To De-lid or Not to......

[ShrimpBrime]

OK I run a FX-9590. I'm simply looking for votes.

Success rate is pretty good. But this cpu is expensive and my second. I killed the first one which I'm sending off to a fella that recently purchased it.

It is said I'll lose maybe 10c and bring core and socket temps closer together. Which they already are close, just a little hot for any overclocking beyond stock 220w.....

SOoooo, should I do this? I've been contemplating for many weeks now. I've really layed back on benchmarking... I just can't get time in for it to really concentrate on proper tweaking. Sadens me crazy.

BUT it only takes minutes to perform a good de-lid. I'm just on the fence to break another expensive Cpu.

One hand says, meh you gotta job. objects are replace-able. The other hand says, 500$ get you a nice new Intel rig..... Not that I want another one.... just putting that there as a scape goat. but it's true....

So what would YOU do. I don't want opinions based on "Ya go for it!!!" Just kinda put yourself into shoes where a budget is always met weekly, savings are not for PC stuff. I buy and sell and trade this hobby (most of the time) used style.

Do it?

If it breaks.... I'd probably cry. just a single tear. but none the less cry....

 

[Mr.Scott]

Nope. Not worth it on a 9590. It'll still be a hotbox and still top out with the same clocks as any other 8 core FX.

 

[ShrimpBrime]

Nope. Not worth it on a 9590. It'll still be a hotbox and still top out with the same clocks as any other 8 core FX.

Yea I think the same thing! But its fun.... lol.

And the other issue is..... It's stock already at the same clocks as every one else's. There's no room for fun

believe me, this is on a back burner. Perhaps waiting till it's so dated.......

 

[RGone]

And the other issue is..... It's stock already at the same clocks as every one else's. There's no room for fun

That is the crux of the reasoning behind not suggesting users buy that hot thing in just an out of the box speed. Had they been uber binnings of 8 cores and capable of really lighting our fires rather than just making us hot...well that would have been a big deal for sure.

That high Vcore called for by the firmware inside the cpu as evidenced by the P-States at faster speeds is always a problem since doing a little further study found the boards are all too dependent on information fed it by the cpu itself.

I doubt you would clap-out two in a row. You have too many positive results to believe it would be two fails in a row on AMD's highest priced processor. If you really used that FX-9590 day in and day on apps that could use 8 cores correctly...well that processor could still see a long life even if forced to run 4.5/4..6Ghz and cooler. I think you could even leave it in AMD type turbo mode and use that CHV-Z to bring down the max Turbo multiplier to a cooler running range.

So De-Lid? Humh. Probably not worth the time nor the risk even if I just cannot see you with two fails in a row on the same highest priced AMD processor in their arsenal now for our desktops.

RGone...ster.

 

[ShrimpBrime]

I do "use it daily" and clocks, core counts and turbos differ from time to time. Depends on how I feel that day or week......

I use 3 main settings.

1. All 8 cores stock. Basically setting defaults and fly. I even leave memory at 667...

2. All 8 cores at 4.9ghz. I can actually leave the v-core stock which seems to run no higher than 1.48750v but typically under stressing or heavy loads will run 1.4150v // I know this is a considerable amount of v-droop, but it's actually not a big deal. This winds up in a freeze or hang at some point usually either right away or never at all. Prime95 doesn't throw errors or anything. It just hangs. This is due to low v-core and I know it. So this setting is used least.

3. I down 4 cores. Run 5ghz on all of them at recommended P-state voltage of 1.5250v. Has passed many hours of gaming, benching and even OCCT linpack and Prime95 (latest rev.) after a couple few hours.

Since I work 5 days 60+ hours during the week, choices 1 and 3.

The last week I've been testing some memory settings for fun. So running choice 1. with 669mhz memory frequency 7-12-7-28-46 clocks and a command rate of 1T. Stable with OCCT and gaming, probably head room for some reference clocks! But I haven't gotten a chance for this.

