Damn 7820x is impressive (my own chip)

You realize the the links you posted show all of the air coolers getting beat by good 240 AIOs right? Even the very best FUMA gets beat by the best AIO and is still within 1C/1dB of the other good AIOs. I guess that's what you meant by "astronomical." I guess my frame of reference was off.

So I decided to post latest review to avoid cherry picking and you still decided to made an statement based on a single one?.

what about this:

1494375025OStBK3D4ap_3_2.png


1494375025OStBK3D4ap_3_4.png


and this?

1487198614HJJmdo8xsz_3_2.png


1487198614HJJmdo8xsz_3_4.png


what about this?
1425569641XP4EChKoVd_3_2.png


1425569641XP4EChKoVd_3_4.png


oh wait...

1430051962VF0AJetlpn_3_2.png


1430051962VF0AJetlpn_3_4.png


another one?

1420032041o3rmje1xpl_3_2.png


1420032041o3rmje1xpl_3_4.png


oh.. just another one more please...

1417707371g9dzEbglg6_3_2.png


1417707371g9dzEbglg6_3_4.png


You know, this is exactly what I was wanting to not do, because lazy people as you just don't want to be properly documented before making an statement in a web forum.. i've been AIO user since the days of CooliT 92mm AIO and the first Corsair H50 Unit.. but Also i've been an High-end Air Cooling guy, I have 13 machines all equiped with AIO/CLC and Air Cooling, I know exactly how loud they can be.. and never, just never AIO offer the same Cooling performance at the same noise levels than mid range to high-end Air cooling.. they may cool couple of degrees better? yes.. at exponentially higher sound levels.
 
So let me take the last picture you posted and look at it. The H100i cools ~5C cooler at low fan speeds than the NH-D15 and is .1dB louder. Sounds like a real trade off to me.

Thankfully with the magic ability of PWM fans you can have cool temps and lower noise too.

I'm not knocking high end air, but you claiming that it's "astronomically" better in both sound and temps is a little outlandish don't you think?
 
Here's what I get so far at 4.6 Ghz. I can't seem to get the system to load when CPU Cache is over 3.0Ghz however.

vsnrpk.png
 
So let me take the last picture you posted and look at it. The H100i cools ~5C cooler at low fan speeds than the NH-D15 and is .1dB louder. Sounds like a real trade off to me.

Thankfully with the magic ability of PWM fans you can have cool temps and lower noise too.

I'm not knocking high end air, but you claiming that it's "astronomically" better in both sound and temps is a little outlandish don't you think?
What doesn’t show in charts is the infamous pump noise. You’ll hear it from any watercooler when the rest of the system is virtually silent. I find it much more distracting than loud fans.
 
What doesn’t show in charts is the infamous pump noise. You’ll hear it from any watercooler when the rest of the system is virtually silent. I find it much more distracting than loud fans.

While I agree that the pump noise is there in a quiet environment, I have two Corsair pumps in my system, and they're still barely audible. Dampened Fractal Design Define R5 helps ;).
 
Ok this is great and all, but this thread is about the 7820x and those whom are interested in it and those whom are getting them and setting them up and reporting their thoughts.
 
Now up to 4.7Ghz, but I had to increase core voltage to 1.250. Haven't tested for stability yet though. Just a quick screenshot of Cinebench.

* Update... well, I was having some random crashes during heavy MMO gaming, so I've dropped it back to 4.6Ghz @ 1.200v. We'll see how this goes for a while.

2ym7eo2.png
 
Last edited:
So I was playing around with the CPU vcore voltage to see how low I could drop the voltage @ 4.7 ghz. Ive dropped it down to 1.18V, and was able to run multiple firestrike runs without issue. CPU temp maxed out around 58C. I set the mesh to 3 ghz, and was getting mid 17000s in firestrike with my setup below.
 
So I was playing around with the CPU vcore voltage to see how low I could drop the voltage @ 4.7 ghz. Ive dropped it down to 1.18V, and was able to run multiple firestrike runs without issue. CPU temp maxed out around 58C. I set the mesh to 3 ghz, and was getting mid 17000s in firestrike with my setup below.

