Ryzen 3950X, 3960X, 3970X Availability

tunatime

2[H]4U
Joined
Sep 15, 2011
Messages
3,235
Well, they're nowhere near TR benches but I did these two

View attachment 212938
That's not bad at all my 3960x gets 1350 or so stock and best I've gotten is little over 15.5k. What's your single core?

Also if you haven't yet play with your ram makes a big difference for me running just under 3800 at 14-14-14-28 and tight sub timings vs the lose 3600 xmp
 

IndyColtsFan

Limp Gawd
Joined
Jul 4, 2017
Messages
356

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
So...a couple of these + a 3D printed "cap" costs a lot less than a CableMod cable set. $7.50 versus $156 I need to remember that.

61utJSwNfvL._SL1200_.jpg
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
Pretty sure I can break 10k on Cinebench R20 if I could stomach the BIOS long enough to do more than change the multiplier and press F10, lol

CBScores.jpg


That's okay enough for me.

Need to drain the loop, remove a couple runs, remake them (they bother me but don't leak or anything) and add the flow meter spinny wheel thing
 
  • Like
Reactions: rUmX
like this

Zarathustra[H]

Official Forum Curmudgeon
Joined
Oct 29, 2000
Messages
29,461
So...a couple of these + a 3D printed "cap" costs a lot less than a CableMod cable set. $7.50 versus $156 I need to remember that.

View attachment 212959
What are these for?

I would highly recommend using any kind of splitter or adapter on either the EPS12v connectors going to the motherboard, or the PCIe power cables going to the GPU. The only safe way to populate EPS and PCIe connectors is with all pins wired all the way back to the PSU, with trusted brand wires and no adapters or extenders in between. (Ask me how I know)

I wouldn't use any extenders or splitters for these things. If the PSU doesn't support it, I'd buy a better PSU. If you can't make the wire routing work in your case, get a different case.

If you need to pinch the pennies, this is seriously not the place to do it.

Queue "Trough the Fire and Flames" by Dragonforce.
 

tunatime

2[H]4U
Joined
Sep 15, 2011
Messages
3,235
What are these for?

I would highly recommend using any kind of splitter or adapter on either the EPS12v connectors going to the motherboard, or the PCIe power cables going to the GPU. The only safe way to populate EPS and PCIe connectors is with all pins wired all the way back to the PSU, with trusted brand wires and no adapters or extenders in between. (Ask me how I know)

I wouldn't use any extenders or splitters for these things. If the PSU doesn't support it, I'd buy a better PSU. If you can't make the wire routing work in your case, get a different case.

If you need to pinch the pennies, this is seriously not the place to do it.

Queue "Trough the Fire and Flames" by Dragonforce.
Qft theses things pull some crazy amperage when you get above the 4.3+ range. That being said if it's made well it should work( I would be suspicious as they don't look that thick in that pick) but as you said get a better case I see about 1in of cable other then the connectors as there a cutout right next to the eps
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
What are these for?

I would highly recommend using any kind of splitter or adapter on either the EPS12v connectors going to the motherboard, or the PCIe power cables going to the GPU. The only safe way to populate EPS and PCIe connectors is with all pins wired all the way back to the PSU, with trusted brand wires and no adapters or extenders in between. (Ask me how I know)

I wouldn't use any extenders or splitters for these things. If the PSU doesn't support it, I'd buy a better PSU. If you can't make the wire routing work in your case, get a different case.

If you need to pinch the pennies, this is seriously not the place to do it.

Queue "Trough the Fire and Flames" by Dragonforce.

These are PCIe 8-pin adapters and they're perfectly safe for what they're intended to do. LOL.
 

Zarathustra[H]

Official Forum Curmudgeon
Joined
Oct 29, 2000
Messages
29,461
These are PCIe 8-pin adapters and they're perfectly safe for what they're intended to do. LOL.
That's what I thought when I bought my 8 pin extension cable at MicroCenter too...

Then this happened when under load for an extended period of time (video reencode)

196494_IMG_20190223_165838.jpg


196496_IMG_20190223_165913.jpg


And that was on my old 3930k. This Threadripper should draw a lot more current...

I don't trust any power adapters or extenders anymore...
 
Last edited:

Zarathustra[H]

Official Forum Curmudgeon
Joined
Oct 29, 2000
Messages
29,461
Outside of benchmarks like Cinebench, I haven't quite figured out how to load this thing up yet.

Right now, I'm dubbing a box of Video8 camcorder tapes in PAL25 format my parents recorded in the 80's, using OBS under Linux.