NOW - TEMPS!

choice 1 with above memory tweak, no real stressing other than my usual gaming and cinebench testing, read out as follows....

Cpu (socket) 27/21/46c
Package (cores) 8/0/52c

This thing can hit 70c very easily. I dare not run 8 cores 5ghz at 1.51250v Standard ambient cooling, thermostat on the wall reads 72f. For the record, just because my desk sits in a basement, doesn't mean it's cold down here. It's really not. My ever going dryer and furnace makes sure of that lol.

And a note about this set up. If you take OC to manual mode, it will not throttle. Most of the time it doesn't even crash (blue screen or errors on stress testing) it just simply low volt hangs. This chip would easily love 1.65v for any type of overclocking. Thus the itch to de-lid. 10c is a lot to loose as far as temperature goes.

So De-Lid? Humh. Probably not worth the time nor the risk even if I just cannot see you with two fails in a row on the same highest priced AMD processor in their arsenal now for our desktops.

RGone...ster.

Well I'd be a LOT more careful this time around. That I can assure you. I have a Phenom II 955BE C2 I can either practice on or sell or perhaps both?

The itch to de-lid this bad boy is there. Almost as bad as when I get the TEC itch. I simply love the challenge.

 

[Silver_Pharaoh]

I'd run 4 cores and OC like mad and call it a day.

No need to have another dead chip on your hands Shrimpy.
You've already got it clocked as high as my 6300 goes, with much less Vcore.

With 4 cores I'm sure you'd be able to clock higher than 5Ghz daily.

 

[ShrimpBrime]

I'd run 4 cores and OC like mad and call it a day.

No need to have another dead chip on your hands Shrimpy.


With 4 cores I'm sure you'd be able to clock higher than 5Ghz daily.

Not really. 5.1ghz x4 I think I can run stability testing. Past that, I'm using 1.6v and beyond. Temps fail. This is by far the most voltage demanding processor I've EVER owned. I could see it easily sucking down 2.2v to hit 8ghz. When I move beyond this platform, I'll LN the **** out of it.

You've already got it clocked as high as my 6300 goes, with much less Vcore.

I can thank AMD for that. Or Asus. I find it extremely odd that on auto everything stock, it only hits 1.5250v while it's turbo'ed to 5ghz. Any other time, it's around the 1.4520v range @ 4.7ghz. Volts as low as 0.876v @ 1400mhz. THIS is what keeps it cool. Clock all cores and stability test, My temps are easily in the 60s in probably 15 minutes. If anything reads 70c ++ the board hangs right up. Or Cpu. Not sure which. It is a pure temp issue. At 5200mhz, the temps spike so fast, that when you click run on cinebench, it's an instant hang. The temp software doesn't even have time to pull and report.

I feel a de-lid would cure 30% heat issues. (rough guesstimate) lols

 

[Silver_Pharaoh]

Not really. 5.1ghz x4 I think I can run stability testing. Past that, I'm using 1.6v and beyond. Temps fail. This is by far the most voltage demanding processor I've EVER owned. I could see it easily sucking down 2.2v to hit 8ghz. When I move beyond this platform, I'll LN the **** out of it.


I can thank AMD for that. Or Asus. I find it extremely odd that on auto everything stock, it only hits 1.5250v while it's turbo'ed to 5ghz. Any other time, it's around the 1.4520v range @ 4.7ghz. Volts as low as 0.876v @ 1400mhz. THIS is what keeps it cool. Clock all cores and stability test, My temps are easily in the 60s in probably 15 minutes. If anything reads 70c ++ the board hangs right up. Or Cpu. Not sure which. It is a pure temp issue. At 5200mhz, the temps spike so fast, that when you click run on cinebench, it's an instant hang. The temp software doesn't even have time to pull and report.

I feel a de-lid would cure 30% heat issues. (rough guesstimate) lols

On 4 cores with 1.6+ Vcore you are hitting 60+C?
Weird. I'm hitting ~ 62C, with the chip being able to handle 70C...
But I'm running 1.7 Vcore for that


Now, de-lid sounds great but these chips are soldered no? De-lidding can't nearly be as effective as say, a de-lid of a 3770K...

