AMD Ryzen Threadripper 2950X Overclocking

DethGasp

Junior Member
Feb 2, 2017
23
9
81
https://www.hardocp.com/article/2018/08/31/amd_ryzen_threadripper_2950x_overclocking/1

Once we got our cooling loop assembled, we used an ice chest and 50lbs of ice to keep our system cool. Even under fully threaded loads we never saw our coolant temperature get above 54F/12C.



We did all of this because we wanted to see if Precision Boost Overdrive would perform as AMD stated it would. Self Contained chiller systems are fairly expensive, as well as phase change systems and quite frankly we did not want to spend the money on one to use it for such a short duration. Also it is important to note, as you can see in the pictures above, we are dealing with condensation through the system as well as condensation down on the TR4 socket itself. This is not the safe way to do things, but we babysitting the system throughout the testing to make sure we kept everything dry that needed to be dry. Should you want to do something like this at home, you would need to do a lot of insulation work in and around the motherboard socket to make sure you would not have condensation down on the board
 
  • Like
Reactions: lightmanek

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,483
14,434
136
They don't compare it to anything else, and when/where can I buy wraith ripper ?
 
  • Like
Reactions: Drazick

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Most of those were at 4.2 ghz, so what are you talking about ?
These. The clocks are all over the place and even the 2950x wasn't entirely stable at 4.2Ghz over any meaningful amount of time. It's like the TR reviews all over again, when lots of lofty claims were made but only a handful of chips could even run stable 24/7 at 4.1Ghz.
We used the 2990WX set at 1.375 vCore and 1.1vSOC for the 4GHz scores that we shared here today. If you noticed the missing Blender Gooseberry score, the system would immediately shut down when trying to run it. I think this was the OverCurrent Protection (OCP) on the PSU kicking in. Running the 2990WX at 4GHz with Prime95 SmallFFTs, we were able to realize a ~590 watt CPU Package Power and saw ~1069 watts being pulled at the wall.
Jesus!

We still have not had the time to figure out a "bullet proof" overclock on the 2990WX at 4GHz, if it is even possible with ambient water cooling. We have more work to do here. Talking to Sami Makinen at AMD, he did not think we could be successful above 4GHz with ambient water cooling, and quite frankly, he is probably very correct.
So I remember how people were making fun of Intel running 28 cores at 5Ghz with a chiller. Seems AMD may need something similar at only 4Ghz. The irony.

We used the same 1.3725 vCore and 1.1v vSOC voltages for overclocking the 2950X to 4.2GHz/3400MHz across all 16 cores. We can run Prime95 for a good 20 minutes before this will crash at 4.2GHz. We likely need to roll back both these voltages a good bit and truly dial in the voltages this for the processor which we have not done yet. That said, we have no issues running any real workloads at 4.2GHz and 3400MHz on the 2950X.
Yeah, these things aren't stable at 4.2Ghz. Maybe for some benching but nothing near stable. So, even the 16 Core 2950x is having trouble above 4Ghz - again.

We still have a lot of work to do on the overclocking front with both the 2950X and the 2990WX CPUs. Overclocking the 2990WX is going to come down to the motherboard you are using, the power supply you have plugged into it, and of course it is going to come down to cooling. Even with our best TR4 CPU water blocks from the likes of XSPC, HeatKiller, and Koolance, overclocking the 2990WX to 4GHz is going to be a tall order using an ambient temperature system. Again, we need to look closely at getting the vCore down as low as possible as you can see above the overall system power increases exponentially as we start to push up the vCore. I do think we can get a bullet proof 4.2GHz out of the 2950X with a bit of effort. Our current water cooling lineup should be very capable of this. Of course we want to get a stable version of Ryzen Master to work with as well and see what Precision Boost Overdrive can do for us.
So, taken together, I think I wasn't far off with my original statement. Granted, these are early days but we hardly give Intel that break so I'm just going to take these numbers as proof until I see better from other reviews.
 
  • Like
Reactions: CHADBOGA

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,483
14,434
136
These. The clocks are all over the place and even the 2950x wasn't entirely stable at 4.2Ghz over any meaningful amount of time. It's like the TR reviews all over again, when lots of lofty claims were made but only a handful of chips could even run stable 24/7 at 4.1Ghz.
Jesus!

