I just delidded my old 1090T [PICS]

[IronBalls]

Hello,

I was bored and decided to remove the heatspreader of my still in use phenom x6. The process was quite easy, took me an hour. I cut around the black silicon, than put a lighter below the cpu and it popped up in less than a minute. Here are some pics I took.







The most time consuming part is the removing of the solder on the die. A real pain. I played a little around and the benefit is really small with a high end watercooling, 2-3 degress at most. I would say it was fun but not worth.

[Lightman]

Looks like you did great job if it still works

Nice pics!

[IronBalls]

It works like a charm. But I was expecting better temps. Doing this for 2-3 degrees is just not worth the trouble. Real world benefit is small too. My cpu does 4 ghz @ 1.425 v ... before it needed 1.45 v. I will do some more testing when it is less hot outside. Right now 35 degrees in my room, doesnt help for oc.

[IronBalls]

I did some testing. It has been less hot today, 27 C in my room.

There is nearly no benefit at all of removing the heatspreader. There are only two things I noticed. The cpu is more "stable". Before delidding it would sometimes crash without reason ( I think it was an IMC issue ). Now it seems to work stable. Secondly, temperature goes up more smoothly, this means that heat is transfered more fast to the waterblock.

Im using gelid extreme. I tried to apply metal paste but it wont stick to the die so I gave up. The surface looks like a mirroir once it is well cleaned. Way better than a sanded IHS.



Here some results :

After 40 minutes occt
Delta room temperature - water temperature is only 2 degrees, bare that in mind to interpret the results.


Last year at the same time ( it wasnt nearly as hot last year )



So there is a benefit, I would say 2-4 degrees. It doesnt however help for oc. I tried 4.3 ghz. Even with vcore @ 1.6v it isnt stable. If I could apply coolaboratory on the die, Im sure this would make a big difference. But it wont stick

[C-N]

Im using gelid extreme. I tried to apply metal paste but it wont stick to the die so I gave up.

Use a cotton bud & rub it in a few times & it will stick.

I find liquid metal works best at getting rid of high heat like what we get with Intel running Linpack etc 70oC +

[felix_w]

After reading the thread's title, looked back to the TS name...then it all made sense...

Well, you gained 0.025V, that is something, but for not being able to get higher OC , it may not be supported by the specific part you have...or it may be other setting responsible.

Does your motherboard have LLC options in bios ? Also, what is your CPU/NB voltage for 2800MHz ?

[IronBalls]

After reading the thread's title, looked back to the TS name...then it all made sense...

Well, you gained 0.025V, that is something, but for not being able to get higher OC , it may not be supported by the specific part you have...or it may be other setting responsible.

Does your motherboard have LLC options in bios ? Also, what is your CPU/NB voltage for 2800MHz ?

I also think the motherboard is the limiting factor. I have a gigabyte 890GPA ud3h. I dont have LLC option. For 2800 mhz im using 1.25v. I am able to do 3000 mhz with 1.35v but performance increase is really low so it isnt worth the extra voltage.

[felix_w]

Maybe with a stronger motherboard you could gain some +200-300MHz stable....I was able to get 4.0/2.8 stable with my 1090T and ASUS M4A79 but needed high voltage in both CPU & CPU/NB.

When i got the CHVF-Z i managed run stable 4.2/3.0 with lower voltages....LLC helped much but also the strong VRM and many bios options..i was even able to run Cinebench @ 4.4/3.0, which was not even a dream with previous motherboard

[IronBalls]

Maybe with a stronger motherboard you could gain some +200-300MHz stable....I was able to get 4.0/2.8 stable with my 1090T and ASUS M4A79 but needed high voltage in both CPU & CPU/NB.

When i got the CHVF-Z i managed run stable 4.2/3.0 with lower voltages....LLC helped much but also the strong VRM and many bios options..i was even able to run Cinebench @ 4.4/3.0, which was not even a dream with previous motherboard

I do think my motherboard is holding me back. Im surprised to see how much stability you won by changing only motherboard. This makes me consider buying the crosshair formula which isnt that expensive anymore on e.bay. I was also able to run Cinebench @ 4.4/3.0 in a suicide run .

[felix_w]

This is your VRM :



Looks worse than my 790X-UD3P's VRM...

I also changed cooling from total 5x120 internal slim (30mm) radiators to 9x120 external thick (48mm) one..

What was your batch and serial ?

[IronBalls]

Yes my motherboard doesnt like high voltages. Anything higher than 1.5v will result in sounds coming from the mobo. Weak vrms. My watercooling is entirely custom made : AquastreamXT ultra, mora-3 (9x120)

As for my batch impossible to tell, as I sanded down the IHS when I bought the cpu in 2011.

