Lapped my FX-8350, 4.3GHz OC'ed temps lowered by 9°C

[inf64]

From multitude of topics on XS AMD sub-forum. There are many FX users there,check it out.

[jvroig]

Originally Posted by inf64

From multitude of topics on XS AMD sub-forum. There are many FX users there,check it out.

You prove my point that people will do things, even crazy things, in the absence of real data + the pursuit of a numerical goal (in this case, GHz).

People settling on 1.55V are probably the same people who were used to Deneb/Thuban being safe at 1.5-1.55V.

That by no means proves 1.5V or beyond is anywhere near safe for AMD's 32nm products. It just means 1.55V isn't "sudden death", and that's hardly surprising.

[inf64]

If you push your CPU one mV over spec you are voiding your warranty (on paper) .If you really push it you are risking electromigration as you probably already know. Since I have not heard of one case of dead FX8350 until now I'd say the CPUs can take a lot of abuse. How much exactly is unknown and the risk is on the owner. We only have some data from forums like this and AMD won't give you a real answer on just how much is max. safe Vcore.

[jvroig]

Originally Posted by inf64

We only have some data from forums like this and AMD won't give you a real answer on just how much is max. safe Vcore

Exactly, and that's why I had to question your "1.55V should be max safe vcore" post, because such a message is presented as authoritative, when in fact, and as you so admit here, it is just people guessing and "well, it's not dead yet".

At least in the Phenom II era, we had AMD's guidelines to fall back on, as well as their published voltage ranges for each SKU.

[inf64]

Originally Posted by jvroig

Exactly, and that's why I had to question your "1.55V should be max safe vcore" post, because such a message is presented as authoritative, when in fact, and as you so admit here, it is just people guessing and "well, it's not dead yet".

At least in the Phenom II era, we had AMD's guidelines to fall back on, as well as their published voltage ranges for each SKU.

I guess English is your 1st language?
"1.55V should be max safe vcore"
There is no authoritative tone there,I just gave my opinion based on what I have seen on the forums elsewhere. IDC knows what he is doing and I doubt he would do some crazy stuff with overboard Vcore and crazy temps.
To be on the safe side subtract 6-7% from above and you will be running the FX8350 at 1.44-1.45V which is what most users hitting ~4.7-4.8Ghz are using.

[Idontcare]

Originally Posted by jvroig

Exactly, and that's why I had to question your "1.55V should be max safe vcore" post, because such a message is presented as authoritative, when in fact, and as you so admit here, it is just people guessing and "well, it's not dead yet".

At least in the Phenom II era, we had AMD's guidelines to fall back on, as well as their published voltage ranges for each SKU.

Yeah I found out earlier that the specs are lacking for most things 32nm. But that isn't just AMD, Intel too won't publish max Vcc specs for Sandy Bridge or Ivy Bridge.

So I wasn't expecting hard specs, more just wanting to know what people found to be unsafe versus what has been tried and found to not result in blue smoke and sparks

Originally Posted by inf64

1.55V should be max safe Vcore you would want to run your chip on. Make sure temps don't go over 65C and it should be OK.

Originally Posted by inf64

From multitude of topics on XS AMD sub-forum. There are many FX users there,check it out.

That works for me. About the best I could hope for in terms of a rough rule of thumb. I'll see where that gets me with the NH-D14, probably no higher than 4.8GHz.

[jvroig]

@inf64
Let's not get into semantics. Your post was presented as the end-all and be-all on the topic of max safe voltage, to wit: "1.55V should be max safe Vcore you would want to run your chip on", that's similar to statements telling someone exactly what to do, as in "This is what you should do and don't go beyond".

I was just making sure that it won't be interpreted as "the" max safe voltage, because it's not.

With the proper clarification that "at least, that's from people at XS", then it's fine. It clearly loses the authoritative quality and qualifies it as something that enthusiasts have settled on by experience, and not as something according to offical specs or guidelines

[inf64]

Ok to be on the safe side don't push past 1.5V. I have just rechecked and many XS users need around ~1.45V for ~4.7-4.8Ghz. If your chip cannot hit it with LLC set to high and with tha tbetter air cooler then it's a dud and I wouldn't torture it much.
Check this topic for example. This guy couldn't get 4.8Ghz stable on air.

@jvroig

I disagree. Should be is not equal as "1.55V is the max. safe voltage" . First one is a suggestion by me, second is authoritative statement. Sorry.

[jvroig]

Originally Posted by Idontcare

Intel too won't publish max Vcc specs for Sandy Bridge or Ivy Bridge.

What, sure they did, they published it but just called it VID, and eveyone on the internet used it as reference


(To everyone else: yeah, that statement above is a joke. Don't go running 1.52V on your Ivy Bridge, VID doesn't mean max safe operational voltages)

[jvroig]

Originally Posted by inf64

I disagree. Should be is not equal as "1.55V is the max. safe voltage" . First one is a suggestion by me, second is authoritative statement.