Using an EasyCap Analog composite in device, I have it set to capture at 1440x1152, downscale to 720x526 using SincLanczos, and compress with x264, VBR, Very Slow, 2500kbps CRF=20.

CPU is hovering at anywhere betwee 8% and 12% load, depending on what is going on...

It's too bad I can't play back these tapes at faster than 100% speed.

I remember the last time I did this back in 2001. It was a really difficult problem to solve. Had to stripe two 7200rpm hard drives in order to get enough write speed, and then real time compress the camcorder stream with HUFFYUV to make it both light enough that the highly overclocked Athlon XP 1800+ CPU could handle it in real time, and sufficiently compressed to be able to write it to the slow hard drives at the same time. It was a real challenge.

Now it is a piece of cake.
 
Last edited:

bobzdar

[H]ard|Gawd
Joined
Jun 6, 2003
Messages
1,736
Outside of benchmarks like Cinebench, I haven't quite figured out how to load this thing up yet.
Download 3df zephyr and run some photogrammetry reconstructions. It's highly threaded and maxes my 3950x and gtx1080 concurrently. Best whole system stress test I've found (and also the primary reason I went for a 3950). It has a built in benchmark but it's a lot more fun to take a bunch of pictures of something and reconstruct it - like your house, a room in it, car or something like that.

Maybe go meta and take a bunch of pics of your new system and create a 3d model of it.
 

mikeo

Gawd
Joined
May 17, 2006
Messages
624
B&H has the 3970x in stock for msrp, did everyone that preordered from there get theirs already?
 

somebrains

Gawd
Joined
Nov 10, 2013
Messages
898
B&H has the 3970x in stock for msrp, did everyone that preordered from there get theirs already?
Might as well cancel that order and preorder 3990x.
I’m curious how many small feeder studios have 3900x/3950x running as stopgap builds while they wait for TR3.
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
I'm evidently receiving 5 meters of white 16mm Bitspower PETG Saturday, thanks to Bitspower :)

Logo follows sometime later, I guess.

I'm cool with that!

Now I can use all the spare parts left over to do a hard line system for my 2700X (which I'm going to transplant from the Phanteks case to the O11 Dynamic)
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
Download 3df zephyr and run some photogrammetry reconstructions. It's highly threaded and maxes my 3950x and gtx1080 concurrently. Best whole system stress test I've found (and also the primary reason I went for a 3950). It has a built in benchmark but it's a lot more fun to take a bunch of pictures of something and reconstruct it - like your house, a room in it, car or something like that.

Maybe go meta and take a bunch of pics of your new system and create a 3d model of it.
That sounds like a cool program to play with.

I'm hoping that my build is finally completed this weekend so I can try that program out but, with the kind of storms predicted for tonight/tomorrow, I may not have electricity after tonight for several days.
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
Did you to a single threaded run in CB20?

I'm curious how much low threaded performance I have up by going with the Threadripper instead of a 3950x

System is currently in many pieces right now in preparation for remaking all hard line runs with the white PETG that's arriving tomorrow...

I had a week to think about routing of the loops so I've already re-made the tubes twice using the clear stuff I originally had, just to try out different routing/bends/etc...

I'm actually pretty happy with the results so far since I was able to remove EIGHT fittings from the loop by re-making the tubes. I also re-did the drain port and fill port (though the fill port is less than optimal due to space constraints.)

I'm also re-doing the pump mount and the plexi plate it mounts to, but hand sawing and polishing plexi is a time consuming process...


The USPS delay meant that when I tried to return the defective TT Pacific kits I was 3 days past the return date. At least I'll have enough mix-n-match crap TTparts leftover (from both kits) to do a single custom loop for my other computer...so that's an upside, I guess...


All this is moot if I lose power for 2-3 weeks or die from the severe weather arriving tonight.
 

Silentbob343

[H]ard|Gawd
Joined
Aug 2, 2004
Messages
1,772
System is currently in many pieces right now in preparation for remaking all hard line runs with the white PETG that's arriving tomorrow...

I had a week to think about routing of the loops so I've already re-made the tubes twice using the clear stuff I originally had, just to try out different routing/bends/etc...

I'm actually pretty happy with the results so far since I was able to remove EIGHT fittings from the loop by re-making the tubes. I also re-did the drain port and fill port (though the fill port is less than optimal due to space constraints.)

I'm also re-doing the pump mount and the plexi plate it mounts to, but hand sawing and polishing plexi is a time consuming process...


The USPS delay meant that when I tried to return the defective TT Pacific kits I was 3 days past the return date. At least I'll have enough mix-n-match crap TTparts leftover (from both kits) to do a single custom loop for my other computer...so that's an upside, I guess...


