load line calibration, the voltage stays at stock.


Yep, I've found that the "AUTO" voltage settings seem to always use way more voltage than is needed. I've even seen cases where it uses so much voltage that the overclock becomes unstable, and manually setting it lower makes it completely stable.
6700K @ 4.6ghz || ASUS Sabertooth Z170 S || Crucial Ballistix DDR4-2400 16GB
ASUS STRIX GTX 970 || EVGA Supernova 750W G2 || Noctua NH-D15 || Fractal Define R5
Crucial MX200 500GB || 2x WD Blue 6TB || 2x WDGreen 2TB
Philips 272G5DYEB || Dell U2312HM
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Frequency vs Temperature on a Phenom II X6

Mon May 28, 2012 10:44 pm

fff888 wrote:
just brew it! wrote:
voltage will droop due to electrical resistance of the power plane and CPU socket

It would be a big step for the computer industry to finally learn how to use superconductivity for the purpose of preventing any and all unwanted voltage drops in CPU sockets. So much money has been invested into such research, does anybody know any positive outcome?

Not for consumer products. Even "high temperature" superconductors still require some pretty extreme cooling (-262 degrees F), making them completely impractical for everyday use.
Nostalgia isn't what it used to be.
 
Crayon Shin Chan
Minister of Gerbil Affairs
Topic Author
Posts: 2313
Joined: Fri Sep 06, 2002 11:14 am
Location: Malaysia
Contact:

Re: Frequency vs Temperature on a Phenom II X6

Mon Aug 18, 2014 6:54 am

just brew it! wrote:
It sounds like LLC is a kludge to avoid having to implement a feedback mechanism (from the power pins at the CPU socket) to control the VCore regulator on the motherboard. IOW they're guesstimating how much the voltage will droop due to electrical resistance of the power plane and CPU socket, and boosting the output by a corresponding amount. They're trying to fix something in the BIOS firmware which really ought to be handled by a physical change to the design of the PCB and VRMs...

Re LLC overcompensating... it probably isn't, or at least isn't overcompensating as much as you think. Unless the sensor you're getting that readout from is on the CPU it will be measuring the boosted voltage, which will be lower by the time it goes through the resistance of the power plane and the CPU socket contacts.

When you're dealing with low voltages and currents approaching 100 amps, resistance of the power plane and the socket becomes very significant!

Some time ago I read an Intel paper on LLC, apparently it was introduced with the Core 2 processors. The thing is, when you draw a lot of current (i.e. when the CPU is loaded), the voltage in any circuit naturally drops. If the drop (called Vdroop) is big enough, the voltage may not be enough for the CPU to be stable. LLC exists to bring the voltage back up and should be on all the time, according to Intel. A large part of the paper discussed the time delay between the CPU suddenly drawing a lot of current and the LLC bringing the voltage back up to compensate, during which the CPU might be unstable. This explains the "CPU Voltage Frequency" option within the M5A99X EVO's UEFI, which controls the "VRM transient response" to these kinds of sudden change behaviours. The higher transient response, the less time the CPU will receive a reduced voltage when it goes from 0-100% usage while waiting for the LLC to kick in.

I don't think the resistance of the power plane and the CPU socket contacts is important. It never changes anyway because the temperature of the board shouldn't change that much.
Mothership: FX-8350, 12GB DDR3, M5A99X EVO, MSI GTX 1070 Sea Hawk, Crucial MX500 500GB
Supply ship: [email protected], 12GB DDR3, M4A88TD-V EVO/USB3
Corsair: Thinkpad X230
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Frequency vs Temperature on a Phenom II X6

Mon Aug 18, 2014 7:13 am

Holy thread necro, Batman! :o (Might have originally been a spam necro, with the spam post since nuked...)

I think you misunderstood the post you were responding to. Of course the voltage will tend to drop when you draw more current. That's why I am questioning why VRMs don't implement some form of remote sensing, i.e. direct feedback from the CPU's power bus to the VRM. Seems to me this would eliminate the need for LLC (unless I am misunderstanding things, and remote sensing would not solve the problem that LLC deals with).

ATX PSUs have implemented remote sensing for years on the 3.3V rail, with a separate sense wire coming back from one of the 3.3V pins to the PSU. This wire carries no current; its purpose is to allow the PSU to sense the voltage at the motherboard end of the wiring harness, after the voltage drop due to the wire resistance.

Also, I never said the resistance of the motherboard's power planes or socket contacts was changing. The amount of current flowing changes, which causes the voltage drop across the (constant) resistance to vary, according to Ohm's Law.
Nostalgia isn't what it used to be.

Who is online

Users browsing this forum: No registered users and 1 guest
GZIP: On