I just delidded my newly bought FX 8320 [PICS]

[FlanK3r]

yes, th PSU seems the limit. With 3D Mark you cna try nice try:

set the CPU clock around 5 GHz for all results
set the HT link at 2200 MHZ, after 2600 MHz and after at 3000 MHZ and if your chip can, 3200 MHz. These results will interesting ,-).

[IronBalls]

yes, th PSU seems the limit. With 3D Mark you cna try nice try:

set the CPU clock around 5 GHz for all results
set the HT link at 2200 MHZ, after 2600 MHz and after at 3000 MHZ and if your chip can, 3200 MHz. These results will interesting ,-).

HT 2200

HT 3000

HT 3200

I dont really see what is the point of this. As my motherboard only lets me choose up to 2600 than I have to play with FSB.

[ShrimpBrime]

Looks really good. You did a very nice job de-lidding that chip! I'm glad to see some one else take the risk and go for it! Been at it since 2009 and felt I was a loner.

Done quite a few processors myself through some time now, all soldered, including and not limited to the one in my sig.

[IronBalls]

Looks really good. You did a very nice job de-lidding that chip! I'm glad to see some one else take the risk and go for it! Been at it since 2009 and felt I was a loner.

Done quite a few processors myself through some time now, all soldered, including and not limited to the one in my sig.

Thanks for writing on my thread. You are one of the reasons I tried this in the first place. You are a living legend when it comes to delidding soldered CPU , I can only look up to you.

[ShrimpBrime]

Thanks for your kind words! Glad to be an inspiration to people as this type of modding really catches my eye. This thread, your thread here is what got me to finally register here although I should have a long time ago as I read here often enough.

The first Phenom II I de-lidded was a 940BE DDR2 chip. I did knock off a couple of transistors, one completely off and another broken. It was still a use-able processor after the de-lid but suffered much degradation from many hard beatings and even with LN2 wouldn't go very fast. I think that one topped out at 4.4ghz for a validation.

I must say, it was nerve racking to de-lid the FX-8320 because these processors aren't exactly cheap. Because of that alone, I had to chime in, pay my 1$ reg fee (which I'm cheap and almost said nah f-it) just to post and tell you "great job" which you whole heatedly deserve. Hats off to you my friend.

[IronBalls]

Thanks for your kind words! Glad to be an inspiration to people as this type of modding really catches my eye. This thread, your thread here is what got me to finally register here although I should have a long time ago as I read here often enough.

The first Phenom II I de-lidded was a 940BE DDR2 chip. I did knock off a couple of transistors, one completely off and another broken. It was still a use-able processor after the de-lid but suffered much degradation from many hard beatings and even with LN2 wouldn't go very fast. I think that one topped out at 4.4ghz for a validation.

I must say, it was nerve racking to de-lid the FX-8320 because these processors aren't exactly cheap. Because of that alone, I had to chime in, pay my 1$ reg fee (which I'm cheap and almost said nah f-it) just to post and tell you "great job" which you whole heatedly deserve. Hats off to you my friend.

Thanks for your kind words too.
Even back then, as a teen, I used to delid Athlon 64 (also X2) processors. But its nothing compared to the excitement you get of delidding soldered CPU.

Again thanks. The funny part is I delidded the fx 8320 2 hours after the chip being delivered . I just quickly put it in the socket to see if it worked. Than took it out and delidded it.
I can not say for sure if it because of the delid that this one clocks so well. Its prime stable 2hours @ 5ghz and 1.4625v. Not bad considering I bought it 110 euros.

[ShrimpBrime]

I just quickly put it in the socket to see if it worked. Than took it out and delidded it.
I can not say for sure if it because of the delid that this one clocks so well. Its prime stable 2hours @ 5ghz and 1.4625v. Not bad considering I bought it 110 euros.

Lol, that's how I do them now. In the beginning was for testing temps and how that could effect overclocking. On average, I was seeing a Very considerable drop in temps bringing socket and core temps pretty close together while gaining 100mhz to the stable overclock and even running slightly lower voltages because of the better temps.

Now Mr.Scott at ClassicPlatforms has the 8320, and I know he can run that cpu at about 5.2ghz without issues. Before the de-lid it was tested stable at 5ghz with roughly the same voltage your using at 1.4625v maybe a tad more (LLC +3%) so it was about 200mhz gain with that processor.

I have a FX-5000 and 955BE sitting here waiting to get de-lidded. The s939 rig needs to come down to make room for the 790X-UD4 so I can test these processors before and after the mod.

[IronBalls]

Lol, that's how I do them now. In the beginning was for testing temps and how that could effect overclocking. On average, I was seeing a Very considerable drop in temps bringing socket and core temps pretty close together while gaining 100mhz to the stable overclock and even running slightly lower voltages because of the better temps.

Now Mr.Scott at ClassicPlatforms has the 8320, and I know he can run that cpu at about 5.2ghz without issues. Before the de-lid it was tested stable at 5ghz with roughly the same voltage your using at 1.4625v maybe a tad more (LLC +3%) so it was about 200mhz gain with that processor.

