Overclock.net banner

Golden 7820X? Feels like winning the lottery!!!

15K views 32 replies 8 participants last post by  marcus556 
#1 ·
In my efforts to build a workstation/gaming computer I preordered a 8600K and a Taichi Z370 motherboard.

But then, a 7820X showed up on Craigslist for $475 new sealed in the box. I thought it was a good deal right??? Well... i called the seller and negotiated it down to $375!!!! Even better deal!!!! At this point i'm debating wether i shoud buy the 7820X and return the 8600K when it arrives... tough decision? Maybe.

To sweeten the deal, I found a Gigabyte Gaming 7 x299 motherboard for $346 on Amazon + a 20% discount and free prime shipping (final price $276.80). The matter is settled: I'll go to the dreaded x299 platform!!!! Wel... No 5ghz for me (i thought).

The amazon x299 motherboard arrived together with the 8600K and the taichi z370 (i already had the 7820X).

I had everything else so put the x299 system together and guess what?

$375 Craigslist i7-7820X running at 5ghz!!! with 1.31V while temperatures stayed at around 75-80C and even passed prime95 torture test for 1 hour!!! Of course running with a water cooler.

I'm a happy camper!!!

Am I right to be happy? Is this a golden chip?

I checked online and I only found 4.9ghz at 1.3. Mine runs at 5ghz 1.3V but not 100 stable. 1.31V stabilizes it.

Let me know what you guys think. A unicorn chip?
 
See less See more
#3 ·
I delidded a devils canyon before so i'm confident i could do it, but i don't think theres much more to gain on this chip by delidding it.

I turned off hyperthreading just to see if the cores were capable of running at 5.1 if temperatures were tamed, i got up to 1.4V and altough it ran windows and even cinebench, it wasnt 100% stable.
 
#6 ·
Mine is very similar. Cinebench at 5.0GHz ~1.32v. I can't get 5.1GHz to finish cinebench all core, but it will finish single core. However, my chips is delided. Let's see those cinebench numbers OP!





The voltage being reported in CPU-z is obviously not being reported correctly.
 
#7 ·
Absolutely!!!! I will oblige. First let me tell you what I did and the results:

So after i typed my last message it got even worse temperatures (i started using the computer while also running OCCT). So bad that it got to 104_C Yikes!!!.

So I pulled out the good ol' vise and went at it!!!! Poof!!!! Delidded in less than 3 minutes.

I also had leftover liquid metal from when I delidded my devil's canyon so that was good.

Only worry I have is this chip has a lot of little surface-mounted little components and I'm afraid if the liquid metal were to drip it could damage it. Tomorrow I will take the Chip back out and put liquid tape all over those components (tomorrow I'll upload pictures of my ghetto delidding tool and the chip with the liquid tape)

Here are the before and after ACCT runs (please note that the delid run was only for an hour and change, but that should suffice):

BEFORE (STOCK THERMAL SOLUTION FROM INTEL INSIDE):



AFTER DELIDDING AND USING LIQUID METAL (THE WE THE THINKERING PEOPLE'S WAY):



What you guys say?
 
#33 ·
Absolutely!!!! I will oblige. First let me tell you what I did and the results:

So after i typed my last message it got even worse temperatures (i started using the computer while also running OCCT). So bad that it got to 104_C Yikes!!!.

So I pulled out the good ol' vise and went at it!!!! Poof!!!! Delidded in less than 3 minutes.

I also had leftover liquid metal from when I delidded my devil's canyon so that was good.

Only worry I have is this chip has a lot of little surface-mounted little components and I'm afraid if the liquid metal were to drip it could damage it. Tomorrow I will take the Chip back out and put liquid tape all over those components (tomorrow I'll upload pictures of my ghetto delidding tool and the chip with the liquid tape)

Here are the before and after ACCT runs (please note that the delid run was only for an hour and change, but that should suffice):


BEFORE (STOCK THERMAL SOLUTION FROM INTEL INSIDE):




AFTER DELIDDING AND USING LIQUID METAL (THE WE THE THINKERING PEOPLE'S WAY):



What you guys say?

Where are the pics of the "ghetto" delidding tool? :) I am looking to purchase the 7900x in the coming weeks and thought about delidding mine.
 
#8 ·
Quote:
Originally Posted by samaelestevez View Post

Absolutely!!!! I will oblige. First let me tell you what I did and the results:

So after i typed my last message it got even worse temperatures (i started using the computer while also running OCCT). So bad that it got to 104_C Yikes!!!.

So I pulled out the good ol' vise and went at it!!!! Poof!!!! Delidded in less than 3 minutes.

I also had leftover liquid metal from when I delidded my devil's canyon so that was good.