I don't want this to devolve into a grammar discussion. If you thought you weren't being authoritative, so be it. It is moot now anyways, since the succeeding postings from both of us have hashed it out very clearly that it is, in fact, from a source that is not official at all.

Originally Posted by inf64

Sorry.

Apology accepted, no harm done

[Jovec]

AMD sets its turbo voltage crazy high, so I've always my kept my 24/7 OCs under that. I haven't played with my 8350 yet, but my 1090t turbo voltage was 1.475v yet stock volts were 1.30 (or maybe 1.35v). Either way, plenty of room for a nice OC, especially since the max multi is 18 (or 18.5 with half steps) for CnQ to downlock during idle (again, for Thuban).

Edit: My 8350 stock vcores: idle@0.8875, stock/turbo off@1.3375v, stock/turbo on@1.4125v. It also seems to be thermal throttling around 52-53.5c, with a random core dropping the multi to 17x and the voltage to 1.2375v for a second or two.

[jvroig]

Originally Posted by Jovec

AMD sets its turbo voltage crazy high, so I've always my kept my 24/7 OCs under that. I haven't played with my 8350 yet, but my 1090t turbo voltage was 1.475v yet stock volts were 1.30 (or maybe 1.35v). Either way, plenty of room for a nice OC, especially since the max multi is 18 (or 18.5 with half steps) for CnQ to downlock during idle (again, for Thuban).

This has also been my observation with my own Thuban, and also my own behavior/decision-making regarding safe voltage OC settings. As long as the Vcore ends up less than what TurboCore would set it to every time it turbos, I call it good. I can't believe I completely forgot about this. :facepalm:

So, IDC, what's the max TurboCore voltage you see at pure stock settings? That seems a reasonable data point in finding a max safe Vcore for the 8350.

[deltree86]

@OP did you not cover the contacts on intel processors? In other words u used your hands directly without covering the processor contacts in any manner?

[Idontcare]

Originally Posted by deltree86

@OP did you not cover the contacts on intel processors? In other words u used your hands directly without covering the processor contacts in any manner?

I did not cover the contacts. I wear gloves when handling the processors.

[Plimogz]

Thanks for the great thread Idontcare! It's a fun read, as usual. (and oooooh the shiny!)

But I was wondering, with regards to the bent pins and your comment

Originally Posted by Idontcare

Now most people, from what I gather, use a styrofoam block to cover the pins before lapping their AMD chips. I intentionally avoided this because I was quite keen to avoid risking an ESD (electrostatic discharge) event and insulating dielectrics such as styrofoam are notorious for accumulating surface charges.

Are the AMD chips still shipping with their pins lodged in a sort of little black, stiff, plastic-y fabric mat? I've used that to afford the pins a little protection when lapping in the past, under the assumption that if it was okay for the manufacturer to package it like that, it shouldn't cause problem.

And I'm wondering if you would extend the above quoted comment to that material as well.

[lyssword]

that is sexy

[Idontcare]

Originally Posted by Plimogz

Thanks for the great thread Idontcare! It's a fun read, as usual. (and oooooh the shiny!)

But I was wondering, with regards to the bent pins and your comment
Are the AMD chips still shipping with their pins lodged in a sort of little black, stiff, plastic-y fabric mat? I've used that to afford the pins a little protection when lapping in the past, under the assumption that if it was okay for the manufacturer to package it like that, it shouldn't cause problem.

And I'm wondering if you would extend the above quoted comment to that material as well.

My FX-8350 did not come with the pins touching anything. It was packaged in a clear plastic clamshell that gripped the CPU's green PCB at the corners and left the pins hanging freely in the air without making contact to any portion of the plastic clamshell. There was no pad, nothing, in contact with the pins.

Very similar to the packaging for Intel's CPUs.

If AMD use to do it the way you describe, and now they don't, then that would suggest they learned something about the way they were packaging their CPUs which caused them to rethink their packaging and go to the trouble/expense of changing it.

[skipsneeky2]

Gotta try this lapping myself,i got a old dell with a pentium 4 that i could use as a test to practice on,would shake like a tree lapping my 3930k.

Wonder if the improvements on my chip would bring anything to the table?

[Idontcare]

Originally Posted by skipsneeky2

Gotta try this lapping myself,i got a old dell with a pentium 4 that i could use as a test to practice on,would shake like a tree lapping my 3930k.

Wonder if the improvements on my chip would bring anything to the table?

On a 3930K? Definitely. I saw a solid 5°C decrease in temps with my 2600k when I lapped it.