I have a FX-5000 and 955BE sitting here waiting to get de-lidded. The s939 rig needs to come down to make room for the 790X-UD4 so I can test these processors before and after the mod.

When I delidded my thuban 1090T, I noticed a 7-9 C temp drop. I was able to get 100 mhz more out of it afterwards.
With this one, fx 8320, I did no testing before so I cant say if it helpled or not but I would say yes.

Are you gonna delid these processors for you? I mean the FX-5000? Nevertheless I wish you good luck

[FlanK3r]

HT 2200

HT 3000

HT 3200

I dont really see what is the point of this. As my motherboard only lets me choose up to 2600 than I have to play with FSB.

The point is HT link...But seems, it will be working with Crossfire maybe, because there are small differencies and R9 290x is not the problem for HT link. I saw scores from Johan user with "huge" difference in score. But I dont remember if it was with 1 card or with 2 cards...

[ShrimpBrime]

When I delidded my thuban 1090T, I noticed a 7-9 C temp drop. I was able to get 100 mhz more out of it afterwards.
With this one, fx 8320, I did no testing before so I cant say if it helpled or not but I would say yes.

Are you gonna delid these processors for you? I mean the FX-5000? Nevertheless I wish you good luck

Nice, good temp drop for a 1090T, I de-lidded a 1090T as well and it lost the dual channel capability from the de-lid. I sold this one some time ago at EOCF.

The FX-5000 is an Athlon x2 5000+ unlocker. This one is a dud. It unlocks L3 cache to Core #0 only. So it's a dual core without L3 stock 5000+ or a single core with L3 cache unlocked. I did get it to clock up to 3000mhz and beyond.

It is mine, just going to de-lid this to sample it next to a real phenom II, they should be identical.

I did a 970T for a guy at OCF and he was pretty pleased running 4100mhz a nice little 100mhz boost for him. And that was done on air cooling.

[BeepBeep2]

I cant help you with that. Im not even sure that thing was a capacitor. If it was I would have had to make the bridge between the pads for the cpu to work. But I did nothing, it worked like that

It is most definitely a capacitor array.
http://www.ebay.com/itm/TDK-0-01uF-1...-/290562041111
http://in.rsdelivers.com/product/avx...c/4057498.aspx

etc etc.

To say that you would have had to make a bridge between the pads for the CPU to work, I don't think so. That may not be the only path for current to flow...

Also, welcome to XS, ShrimpBrime!!!

[IronBalls]

I just found out that phobya liquid dissolves the solder used by amd. With a little isopropyl alcohol it is possible to get a mirror finish of the die. Just look for yourselves

[IronBalls]

Nice, good temp drop for a 1090T, I de-lidded a 1090T as well and it lost the dual channel capability from the de-lid. I sold this one some time ago at EOCF.

The FX-5000 is an Athlon x2 5000+ unlocker. This one is a dud. It unlocks L3 cache to Core #0 only. So it's a dual core without L3 stock 5000+ or a single core with L3 cache unlocked. I did get it to clock up to 3000mhz and beyond.

It is mine, just going to de-lid this to sample it next to a real phenom II, they should be identical.

I did a 970T for a guy at OCF and he was pretty pleased running 4100mhz a nice little 100mhz boost for him. And that was done on air cooling.

Do delidded CPU sell for higher than the ones with IHS?

[ShrimpBrime]

Do delidded CPU sell for higher than the ones with IHS?

No they sell for less money as its considered damaged by most people, and they do sometimes get damaged as you've. Already shown us. I just do it for the hobby of it, much like running extreme cooling.

[ShrimpBrime]

Also, welcome to XS, ShrimpBrime!!!

Many thanks

[IronBalls]

After some fine tuning I finally got power consumption @ an acceptable level. 400 watts fully stressed while running prime95. I need 1.4625v for that, the calculated TDP is 240 watts. It is high, but in real life, this worst case scenario never happens. So it is acceptable. 5.1 ghz is also achieveable, but I need 1.5125v to get it completely stable. Thats nearly a TDP of 260 watts. So I settle @ 5 ghz. I wonder if the broken capacitor has actually improved the oc potentiel ...

[ShrimpBrime]

After some fine tuning I finally got power consumption @ an acceptable level. 400 watts fully stressed while running prime95. I need 1.4625v for that, the calculated TDP is 240 watts. It is high, but in real life, this worst case scenario never happens. So it is acceptable. 5.1 ghz is also achieveable, but I need 1.5125v to get it completely stable. Thats nearly a TDP of 260 watts. So I settle @ 5 ghz. I wonder if the broken capacitor has actually improved the oc potentiel ...

Looks Good Iron_B. I like the mirror finish, but would like to comment on it really quick if that's ok...

Couple of pointers, maybe they could help, maybe not. One extra lapping of the silicon will get your water block closer to the cores. Leaving the scoring or scratches or non mirror finish actually leave for extra surface area believe it or not. I use a nano diamond compound and try to take advantage of that microscopic surface area. I haven't tested this, but I figure it isn't bad when the Cpu drops an easy 8-10c.