Only worry I have is this chip has a lot of little surface-mounted little components and I'm afraid if the liquid metal were to drip it could damage it. Tomorrow I will take the Chip back out and put liquid tape all over those components (tomorrow I'll upload pictures of my ghetto delidding tool and the chip with the liquid tape)

Here are the before and after ACCT runs (please note that the delid run was only for an hour and change, but that should suffice):

BEFORE (STOCK THERMAL SOLUTION FROM INTEL INSIDE):



AFTER DELIDDING AND USING LIQUID METAL (THE WE THE THINKERING PEOPLE'S WAY):



What you guys say?
Nice! You are gutsier than I am. I sent mine to Silicon Lottery to be delid.

OCCT is a decent stress test. See what your temps are like with 30 minute runs of Prime95 or Realbench. It was much easier for me to get OCCT stable than either of those two. Also, running handbrake using x264 can expose some instabilities. I settled for an everyday overclock of 4.8GHz @ 1.285v. Anymore voltage than that and temps while running prime got too close to 90c for my liking. What kind of cooling are you using?

Here were my results 12hr runs of Prime95...one with AVX and one without AVX.

Here is 12hrs without AVX...



Here is 12hrs with AVX...



Running AVX without an offset holding it to ~4.3GHz gives temps higher than I like. You can see that 4.3GHz with AVX pulls max power 10w higher than 4.8GHz without.

I also have a Gigabyte 7 board. If you have not already, you are going to want to set "Package Power Limit 1 - TDP (Watts)" to 450, Package Power Limit 2 (Watts) to 450, "CPU VRIN External Override" to 1.900v, and "CPU Vcore Current Protection" to Extreme. Those setting will keep it from phantom throttling.
 
#9 ·
Which one of the prime95 torture tests did you run?

I haven't run any tests with AVX yet. But I don't mind offsetting it like you did.

Here is the promised CINEBENCH R15 run. Yours is a tad higher I guess because you have a much better memory kit than mine.



Now I'll try 5.1GHZ just for kicks
biggrin.gif
 
#10 ·
Quote:
Originally Posted by samaelestevez View Post

Which one of the prime95 torture tests did you run?

I haven't run any tests with AVX yet. But I don't mind offsetting it like you did.

Here is the promised CINEBENCH R15 run. Yours is a tad higher I guess because you have a much better memory kit than mine.



Now I'll try 5.1GHZ just for kicks
biggrin.gif
Nice man. These are fast little 8 cores. For Prime95, when I am running them for extended time like that, I use the "Blend". For quick and dirty temp tests I use Small FFT. Goodluck with 5.1GHz. I have not been able to do it yet.
 
#12 ·
Quote:
Originally Posted by samaelestevez View Post

I did!!! 5.1ghz. Not too hot either. I will try to overclock my memory to the max now. I'm crazy LOL

Nice!
 
#14 ·
Quote:
Originally Posted by samaelestevez View Post

I will run Prime 26.6 (Non-AVX) overnight at 5.1. I'll post the results in the morning.

Thanks for your help TahoeDust
No problem man. I would take a peek at temps after 30 minutes before I let it rock all night. At 1.375v temps are going to be intense. What kind of cooling are you running?
 
#16 ·
Since liquid metal scales up with increasing heat load, you could use it on the IHS -> Waterblock. https://forums.anandtech.com/threads/lapped-my-fx-8350-4-3ghz-oced-temps-lowered-by-9%C2%B0c.2300800/ If you don't care about warranty, you could lap it too.

There's also a method to completely remove the liquid metal residue without scratching the surface.

http://www.overclock.net/t/1588116/thermal-grizzly-conductonaut-73-w-mk/130#post_26305946
He managed to get 8C lower temp vs thermal paste on a soldered cpu ( delidded cpus are 4-6C lower temps than solder ). The 7820X with AVX512 will produce more heat than that 2011 processor.

http://www.overclock.net/t/923448/user-review-update-11-10-26-supernova-mora-2-mora-3-phobya-nova-1080
There's also these rads if you wanted even lower water temps.
 
#17 ·
Quote:
Originally Posted by czin125 View Post

Since liquid metal scales up with increasing heat load, you could use it on the IHS -> Waterblock. https://forums.anandtech.com/threads/lapped-my-fx-8350-4-3ghz-oced-temps-lowered-by-9%C2%B0c.2300800/ If you don't care about warranty, you could lap it too.

There's also a method to completely remove the liquid metal residue without scratching the surface.

http://www.overclock.net/t/1588116/thermal-grizzly-conductonaut-73-w-mk/130#post_26305946
He managed to get 8C lower temp vs thermal paste on a soldered cpu ( delidded cpus are 4-6C lower temps than solder ). The 7820X with AVX512 will produce more heat than that 2011 processor.

http://www.overclock.net/t/923448/user-review-update-11-10-26-supernova-mora-2-mora-3-phobya-nova-1080
There's also these rads if you wanted even lower water temps.
Thanks for all these suggestions!!! My current radiators configurstion is not too gar from those lol. Total overkill.
wink.gif


I will definitely try liquid metal between IHS --> waterblock.
 