 

[Mr.Scott]

Now, de-lid sounds great but these chips are soldered no? De-lidding can't nearly be as effective as say, a de-lid of a 3770K...

Yes. Soldered, but still can be de-lidded. Much more effective than on a 3770. AMD is not Intel. I've caught 20+ degree temp drops after de-lidding.

 

[Silver_Pharaoh]

Yes. Soldered, but still can be de-lidded. Much more effective than on a 3770. AMD is not Intel. I've caught 20+ degree temp drops after de-lidding.

Celsius?

I was under the assumption that solder was better than TIM, so a de-lid wouldn't gain anything (or drop anything ) since it's metal to metal, and there's nothing more conductive of heat to metal than more metal...

I should de-lid my 6300 then

 

[ShrimpBrime]

Celsius?

I was under the assumption that solder was better than TIM, so a de-lid wouldn't gain anything (or drop anything ) since it's metal to metal, and there's nothing more conductive of heat to metal than more metal...

I should de-lid my 6300 then

Yes Celsius lol.

Solder is gotta be one of THE worst conductors of heat. Electricity, no problem.... But the lead in solder is one of the poorest conductual materials on the table brother.

I've gotta Morgan Silver dollar lapped down both sides to take the place of the Copper IHS plate. While it offers little in temp drops, it does ensure rapid heat transfer, the only barrier being thermal paste. I prefer to run block to cpu silicon which Ive also have lapped down pretty far, but to no better effects.

You could de-lid it. It's not super difficult, however is risky. That's the entire point of this thread actually. I know it could only bring better temps. But man the $$ . Chips not cheap. I've done all the cheap ones......

 

[Silver_Pharaoh]

Yes Celsius lol.

Solder is gotta be one of THE worst conductors of heat. Electricity, no problem.... But the lead in solder is one of the poorest conductual materials on the table brother.

I've gotta Morgan Silver dollar lapped down both sides to take the place of the Copper IHS plate. While it offers little in temp drops, it does ensure rapid heat transfer, the only barrier being thermal paste. I prefer to run block to cpu silicon which Ive also have lapped down pretty far, but to no better effects.

You could de-lid it. It's not super difficult, however is risky. That's the entire point of this thread actually. I know it could only bring better temps. But man the $$ . Chips not cheap. I've done all the cheap ones......

Thanks for the video!!!
(Stupid Youtube copyright crap, audio would have been nice! )

The only things I'm worried about are killing the chip in the process, and removing the solder afterwards.
The latter I see now being done with a razor and some probably wet fine grit sandpaper.

 

[RGE]

Intels solder is 87 w/mk. The best paste on the market is about 5 w/mk. liquid metal tim, clu/clp is about 40 w/mk, ie mostly gallium.

On an intel cpu like 3770k or 4770k/4790k, if you delid intels polymer tim paste at 3-5 w/mk (plus intels paste has a thick bondline ie 60 microns or so) and replace with 40 w/mk at a thinner bondline 10-20 microns thick, you can drop temps 20C if testing at 180w load.

If you delid intels indium solder (87 w/mk) like on an i950 cpu, and replace with 5 w/mk paste, you get worse temps, many have done it. The reason is surface area is too small at critical tim1 junction (hot spots on die) to use paste at 5 w/mk. Your better off leaving solder with heat conducting at 87 w/mk then through copper ihs at 400 w/mk, then with much larger surface area before hitting crappy paste. If you delid and use liquid metal for tim (40 w/mk) then you can drop temps 1-2C, or sometimes get 1-2C worse temps, depending on how well you applied liquid metal and mounting pressure when going bare die to cooling block. But there is less than 5C gradient through 87 w/mk solder tim and 400 w/mk copper, so zero chance of getting more than 5C better temps.