[felix_w]

Well, since we have similar watercooling system (Phobya 1080, D5, D-Tek Fuzion V2, but placed outside) i guess that you have concluded well, your m/b is the limit....
I only heard squeaks from the chokes or something from the VRM on the 790X-DS4 when used with 1090T, while struggling to give power to the cpu...

[BeepBeep2]

The 890GPA-UD3H has a 4+1 phase VRM but there is most certainly nothing wrong with it. Stronger than ASUS 890X/890GX/890FX non ROG/Sabertooth offerings.

I had two, and the first one was able to handle Deneb at 4.2 GHz / 1.6v on water rather easily. I remember the coil whine, which was very loud, but not really anything to worry about more than an annoyance IMHO.

[IronBalls]

The 890GPA-UD3H has a 4+1 phase VRM but there is most certainly nothing wrong with it. Stronger than ASUS 890X/890GX/890FX non ROG/Sabertooth offerings.

I had two, and the first one was able to handle Deneb at 4.2 GHz / 1.6v on water rather easily. I remember the coil whine, which was very loud, but not really anything to worry about more than an annoyance IMHO.

This might be true, but I still think the motherboard is holding me back. 4.2 ghz is stable with 1.5v ... but 4.3 ghz isnt stable even with 1.6v ... If I can get a asus 990FX i will try and see for myself. I might even buy a FX 8320 and delid it too

[cab]

so are all Phenoms soldered?!

[IronBalls]

so are all Phenoms soldered?!

Yes, all phenoms are soldered.

[cab]

Yes, all phenoms are soldered.

I can see then why not much improvement, the die is flatter than the heatspreader so it has only a slight advantage. Looks like a lot of solder, still better than haswell/ivy bridge.

Is the fx-8350 soldered too?

[IronBalls]

Improvement is perhaps bigger than I reported because I had previously sanded the IHS. It was really flat. But once the die looks like a miror. I think if I had compared " default IHS " vs direct contact we would see an 5-6 C difference.

All FX processors are soldered too.

[cab]

Improvement is perhaps bigger than I reported because I had previously sanded the IHS. It was really flat. But once the die looks like a miror. I think if I had compared " default IHS " vs direct contact we would see an 5-6 C difference.

All FX processors are soldered too.

I noticed that the circuits surrounding the die are epoxied, perhaps to prevent any solder-splatter from shorting them?

My Phenom II will also do 4 GHz @ 1.425v but I use 1.45 just to be safe. I have a couple of 890GPAs in new condition, you're welcome to try one to see if it's any better.

[BeepBeep2]

This might be true, but I still think the motherboard is holding me back. 4.2 ghz is stable with 1.5v ... but 4.3 ghz isnt stable even with 1.6v ... If I can get a asus 990FX i will try and see for myself. I might even buy a FX 8320 and delid it too

Some Thuban chips do not scale well on air at high voltage. Typically 1.55v+ is a land of diminishing returns...

BTW, you're the first person to successfully delid a Phenom II X6, chew* tried this in the past and basically called it impossible due to the extremely high probability of breaking off the SMT components surrounding the die. Congrats!

He warned about issues with chips dying at higher voltage when delidded to what appeared to be quick temperature swings and the IHS acting as "a buffer"...but I think bad contact was the real problem.

I'm not saying you can't gain some MHz from a top end board, but I personally don't think the 890GPA-UD3H has a terribly bad VRM.
I actually might still have one of these in the house but I think I killed it running the northbridge IGP at high volts on LN2.

I noticed that the circuits surrounding the die are epoxied, perhaps to prevent any solder-splatter from shorting them?

My Phenom II will also do 4 GHz @ 1.425v but I use 1.45 just to be safe. I have a couple of 890GPAs in new condition, you're welcome to try one to see if it's any better.

Most likely. Phenom II X4 weren't made this way...
http://forums.extremeoverclocking.co...8&d=1259826720

I'm certainly interested.
Phenom II was so much fun -
http://www.xtremesystems.org/forums/...s-5-GHz-on-air
FX 8000 series are fun too but not quite as satisfying with low IPC and are more fragile.

[tbone8ty]

Since the die is soldered your better off lapping the heatspreader for flatness

[IronBalls]

Some Thuban chips do not scale well on air at high voltage. Typically 1.55v+ is a land of diminishing returns...

BTW, you're the first person to successfully delid a Phenom II X6, chew* tried this in the past and basically called it impossible due to the extremely high probability of breaking off the SMT components surrounding the die. Congrats!