You won the lottery.
 
Firestrike isn't a good test. My 7820x can do similar, but once I run some WCG crunching then no go.
Im not making wild assumptions. I figure that he know if he's stable and I was commending him on it.

As much as I know we like to reference Cinebench it really does load a processor hard enough to actually deem if it is going to crash under load. I use it all the time to load test the processor with something is dynamic.
 
Last edited:
I've been playing around a bit with mine. Screenshot is an hour of non-AVX prime95. It got pretty hot. Is an hour long enough to be reasonably certain it's stable?
I am thinking of setting 4 core ratio to 46 and 8 core to 44 for now...

Kraken x62 for cooling with 4 fans push/pull - it's pretty much maxed so I am at a thermal wall unless I start tweaking cores individually (not something I am interested in doing)
 

Attachments

  • prime95_hour.PNG
    prime95_hour.PNG
    1 MB · Views: 52
Running mostly stock clocks wioth my mesh at 3 GHz, idle is around 20-25° and load depending on the game but 50-67° ish.
 
Running custom loop full x299 Prime Monoblock...

Vrms around 50c wide open load

Cpu 20s low 30s idle depending on ambient.

Full load is 80s at 4.8 x8 and 3ghz mesh.

40s to low 60s gaming depending on core demand but avg. Is 55 to 60.
Running triple radiator with push pull all corsair fans.

If I add another rad I can shave 10 off delta T but CPU core temps are consistent no matter what.

Havent tried for 5.
 
Running custom loop full x299 Prime Monoblock...

Vrms around 50c wide open load

Cpu 20s low 30s idle depending on ambient.

Full load is 80s at 4.8 x8 and 3ghz mesh.

40s to low 60s gaming depending on core demand but avg. Is 55 to 60.
Running triple radiator with push pull all corsair fans.

If I add another rad I can shave 10 off delta T but CPU core temps are consistent no matter what.

Havent tried for 5.
Does triple radiators require more than one pump? Would be interesting to see how flow rates are impacted.
 
Last edited:
I believe OP is referring to a single 360 radiator, not three individual radiators...
Oh so triple 120mm fans, i got it. Still would be interested to know if 3 rads requires multiple pumps just out of curiosity
 
Oh so triple 120mm fans, i got it. Still would be interested to know if 3 rads requires multiple pumps just out of curiosity

I would think 3 rads would require 1 strong pump with a good head pressure like the mcp35x.

Also you can trick a low head pressure pump by gravity assisting the flow i.e. place the pump above your radiator such as that the pump works less to push X volume of water through the loop. People that place a weak pump at the bottom of the case may encounter velocity problems when meshing many rads in series. Another option if space allows is older rads would allow parallel operation.

Or one could use 2 weaker pumps in series to effectively flow to decent velocities.

I pump my liquid such that the velocity creates enough turbulence to give as many water molecules chance to receive enough energy without having too high a velocity. Lower pressure higher velocity is how I like to run my loops. So I use a smaller hardline diameter than the bigger common size.

Yes I did mean a triple fan radiator sorry.

As far as adding a second radiator my Enermax Combo res has enough oomph to easily create high velocity at lower RPMs. I run my pump at about 2100rpm and it never changes no matter the load. There is no need to run a variable speed pump that changes with thermal load. Water has an excellent heat capacity 4kj/k. Ive toyed with the idea of using liq. Amonnia as it has a 6kj/kelvin cap but in all honesty all that means is your parts will still hit thirr average load temps but your coolant will just take longer to reach equilibrium between the cooling capacity of your radiator and the max temp the liquid will hold before radiators transfer the energy to the atmosphere.

But it would be fun as hell and probably destructive to my parts and toxic as hell haha.
 
Last edited:
4.7Ghz @ 1.300v seems to be holding steady so far. This is with 3.0Ghz Mesh.
I can run mine at 1.25v 4.8. Be a little more aggressive on that thing. It might surprise you.