All this is moot if I lose power for 2-3 weeks or die from the severe weather arriving tonight.
Try not to die .

Joking aside, stay safe.
 

Zarathustra[H]

Official Forum Curmudgeon
Joined
Oct 29, 2000
Messages
29,461
System is currently in many pieces right now in preparation for remaking all hard line runs with the white PETG that's arriving tomorrow...

I had a week to think about routing of the loops so I've already re-made the tubes twice using the clear stuff I originally had, just to try out different routing/bends/etc...

I'm actually pretty happy with the results so far since I was able to remove EIGHT fittings from the loop by re-making the tubes. I also re-did the drain port and fill port (though the fill port is less than optimal due to space constraints.)

I'm also re-doing the pump mount and the plexi plate it mounts to, but hand sawing and polishing plexi is a time consuming process...


The USPS delay meant that when I tried to return the defective TT Pacific kits I was 3 days past the return date. At least I'll have enough mix-n-match crap TTparts leftover (from both kits) to do a single custom loop for my other computer...so that's an upside, I guess...


All this is moot if I lose power for 2-3 weeks or die from the severe weather arriving tonight.
Yeah, I don't know where you are, but hope the weather wasn't as bad as expected and you have no damage.
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
Try not to die .

Joking aside, stay safe.
Yeah, I don't know where you are, but hope the weather wasn't as bad as expected and you have no damage.
My area wasn't hit as hard as others who work on the base with me. The squall line was actually at its "thinnest" over my area of the city.

The weather station made the storm out to be of apocolyptic proportions but it seems to lose strength when it crossed the Mississippi River.

Got the Bitspower tubing and lemmetellyasumthin...that tubing is the most difficult tubing I've ever worked with in my whole life. That is not even a smidge of exaggeration.

I've never seen tubing of its like in my whole life. I've cut copper tubing, gas line, plexi, acrylic, rubber, glass, aluminum, brass, and about 40-leven other types of tubing in my life and this was, by far, the most difficult. Period.

You cannot cut it with a pipe cutter. Not even a brand new one. (mine is) This is the result of a 20 minute attempt to cut it. That's right. 20 frikkin minutes making 1 cut, that failed. Miserably.

The "brown" stuff is the oil on the metal cutting wheel of the new pipe cutter. I was able to get only 1/10th of 1mm into the pipe before it shattered.(the pipe, not my cutter)


IMAG0847.jpg


That's how far it made it in before the innter material just shattered and broke off. It's the most brittle material I've ever worked with. Eggshells are stronger. Again...not exaggerating.

Although that was a decent picture given my phone's capabilities, you can't see that when the inner part shatters, it snaps all the way to the cut-penetration line so there's literally 1/10th of a millimeter of wall material to seal with, so you aren't using that for anything that matters.

If you cut it with a dremel, even as low as the lowest setting, you get nothing but plastic particles in the air (50%) and fused together particles like welding slag all over the part. VERY difficult to work with a dremel because if you try to avoid the slag, the tool bogs down...If you up the speed to prevent bogging, the inner plastic of the tube bonds so well with the rest of the tube once it melts, that you can't remove it without chipping large gouges out of the pipe.

But...it can be worked with, and worked pretty easily, actually. You need to use a hack saw with a minimum of 24tpi (teeth per inch) and take it slowly. That works just fine actually...unless you don't have 24tpi or higher bimetal blades to use.

Again...use a hacksaw with a high TPI blade and everything will be right as rain. Wish I'd known that since it took a full meter of material to make my first pipe. (Literally...a straight pipe from the bottom radiator to the top)

It's also more difficult to bend properly. I needed to heat the tubes 2-3 times as long as normal Acrylic/PETG. It sets far faster too so you better have your shit together on exactly what angle you need.

That said, I was able to complete the loop. Here's the redesigned loop from the front. on the lower right side you can see my re-designed drain. T fitting to 90° to CheckValve to Cap. Same setup on top right but not on the "front" of the Rad because I can't move the rad farther back because of the IO Heatsink

IMAG0851.jpg



Here's a side shot of the case. (Don't mind Onoda Sakamichi. He falls a lot)

IMAG0839.jpg



Currently in hour 5 of leak test. Should be able to boot up tomorrow...but I'm re-working the wiring while it's leak testing.
 

Zarathustra[H]

Official Forum Curmudgeon
Joined
Oct 29, 2000
Messages
29,461
My area wasn't hit as hard as others who work on the base with me. The squall line was actually at its "thinnest" over my area of the city.