He warned about issues with chips dying at higher voltage when delidded to what appeared to be quick temperature swings and the IHS acting as "a buffer"...but I think bad contact was the real problem.

I'm not saying you can't gain some MHz from a top end board, but I personally don't think the 890GPA-UD3H has a terribly bad VRM.
I actually might still have one of these in the house but I think I killed it running the northbridge IGP at high volts on LN2.


Most likely. Phenom II X4 weren't made this way...
http://forums.extremeoverclocking.co...8&d=1259826720

I'm certainly interested.
Phenom II was so much fun -
http://www.xtremesystems.org/forums/...s-5-GHz-on-air
FX 8000 series are fun too but not quite as satisfying with low IPC and are more fragile.

I'm not the first one to achieve this. To my knowledge, two people achieved it before me.
The first one goes by the name of shrimp

http://forums.extremeoverclocking.co...d.php?t=332705

The second one, who inspired me to do this, posted a video on youtube

http://www.youtube.com/watch?v=ukq783v0nMQ

So I must be the third or so.

"He warned about issues with chips dying at higher voltage when delidded to what appeared to be quick temperature swings and the IHS acting as "a buffer"...but I think bad contact was the real problem."

Must be a bad contact. I noticed that temps go up and down less fast. So heat transfer should be better. After all, there is direct contact between the core and the waterblock.

[cab]

I cringed watching the torch job. There’s very little heat needed to melt solder, besides it’s probably low temp silver braze material, like they use to assemble racing bicycle frames. The German guy (2nd video) had the right idea, his mechanics-thermodynamics are excellent.

If they’re soldered, I don’t see the point in delidding, if anything the heatspreader gives a greater area for heat transfer. As for the IHS having some heat capacitance, it makes sense intuitively, especially since it appears (in the videos) to be lined with copper. Mind you I’m just a spectator here, I don’t mod anything.

I have GA-890GPA-UD3H and GA-Z77X-UD5H, both in new condition, and soon I’ll be taking one of my Sabertooths R2.0 out of service, if interested message me.

[IronBalls]

I cringed watching the torch job. There?s very little heat needed to melt solder, besides it?s probably low temp silver braze material, like they use to assemble racing bicycle frames. The German guy (2nd video) had the right idea, his mechanics-thermodynamics are excellent.

If they?re soldered, I don?t see the point in delidding, if anything the heatspreader gives a greater area for heat transfer. As for the IHS having some heat capacitance, it makes sense intuitively, especially since it appears (in the videos) to be lined with copper. Mind you I?m just a spectator here, I don?t mod anything.

I have GA-890GPA-UD3H and GA-Z77X-UD5H, both in new condition, and soon I?ll be taking one of my Sabertooths R2.0 out of service, if interested message me.

Using a torch is not needed and dangerous. He should have stopped after 1 min because there was something wrong with the cpu.
The first method I used was the one the german used but it did not work. I stopped after 1 min because the cpu was extremely hot and it wasnt moving at all. Then I decided to try with a simple lighter. After 45-60 seconds it popped up without issue. I did some research and the solder used melts at 140-170 C. The PCB can sustain 220C and more. So there is some " room " to play with, not as dangerous as one might think at first.

Once my phobya liquid arrives I will try it. Liquid metal has really high conductivity, nearly 10 times more than any thermal paste. With only the die to cool down, this might make a huge difference in temperature.

[BeepBeep2]

I'm not the first one to achieve this. To my knowledge, two people achieved it before me.
The first one goes by the name of shrimp

http://forums.extremeoverclocking.co...d.php?t=332705

The second one, who inspired me to do this, posted a video on youtube

http://www.youtube.com/watch?v=ukq783v0nMQ

So I must be the third or so.

"He warned about issues with chips dying at higher voltage when delidded to what appeared to be quick temperature swings and the IHS acting as "a buffer"...but I think bad contact was the real problem."

Must be a bad contact. I noticed that temps go up and down less fast. So heat transfer should be better. After all, there is direct contact between the core and the waterblock.

ShrimpBrime de-lidded Phenom II X4, but not X6...
The video on youtube is interesting. I hadn't known about that one, but they probably already knew to be careful about the SMT components touching the rubber seal.

I think the problem chew* had, were probably small hotspots on the die (possibly caused by air bubbles) or etc. It is good to keep in mind though that the indium solder does indeed act as a "buffer" and heatspreader to avoid hotspots when contact is good. For example, with perfect contact, and if one part of the die, say, the IMC heats up rapidly or more than other parts, the solder can help spread that heat across the die.

@cab - The solder has a melting point around 110-120c iirc and thermal conductivity around 80W/mK vs ~5W/mK for normal thermal paste. It's mostly indium...