I use Adaptive offset now.

I bet ar 4.7 that thing feels like a little monster. Mine does does for sure for an 8 core.
 
What doesn’t show in charts is the infamous pump noise. You’ll hear it from any watercooler when the rest of the system is virtually silent. I find it much more distracting than loud fans.
I've noticed that pump noise in AIO for me is a setup issue. has to do something with how the hoses are and where are the air pockets inside. Every time I had audible pump noise it disappeared completely after rearranging the hoses and/or flipping the pump around. At least how it works with the cooler I have. I suspect it could be the same for others.
 
Where on Earth besides Caltech and Dolby do you get a room quiet enough that you can actually hear a water pump? Maybe if its running wide open but that just means you dont know anything about water cooling and thermodynamics or something is improperly addressed in your setup.

I can hear my furnace air blowing through my registers in my room more than I can hear my waterpump.

I have no idea how or why people think a watercooler is supposed to be perfectly silent compared to what? Case fans?

If your running a 300 watt CPU @ 4.5 plus ghz all core OC and you have OC ram and big ass heat generating GPUs what makes people think they can get true passive silence out of thier rigs. If your not OC ing your rig and running adequate cooling your welcome to ask to start a subforum for ultra quiet computing because its a tiring subject over here in the performance sections of hardforums.
 
Last edited:
Where on Earth besides Caltech and Dolby do you get a room quiet enough that you can actually hear a water pump?

I can hear the pump in the H80i in my old setup, for an anecdotal example.

I'll be playing with it to see if I can quiet it down.
 
Currently running mine at 4.7GHz @ 1.229v stable. Temps are high, but I'm waiting for a delid tool :). Wonder how far I can push this thing...

Quick info: Mesh at 3.3GHz, -5 AVX off-set. RAM is 32GB Ripjaws V 2666MHz CL15 running 3200MHz CL17 (got these for a steal on black-friday).

MSI x299 SLI Plus, max VRM temp was 85 *C during stress testing.

Custom loop with one 120mm and one 280mm radiator.

7820x 4.7ghz.jpg 7820x 4.7ghz_stable.jpg
 
Currently running mine at 4.7GHz @ 1.229v stable. Temps are high, but I'm waiting for a delid tool :). Wonder how far I can push this thing...

Quick info: Mesh at 3.3GHz, -5 AVX off-set. RAM is 32GB Ripjaws V 2666MHz CL15 running 3200MHz CL17 (got these for a steal on black-friday).

MSI x299 SLI Plus, max VRM temp was 85 *C during stress testing.

Custom loop with one 120mm and one 280mm radiator.

View attachment 47022 View attachment 47023

Nice. Delid will help alot. I run a monoblock so my vrms cruise under load at like 35c to 45c no hotter. Those vrm chips are tiny. It doesnt take much excess heat to kill one.
 
Currently running mine at 4.7GHz @ 1.229v stable. Temps are high, but I'm waiting for a delid tool :). Wonder how far I can push this thing...

Quick info: Mesh at 3.3GHz, -5 AVX off-set. RAM is 32GB Ripjaws V 2666MHz CL15 running 3200MHz CL17 (got these for a steal on black-friday).

MSI x299 SLI Plus, max VRM temp was 85 *C during stress testing.

Custom loop with one 120mm and one 280mm radiator.

View attachment 47022 View attachment 47023

Nice score dude! A bit higher than my 7820X at 4.7Ghz. Although, I'm only running a 3.0Ghz Mesh and 3000 CL16 DDR4, so maybe that's the difference. :smug:

2nsncsk.png
 
Decided to fool around and got 5.03ghz reasonably stable. I say reasonaby because the ambient temp is less than 15c (I have my window open), so it definitely isn't a day to day overclock:

2mpit7t.png
 
Overclock temps are just HIGH man regardless...that's what I'm gonna try when the Loop comes in gonna hook up a 2nd rad to it, try that.
 
You have to delid period when running high oc.