Lap it and get some Liquid Ultra on it, you won't be disappointed

[Idontcare]

Originally Posted by jvroig

So, IDC, what's the max TurboCore voltage you see at pure stock settings? That seems a reasonable data point in finding a max safe Vcore for the 8350.

I just checked, at pure stock conditions the VID for 4GHz when running 8 threads on a program like Prime95 is 1.3875V.

When I run with just one thread, so the chip turbo's up to 4.2GHz on one core, the VID jumps to 1.4250V (an increase of 0.0375V).

I figure I'm probably safe to take it another 10% higher than that, safe meaning "provided I am OK with reducing the expected lifespan from 10yrs to 3-4 yrs".

1.1 x 1.4250 = 1.5675V

Which, ironically enough, is exactly enough for me to hit 4.8GHz stable



Nice power reductions are seen in using the NH-D14 to lower the operating temperatures (minor voltage reduction seen as well).

When you can lower the CPU's power consumption by 54W by simply reducing the operating temperature some 34°C, you know some serious leakage is going on

Still, I was able to get 4.8GHz stable with prime95 LargeFFT within my arbitrary threshold of <10% over-voltage above the max turbo-core voltage AMD spec's the processor to run at. I consider that rather amazing, you won't get that kind of OC (20%) on an Intel Ivy Bridge for the same over-voltage budget.

But man look at that power consumption climb The first 10% OC pushes power usage up by some 51W (not very pretty I suppose) but that final 10% OC, from 4.4 to 4.8GHz, piles on another 143W!

Now obviously this result is skewed at the high end because that last 100Mhz, going from 4.7GHz to 4.8GHz, brings on half of the increase (70W). And if we look at the voltage graph we can see that 4.8GHz is really a different animal versus all the other clockspeed steps preceding it. (nice smooth evolution in the Vcc graph, until 4.8GHz which obviously breaks with the trend)



I would say 4.8GHz is not sustainable for this chip given the voltage and power bump it takes. At least not at the air-cooled operating temperature that goes along with it.

Time to try the H100!

[Plimogz]

Sheesh, 70W for 100MHz That's a pretty bitter pill to swallow. Talk about diminishing returns...

It sure does run cool under that NH-D14, though.

[Gikaseixas]

Originally Posted by Idontcare

I would say 4.8GHz is not sustainable for this chip given the voltage and power bump it takes. At least not at the air-cooled operating temperature that goes along with it.

I can do 4.9 but even 4.8 i see that i'm pushing too much so i settled @ 4.7
Sweet spot for me, it's a beast at that speed

[lau808]

idc, have you had a chance to test with the h100 yet?
eagerly awaiting more great reads from u :-)

[Idontcare]

Originally Posted by lau808

idc, have you had a chance to test with the h100 yet?
eagerly awaiting more great reads from u :-)

Did I forget to do my H100!? Doh!

I've been busy building out replacement boxes for my aging Q6600 rigs and for some reason I had convinced myself I was done with the FX8350 for now. Whoops!

Things are a little crazy here at homebase as we are in the 11th hour mad-rush leading up to putting our house on the market while simultaneously dealing with all that goes into moving to Taiwan (passports, visas, finding a house, getting kids registered for school, etc).

My "woe is me" whining aside, I have all the hardware sitting on my desk right this very minute so I imagine I'll make the time to assemble it and get tests done in a day or two.

I also want to figure out a way to ghetto-strap my VaporPhase LS onto the FX8350. It scales so well with cold in those suicide OC sessions AMD held a while ago, I got to imagine it will do well under vaporphase cooling provided the heat output doesn't overwhelm the cooling capacity. (I know my vapoLS can handle at least 270W which is what my QX6700 put out at 4GHz)

[sequoia464]

Originally Posted by Idontcare

Did I forget to do my H100!? Doh!

I've been busy building out replacement boxes for my aging Q6600 rigs and for some reason I had convinced myself I was done with the FX8350 for now. Whoops!

Things are a little crazy here at homebase as we are in the 11th hour mad-rush leading up to putting our house on the market while simultaneously dealing with all that goes into moving to Taiwan (passports, visas, finding a house, getting kids registered for school, etc).

My "woe is me" whining aside, I have all the hardware sitting on my desk right this very minute so I imagine I'll make the time to assemble it and get tests done in a day or two.

I also want to figure out a way to ghetto-strap my VaporPhase LS onto the FX8350. It scales so well with cold in those suicide OC sessions AMD held a while ago, I got to imagine it will do well under vaporphase cooling provided the heat output doesn't overwhelm the cooling capacity. (I know my vapoLS can handle at least 270W which is what my QX6700 put out at 4GHz)

Thought you lived in blizzard country - couldn't you just open a couple of windows and run a pseudo VapoPhase??

Good luck with the house - pricing is finally stabilized around here and even edging up a bit.