AMD hasnt listed solder conductance specs like intel...but all of the solders are at least in 40 w/mk conductance range....which makes zero sense in delidding unless you use liquid metal and run bare die....you might get 5C better temps.

quote form intel tech journal regarding 87 w/mk solder:
http://www.lsi.usp.br/~acseabra/gra...nmentally Green Microelectronic Packaging.pdf

The primary role of the IHS is to spread the heat out
evenly from the die and to provide a better bondline
control of the interface material. This can be achieved by
increasing the area of the IHS and by using a high thermal
conductivity thermal interface material with low
interfacial resistances. In order to meet thermal dissipation
targets, Intel introduced polymer thermal interface
materials (PTIM) initially with 3-4 W/moK bulk thermal
conductivity and then successfully transitioned to Pb-free
solder-based thermal interface material to meet the ever
increasing demand for thermal cooling capability as
shown in Figure 16.....
thermal conductivity and the mechanical compliance requirements resulted in
the development and qualification of low melting
temperatures (157oC Tm), low mechanical yield strength
(4-6 MPa), and relatively high thermal conductivity (~87
W/moK) pure Indium (In) metal for STIM applications.

 

[ShrimpBrime]

OK so your saying re-using the IHS plate makes only maybe 5c difference at best. Ya I can see that....

The point of why I de-lid is to not use the IHS plate at all. Why spread heat, I wanna transfer it before it's spread.

A later thought comment:



Replace Solder with TIM to directly contact the waterblock. Diamond particles measuring 0.0000015 cm rated at 8.3 w/mK or something of the sort within the paste.

Now removing that plate that seems to accumulate heat, then transfer it to the water block seems to make for a slow process.

Since Copper and Silver are really close together for conducting heat, It doesn't matter which is used. temps don't drop enough to replace copper plate with silver. But every little bit helps, so along the way, tried it.

In theory the IHS plate isn't really needed. It's dummy proofing really.

In the end solder is garbage for conducting heat. Taking IHS plates off AMD chips has done nothing but helped.

 

[Mr.Scott]

Taking IHS plates off AMD chips has done nothing but helped.

I totally agree.
Worst case de-lid I've ever had has been no change in temp overall, but equalized the core temps......and that was the worst case. Most average roughly a 10c decrease in temp.
I'm with Jon BTW, when I de-lid it's to run on a bare die.

 

[chrisjames61]

I would sell it and try to buy two of the really cheap 8 cores.

 

[caddi daddi]

I don't think i'd de-lid.
my 8350 would do 5.6 for benching, 4.6-4.8 for real work around the clock on rads and waterblock, if you plan on going super cooling then that's a question for Rgone or Mandrake.
I have been running some sims in openfoam with it, rather simple stuff, just dry air through round tubes and the 8350 does pretty well and the temps are well in hand, but in the near future I'll start with some "real" sims that will take many more calulations and much more time to complete and i don't expext the heat to be such an issue than a block and rad can't handle.

 

[dyckah]

I would sell it and try to buy two of the really cheap 8 cores.

um... brand new 9590 is $208... cheapest 8 core fx i see is $130... doesnt sound like selling the one to buy two is really that possible...

 

[ShrimpBrime]

I don't think i'd de-lid.
my 8350 would do 5.6 for benching, 4.6-4.8 for real work around the clock on rads and waterblock, if you plan on going super cooling then that's a question for Rgone or Mandrake.
I have been running some sims in openfoam with it, rather simple stuff, just dry air through round tubes and the 8350 does pretty well and the temps are well in hand, but in the near future I'll start with some "real" sims that will take many more calulations and much more time to complete and i don't expext the heat to be such an issue than a block and rad can't handle.

Eventually super cooling , yessir.

When I had an 8350, it was benching at 6100mhz DICE'd.

Air cooling is not an option.... sadly... but true. I tried it. The 9590 is sold with liquid cooling for some strange reasons. lol.

______

Getting a lot of nah's and no's. I'm actually kinda shocked!!

Thought there would be some heavy peer pressure here.....

 

[dyckah]

oh, forgot to give my input, but I say yes, why not?