All the exotic cooling doesnt help until you remove the insulator snd replace with functional material like Liq. Metal.
 
Mine is at 1.24v at 4.7 right now mesh at 3.0ghz with a small pos offset bump in voltage.

4.8 should be easy at 1.2x volts unless you lost the Si lottery.

Many folks are eother too aggressive or too under aggressive with thier load line calibration.

Remember LLC is a logarithmic increase in voltage in response to load until the limit you set.

Finding the sweet spot is where it is at.

I also run SpeedStep and C states and adaptive voltage.

100% stable in everything.

I am against setting a manual voltage as our chips are so advanced there is no need too do that anymore. Plus it potentially places undo wear and tear on your reg. Mosfets and the Si on chip.

Also dont forget to elevate your current limit.

For many auto OC works but its just not fulfilling and is rarely even ballpark in efficiency.
 
Last edited:
Hey, if anyone is using the MSI board, can you take a pic or two of your bios regarding the voltages and post them? I haven't used a MSI board in nearly half a decade and would be tankful for some crib sheets. :love:
 
Pics as promised.

Important note: my settings may be too much or too little for your CPU, depending on your luck. My stock volts are 1.2xx . But as always, it's pretty conservative. That's why i'm going with a negative offset while still overclocking the chip. Another thing i've found the hard way when searching for a 24/7 voltage is that the weaker cores are the ones demanding more voltage (doh right?). So basically putting a fixed voltage as one would do in the past (at least i did, and tried with this platform as well) would keep your better cores overvolted. WIth the voltage offset mode this is done automatically, based on your cpu's core paired in 2. From what i understand on some Asus boards you can control this voltage individually per core but i haven't found a way to do it on my MSI.

HWinfo64 is a great tool to help with monitoring different aspects, temps of cpu and vrm's, voltages, frequencies and throttling aspects.
 

Attachments

  • 1.JPG
    1.JPG
    230.4 KB · Views: 24
  • 2.JPG
    2.JPG
    193.9 KB · Views: 25
  • 3.JPG
    3.JPG
    172.4 KB · Views: 26
Pics as promised.

Important note: my settings may be too much or too little for your CPU, depending on your luck. My stock volts are 1.2xx . But as always, it's pretty conservative. That's why i'm going with a negative offset while still overclocking the chip. Another thing i've found the hard way when searching for a 24/7 voltage is that the weaker cores are the ones demanding more voltage (doh right?). So basically putting a fixed voltage as one would do in the past (at least i did, and tried with this platform as well) would keep your better cores overvolted. WIth the voltage offset mode this is done automatically, based on your cpu's core paired in 2. From what i understand on some Asus boards you can control this voltage individually per core but i haven't found a way to do it on my MSI.

HWinfo64 is a great tool to help with monitoring different aspects, temps of cpu and vrm's, voltages, frequencies and throttling aspects.

Thats funny that your bios reports the mesh as a RING? Thats odd that they would do that considering the 7820x doesn't use a ring bus.

I am curious why you are using a negative offset and yet gaining a positive increase over Vid?
 
Yeah, not sure why the mobo industry cannot have the same names for these kinds of things... And sometimes the description of them doesn't help either.

Because i'm actually going lower than my Vid, for most of my cores. My best cores run of 1.15x and the worst something like 1.21x. I'll have to see if delliding helps in this case, because in full load i've got a 16C difference between my coldest and hottest core.
 
Man, this is freaking hilarious. I just installed the 7820 with monobloc. I didn't bother to relid yet, after I get a baseline. Anyhow, I throw it all back in, and I accidentally boot into Windoze instead going to bios first, doh. Then when its ready I get back to bios and setup a base oc, and setup the rest of the bios, pump pwm, etc etc. I boot back into Windoze, and lol its already setup with not one question mark in Device Manager. Downtime = zero, coming from the outgoing Z170 Skylake board. I love when that happens, of course I do have to wipe it and load a fresh one with the correct key later. But for now, booyah!
 
Back
Top