Since the 5000 was only 10$ us, I have a good mind to lap and lap all the way down as far as possible. Maybe all the way down to the PCB which will make it shorter than the capacitors and will require a possible modification to a water block, or perhaps design some other method for cooling, possible a water block glued to the PCB with the bottom cut off and run the water directly onto the silicon. This will be curious to see how far down I can go, but I believe the actual core layout is inside the PCB, or near the top portion of it at least.

[BeepBeep2]

Looks Good Iron_B. I like the mirror finish, but would like to comment on it really quick if that's ok...

Couple of pointers, maybe they could help, maybe not. One extra lapping of the silicon will get your water block closer to the cores. Leaving the scoring or scratches or non mirror finish actually leave for extra surface area believe it or not. I use a nano diamond compound and try to take advantage of that microscopic surface area. I haven't tested this, but I figure it isn't bad when the Cpu drops an easy 8-10c.

Since the 5000 was only 10$ us, I have a good mind to lap and lap all the way down as far as possible. Maybe all the way down to the PCB which will make it shorter than the capacitors and will require a possible modification to a water block, or perhaps design some other method for cooling, possible a water block glued to the PCB with the bottom cut off and run the water directly onto the silicon. This will be curious to see how far down I can go, but I believe the actual core layout is inside the PCB, or near the top portion of it at least.

That "extra surface area" is bad. Your waterblock does not have mirrored matching scores to fit precisely into those. The idea is to have the best contact possible between the surfaces, which is why a mirror finish (4000 grit etc) is best. You're right, it does create extra surface area, but it actually hurts performance because the thermal paste does not move heat nearly as well as the surface of the block. In an ideal world, you'd have two perfectly flat surfaces and not have to use thermal paste. We use thermal paste because air is an insulator and microscopic air particles would fill in between the surfaces if we didn't.

As far as getting closer to the cores, this is really not going to do anything. Getting 0.05mm closer to the heat source may result in an unnoticeable gain. Silicon already has terrific heat transfer.

Having cracked a dead Deneb chip open myself, I'm also warning you not to lap any further than you need to. There are multiple layers within the die that go all the way near the top. If you were to lap it even close to the PCB, you would remove several (...most) of those layers.

[ShrimpBrime]

We shall see about that

[IronBalls]

Looks Good Iron_B. I like the mirror finish, but would like to comment on it really quick if that's ok...

Couple of pointers, maybe they could help, maybe not. One extra lapping of the silicon will get your water block closer to the cores. Leaving the scoring or scratches or non mirror finish actually leave for extra surface area believe it or not. I use a nano diamond compound and try to take advantage of that microscopic surface area. I haven't tested this, but I figure it isn't bad when the Cpu drops an easy 8-10c.

Since the 5000 was only 10$ us, I have a good mind to lap and lap all the way down as far as possible. Maybe all the way down to the PCB which will make it shorter than the capacitors and will require a possible modification to a water block, or perhaps design some other method for cooling, possible a water block glued to the PCB with the bottom cut off and run the water directly onto the silicon. This will be curious to see how far down I can go, but I believe the actual core layout is inside the PCB, or near the top portion of it at least.

I see your point when you talk about extra surface. Dont know if there will be a real life benefit from doing that. Before using liquid metal, I used gelid extreme. The only thing I can say, liquid metal beats the hell out of gelid extreme.

I think you are right when you say the actual die is " within " the pcb. For the sake of science, I sanded down an old athlon 3800+ which I killed while doing a suicide run. The cpu didnt like 2.1v .
The pics show clearly that the die is below the level of the capacitors.

[ShrimpBrime]

Good pictures Iron_B!! Looks like we can get a good sanding pretty far down there. Is it Necessary? Meh, who cares really. Just want to see if it's possible?

LM is good stuff. Haven't used it yet. My water blocks go right on silicon, especially the AMD 125w processors.

[ShrimpBrime]

OK the De-lid is completed. Begining lapping now, just so far have taken the solder out of the silicon and taking off material. Going to try and take off as much as possible, this may take a few hours of concentration, probably have it down to the capacitors and be stumped at that point. I'm trying to imagine how to lap beyond the little caps, they are in the way..... For now, down to them as close as possible. Then I may test the cpu. I'll have to put a coating on them, perhaps some clear nail polish or some casket maker or something insulating so the caps don't touch the water block.

[ShrimpBrime]

Here is about 45 minutes with 1800b grit sand paper.

[IronBalls]

Here is about 45 minutes with 1800b grit sand paper.

Looks good, but it seems to me you can go a little further down. It will take time to get to the level of the capacitors. I used 180b sand paper, but thats only because the cpu was already dead so I didnt care to crack the core ( and it did happen at the end ).

Im curious to see the results of your experiment

[ShrimpBrime]

Kept on going.... and going.... and then started to see what looks like hair line cracks! Not sure how those got there, but I haven't dropped the processor or anything like that. This has not been mounted under a water block yet.... I think the lapping can stop here??

The small scratches in the PCB is from removing the glue. I used my pocket knife....