The weather station made the storm out to be of apocolyptic proportions but it seems to lose strength when it crossed the Mississippi River.

Got the Bitspower tubing and lemmetellyasumthin...that tubing is the most difficult tubing I've ever worked with in my whole life. That is not even a smidge of exaggeration.

I've never seen tubing of its like in my whole life. I've cut copper tubing, gas line, plexi, acrylic, rubber, glass, aluminum, brass, and about 40-leven other types of tubing in my life and this was, by far, the most difficult. Period.

You cannot cut it with a pipe cutter. Not even a brand new one. (mine is) This is the result of a 20 minute attempt to cut it. That's right. 20 frikkin minutes making 1 cut, that failed. Miserably.

The "brown" stuff is the oil on the metal cutting wheel of the new pipe cutter. I was able to get only 1/10th of 1mm into the pipe before it shattered.(the pipe, not my cutter)


View attachment 215241

That's how far it made it in before the innter material just shattered and broke off. It's the most brittle material I've ever worked with. Eggshells are stronger. Again...not exaggerating.

Although that was a decent picture given my phone's capabilities, you can't see that when the inner part shatters, it snaps all the way to the cut-penetration line so there's literally 1/10th of a millimeter of wall material to seal with, so you aren't using that for anything that matters.

If you cut it with a dremel, even as low as the lowest setting, you get nothing but plastic particles in the air (50%) and fused together particles like welding slag all over the part. VERY difficult to work with a dremel because if you try to avoid the slag, the tool bogs down...If you up the speed to prevent bogging, the inner plastic of the tube bonds so well with the rest of the tube once it melts, that you can't remove it without chipping large gouges out of the pipe.

But...it can be worked with, and worked pretty easily, actually. You need to use a hack saw with a minimum of 24tpi (teeth per inch) and take it slowly. That works just fine actually...unless you don't have 24tpi or higher bimetal blades to use.

Again...use a hacksaw with a high TPI blade and everything will be right as rain. Wish I'd known that since it took a full meter of material to make my first pipe. (Literally...a straight pipe from the bottom radiator to the top)

It's also more difficult to bend properly. I needed to heat the tubes 2-3 times as long as normal Acrylic/PETG. It sets far faster too so you better have your shit together on exactly what angle you need.

That said, I was able to complete the loop. Here's the redesigned loop from the front. on the lower right side you can see my re-designed drain. T fitting to 90° to CheckValve to Cap. Same setup on top right but not on the "front" of the Rad because I can't move the rad farther back because of the IO Heatsink

View attachment 215243


Here's a side shot of the case. (Don't mind Onoda Sakamichi. He falls a lot)

View attachment 215244


Currently in hour 5 of leak test. Should be able to boot up tomorrow...but I'm re-working the wiring while it's leak testing.
Looking sharp.

I feel like I am going to have to give hard tube a try next time around, if for no other reason than minimizing my use of swivel bends. Those things scare me
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
Looking sharp.

I feel like I am going to have to give hard tube a try next time around, if for no other reason than minimizing my use of swivel bends. Those things scare me
You should definitely give it a whirl.

Soft lines with barbs is a pretty quick setup process. Hard lines force you to take time planning the layout and I liked the challenge. I didn't like it when the tubes were shattering but that's an outlier experience. Had I had more experience with Bitspower tubes I'd have known to use a hacksaw from the time I opened the package.

That said, the failures were learning experiences and I was still able to do the whole loop with less than 2 tubes, so I have 3 meters left (more than three, actually...about 3.3 meters)

The largest run (the simplest and first one I tried) failed, twice, but the pieces were still viable to use for the other runs since I caught on quickly about using a hack saw.

I have way more than enough parts left over to transplant my current system (#2 below) into the Lian Li non-XL and do a hardline loop.

If I can migrate all the settings/tweaks over to the 3950X tomorrow I can start this coming week on the second system.

Since Lian Li is going to start selling a mini version I may do a third system (DTX?) in it.
 

blade52x

2[H]4U
Joined
Sep 9, 2006
Messages
3,021
Is the best way to OC these manually? Seems like PBO is supplying way too much voltage and has a high power draw. I was able to reduce 100W+ on my 3970x manually overclocking it to 4.2ghz all core boost.
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
Is the best way to OC these manually? Seems like PBO is supplying way too much voltage and has a high power draw. I was able to reduce 100W+ on my 3970x manually overclocking it to 4.2ghz all core boost.
That's a good question. If I use AUTO in BIOS for voltage control MSI wants to send 1.49V to my CPU for stock clocks. It'll run stock clocks at 1.2V all damn day. (Haven't tried going lower but I believe it could do it)

If I manually set voltage to 1.35v I can change the multi for an all-core overclock of 4.3GHz. At 1.37v I hit 4.375GHz

I use far less "power" when manually setting my voltage...sure...I don't see "boost clocks" of 4.6GHz but those are so random they don't contribute to performance in any meaningful way...