So I remember how people were making fun of Intel running 28 cores at 5Ghz with a chiller. Seems AMD may need something similar at only 4Ghz. The irony.

Yeah, these things aren't stable at 4.2Ghz. Maybe for some benching but nothing near stable. So, even the 16 Core 2950x is having trouble above 4Ghz - again.

So, taken together, I think I wasn't far off with my original statement. Granted, these are early days but we hardly give Intel that break so I'm just going to take these numbers as proof until I see better from other reviews.
And I have benchmarks@4.1 and have had no problems at all, and I was only using 1.325 vcore. But I guess you don't read any of my posts.

And the 2950's are stable@4.2

Why don't you leave to comments up to people who own or have used these chips.
 

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
And I have benchmarks@4.1 and have had no problems at all, and I was only using 1.325 vcore. But I guess you don't read any of my posts.

And the 2950's are stable@4.2

Why don't you leave to comments up to people who own or have used these chips.
I do read your posts. Unfortunately, benchmarking at 4.2Ghz is not the same as being stable at 4.2Ghz. The last time I checked your thread, you stated you were running at 3.7Ghz. Are you now able to run 4Ghz+ daily? Here's @Fir whose 2990WX locks up immediately @ 4.1Ghz while loading blender benchmark. Again, my statement is based on the links in this thread, and I've quoted extensively from the article already. As for your last statement, since when did one have to own a piece of hardware before they qualify to comment on it?
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
I do read your posts. Unfortunately, benchmarking at 4.2Ghz is not the same as being stable at 4.2Ghz. The last time I checked your thread, you stated you were running at 3.7Ghz. Are you now able to run 4Ghz+ daily? Here's @Fir whose 2990WX locks up immediately @ 4.1Ghz while loading blender benchmark. Again, my statement is based on the links in this thread, and I've quoted extensively from the article already. As for your last statement, since when did one have to own a piece of hardware before they qualify to comment on it?

You're in a 2950X thread conflating it with 2990WX overclocking. Tread lightly.
 

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
And I was replying to your post number 9 here that started by badmouthing all the 2x series Ryzens, in a long winded rant.
Why, you don't like the quotes from the article? I didn't write it. Check the title of the thread and links please.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,483
14,434
136
Why, you don't like the quotes from the article? I didn't write it. Check the title of the thread and links please.
The quotes you included do NOT come from the link in the OP, and there is no mention of the 2990wx in that article.. And in this article, 4.2 was stable, and power draw was 450 watts at the wall. Post number 5 is where you got your quotes from, and frankly, they are idiots, because I have not had any of the issues they did. @4.1 mine draws 546 watts from the wall with a 1080TI and under load.

Now stick to 2950x discussions.
 

dnavas

Senior member
Feb 25, 2017
355
190
116
At the risk of going the other direction, I'm preparing for my 2950x build by recasing my 1800X into an R5 from the Antec Solo it's been in -- I'm testing for sufficient airflow and noise quality. In the move, my DVD writer appears to have decided to prevent proper boot and shutdown process (plugged into the sata power cable, the machine will get to Windows login only 1/3 of the time, and will fail to shutdown roughly the other third). During the debugging process (I certainly had no inkling that an optical drive could fail this way), I noticed that my box had become horrifyingly unstable. It turns out my CPU and memory overclocks had been disabled, and that every time the 1800 tried to drive over 4Ghz single-core, that process crashed (goodbye youtube!). Has anyone else seen this on their Ryzen processors? It certainly wasn't the behavior of the processor back in March, though I was no silicon winner for sure. I'm running at 3.85 right now, rock steady.... Might have been the 3200 RAM XMP/"OC" as well, not sure.

This makes me slightly leery of overclocking Ryzen2, though I'd surely love to get the benefits of running at least PBO. the article speaks about setting up 300/300/300 -- but I'm not sure what those three numbers refer to. Isn't one of the points on the triangle temperature? 300C is a bit high, while 300K is a bit low, so ... ?
 