#18 ·
How did the overnight Prime95 go?
 
#19 ·
I left it overnight at 1.35 / 5.1ghz but failed on two of the threads 1hr 48 minutes later while doing one of the small FFT passes
frown.gif


I didnt know that's how it worked, basically a prime blend test like you suggested is good because it throws different things at the processor. Thanks for the awesome suggestion!!!
smile.gif
.

Maybe I should've left it at 1.375 like i had it. Temps didnt really broke the low 80s even at 1.36V

Right now i'm working at 5ghz 1.31V.

Tonight after I'm done working ill try again at 1.375V / 5.1ghz.

Footnote: VRMS got over 100 degrees when i started the last tests last night. I added a fan with zip ties to cool them down
wink.gif
 
#20 ·
Quote:
Originally Posted by TahoeDust View Post

How did the overnight Prime95 go?
One question, since you seem to know about this stuff much more than i do: i was just going out for a little lunch break and decided to leave the computer running realbench for an hour or so (seems like a well regarded stability tests i hear). The computer as it stands can run Prime95 small FFT no problem for more than an hour, but Realbench freezes right away when the stress test starts. What can this be?

I took away memory and cache overclocks to leave it at pure CPU overclock and it stills crashes right away. Downloaded a fresh copy of realbench, still crashes right away. Any ideas?

BTW i left it running prime 95 small FFT and it is still going fine, but realbench fails right away. What do you think?
 
#21 ·
Quote:
Originally Posted by samaelestevez View Post

One question, since you seem to know about this stuff much more than i do: i was just going out for a little lunch break and decided to leave the computer running realbench for an hour or so (seems like a well regarded stability tests i hear). The computer as it stands can run Prime95 small FFT no problem for more than an hour, but Realbench freezes right away when the stress test starts. What can this be?

I took away memory and cache overclocks to leave it at pure CPU overclock and it stills crashes right away. Downloaded a fresh copy of realbench, still crashes right away. Any ideas?

BTW i left it running prime 95 small FFT and it is still going fine, but realbench fails right away. What do you think?
Which versions of Prime95 and Realbench are you running? Newer versions of Realbench will use AVX, so if you were using a version of Prime95 without it, that could be causing the crash. You may need to increase your AVX offset. Also, Realbench stresses GPU, so if your GPU is overclocked, it could be that.
 
#22 ·
Smart man... I will try that next (the AVX stuff).

Here are the results of my "Lunch Time Stability Test"

It was Prime 95 26.6 (Non-AVX) running 5ghz @ 1.28V

I'm a happy Camper!!!!

I really don't want to take the CPU out of there but I have to. I won't be able to sleep well if I don't put liquid tape on those Surface Mounted little things around it.

 
#23 ·
Quote:
Originally Posted by samaelestevez View Post

Smart man... I will try that next (the AVX stuff).

Here are the results of my "Lunch Time Stability Test"

It was Prime 95 26.6 (Non-AVX) running 5ghz @ 1.28V

I'm a happy Camper!!!!

I really don't want to take the CPU out of there but I have to. I won't be able to sleep well if I don't put liquid tape on those Surface Mounted little things around it.

Nice man!

Yeah, I would get that liquid tape on there. Don't want to ruin a good chip.
 
#24 ·
Hi Guys! Just a little update:

This truly is an amazing chip sample.

I even managed to run the computer at 5.2ghz even ran some benchmarks and short stability tests. Temperatures were in the high 80's and power consumption went through the roof so I settled for a 5.1GHZ overclock for everyday usage.

Final Overclock Results:

  • i7-7820X Delidded
  • 5.1 ghz -2 AVX -5 AVX 512
  • 1.36 VCore
  • 3200mhz DDR4 15-16-16-30 (overclocked from 3000mhz modules)
  • 3500 mhz uncore or cache
.
Benchmarks Results

Geekbench Score: 6,433 single core/ 37,571 multi-core
Cinebench Score: 231 single core / 2,304 multi-core
AIDA 64 Stable for 6 hrs

Thanks to everybody on this forum for all their help! This truly is an amazing community.



 
#26 ·
Quote:
Originally Posted by czin125 View Post

If you can manage to get it to run 4000C17 or 4000C16, you might be able to get lower ns than this.

http://www.xtremesystems.org/forums/attachment.php?attachmentid=132407&d=1503073527
That's some crazy fast memory you have over there!!!
thumb.gif


To my dismay, i don't think my memory modules will like that high speed very much. They are rated for 3000 c15 and i couldn't get them to post anything above 3200 no matter what timings I used. One of the modules (i tried one by one) posted at 3400 but that was very unstable. Didn't even booted to windows.
mad.gif
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top