KompuKare

Golden Member
Jul 28, 2009
1,012
923
136
So I remember how people were making fun of Intel running 28 cores at 5Ghz with a chiller. Seems AMD may need something similar at only 4Ghz. The irony.
At the risk of continuing down a mostly off-topic discussion, what I remember about Intel's 28c@5Ghz stunt was that Intel PR tried to hide how they got there and lot's people (general readers and tech 'journalists') reporting it as being a possible new retail product coming soon.

The criticism was the way Intel PR presented it, not that the process wasn't up to it. It's well known that Intel's 14+/14++ process can get close to 5GHz, but equally it is well known how much power that would take. So it was easy for tech enthusiast to know that something about that PR stunt wasn't right, just a pity so many 'tech' websites didn't see straight through it.
 
  • Like
Reactions: Campy

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,483
14,434
136
At the risk of going the other direction, I'm preparing for my 2950x build by recasing my 1800X into an R5 from the Antec Solo it's been in -- I'm testing for sufficient airflow and noise quality. In the move, my DVD writer appears to have decided to prevent proper boot and shutdown process (plugged into the sata power cable, the machine will get to Windows login only 1/3 of the time, and will fail to shutdown roughly the other third). During the debugging process (I certainly had no inkling that an optical drive could fail this way), I noticed that my box had become horrifyingly unstable. It turns out my CPU and memory overclocks had been disabled, and that every time the 1800 tried to drive over 4Ghz single-core, that process crashed (goodbye youtube!). Has anyone else seen this on their Ryzen processors? It certainly wasn't the behavior of the processor back in March, though I was no silicon winner for sure. I'm running at 3.85 right now, rock steady.... Might have been the 3200 RAM XMP/"OC" as well, not sure.

This makes me slightly leery of overclocking Ryzen2, though I'd surely love to get the benefits of running at least PBO. the article speaks about setting up 300/300/300 -- but I'm not sure what those three numbers refer to. Isn't one of the points on the triangle temperature? 300C is a bit high, while 300K is a bit low, so ... ?
My 2700x's work almost better with OC, the new auto-boost works great ! As far as the 1800x goes, I had nothing but problems at anything over 3850, so its interesting thats where you are now, I would say the hardware is fine.
 
  • Like
Reactions: dnavas and Drazick

DrMrLordX

Lifer
Apr 27, 2000
21,583
10,785
136
Yeah, these things aren't stable at 4.2Ghz. Maybe for some benching but nothing near stable. So, even the 16 Core 2950x is having trouble above 4Ghz - again.

What? From the article:

We used a 1.376 vCore and stock vSOC voltages for overclocking the 2950X to 4.2GHz/3400MHz across all 16 cores. We can run Prime95 with Small FFTs for up to 8 hours before we will freeze the system at the desktop. Quite frankly, we are calling this a win...and subsequently stable. In our initial review, I thought we would be able to pull back on the voltages and still get long term Prime95 stability, but that was not the case. The screenshot below shows us 3200MHz RAM, but we were able to get it dialed in at 3400MHz with 16-16-16-36-1T timings at 1.45v. The RAM is rated for 3200MHz at 1.25v.

You pulled your quote from elsewhere in the article where they basically say, if it goes for 20 minutes, they're okay with it. Never mind that it's stable out to 8 hours with those settings!

I'm still personally not THAT impressed - I'd rather see 4.4 GHz out of the 2950X as an all-core overclock with good water. Granted their water cooling system is still allowing temps as high as 83C, which is pretty bad actually. I get annoyed when my NH-D15 allows temps in the high 70s on my 4 GHz Ryzen. If I could find a better air cooler, I would use it! Not ready to go water just for an old 1800x.
 

Elfear

Diamond Member
May 30, 2004
7,096
640
126
My 2700x's work almost better with OC, the new auto-boost works great ! As far as the 1800x goes, I had nothing but problems at anything over 3850, so its interesting thats where you are now, I would say the hardware is fine.

Same here. I went from a 1700 that wasn't fully stable at 3.9Ghz and 1.38V to a 2700X that boosts to 4.35Ghz out of the box (on what appears be 1-3 cores). When all cores are under a heavy load it drops to 4.1Ghz at ~1.325V. I've been very impressed with Ryzen v2.
 
  • Like
Reactions: lightmanek