So I'll second the question and ask is it better to manually overclock?
 
  • Like
Reactions: mikeo
like this

thesmokingman

[H]ardness Supreme
Joined
Nov 22, 2008
Messages
5,699
Is the best way to OC these manually? Seems like PBO is supplying way too much voltage and has a high power draw. I was able to reduce 100W+ on my 3970x manually overclocking it to 4.2ghz all core boost.
Manually no. I'll leave it to the stilt to answer this.

In short: Leave the CPU alone, it knows what its doing. There is no guesswork involved when it comes to something as essential, as determining the default silicon characteristic specific voltage-frequency curve of a CPU.
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
Manually no. I'll leave it to the stilt to answer this.

I have a hard time swallowing that my CPU needs 1.49V based on the word of "random dude on the internet", no matter how accomplished he is at overclocking, when the CEO of AMD says no more than 1.4-ish for safe 24/7 operation
 

thesmokingman

[H]ardness Supreme
Joined
Nov 22, 2008
Messages
5,699
I have a hard time swallowing that my CPU needs 1.49V based on the word of "random dude on the internet", no matter how accomplished he is at overclocking, when the CEO of AMD says no more than 1.4-ish for safe 24/7 operation
You're assumptions on voltage on Zen is wrong. And the stilt is not some random guy lmao.
 

mikeo

Gawd
Joined
May 17, 2006
Messages
624
That's a good question. If I use AUTO in BIOS for voltage control MSI wants to send 1.49V to my CPU for stock clocks. It'll run stock clocks at 1.2V all damn day. (Haven't tried going lower but I believe it could do it)

If I manually set voltage to 1.35v I can change the multi for an all-core overclock of 4.3GHz. At 1.37v I hit 4.375GHz

I use far less "power" when manually setting my voltage...sure...I don't see "boost clocks" of 4.6GHz but those are so random they don't contribute to performance in any meaningful way...

So I'll second the question and ask is it better to manually overclock?
If you care about maximizing every bit of performance out of the chip in all core OC mode it seems like it.
 

bobzdar

[H]ard|Gawd
Joined
Jun 6, 2003
Messages
1,736
That's a good question. If I use AUTO in BIOS for voltage control MSI wants to send 1.49V to my CPU for stock clocks. It'll run stock clocks at 1.2V all damn day. (Haven't tried going lower but I believe it could do it)

If I manually set voltage to 1.35v I can change the multi for an all-core overclock of 4.3GHz. At 1.37v I hit 4.375GHz

I use far less "power" when manually setting my voltage...sure...I don't see "boost clocks" of 4.6GHz but those are so random they don't contribute to performance in any meaningful way...

So I'll second the question and ask is it better to manually overclock?
No - you'll burn your chip up doing that. It only hits 1.47 volts on a single core, 1.325V is the maximum for all core. Above that and you will impact silicon reliability - it's not the voltage that's the issue but the amount of current and localized heat the chip has to handle when all cores are running at that voltage. With a single core, it's not taking much current even at 1.47V and not generating a lot of heat. With all cores loaded, each CCX is taking a considerable amount of current and localized heat, even at 1.325V. You don't seem to fully understand how these chips run on auto, so you may want to do some more research before you start manually pushing that much voltage through all cores...Just seeing max voltage of 1.48 in hwinfo64 and thinking that's the max it can or what it puts out in all loads is wrong and using the flawed assumption to base your o/c voltage on will burn it up in short order.
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
No - you'll burn your chip up doing that. It only hits 1.47 volts on a single core, 1.325V is the maximum for all core. Above that and you will impact silicon reliability - it's not the voltage that's the issue but the amount of current and localized heat the chip has to handle when all cores are running at that voltage. With a single core, it's not taking much current even at 1.47V and not generating a lot of heat. With all cores loaded, each CCX is taking a considerable amount of current and localized heat, even at 1.325V. You don't seem to fully understand how these chips run on auto, so you may want to do some more research before you start manually pushing that much voltage through all cores...Just seeing max voltage of 1.48 in hwinfo64 and thinking that's the max it can or what it puts out in all loads is wrong and using the flawed assumption to base your o/c voltage on will burn it up in short order.
An excellent response. Thanks for the illuminating post. MSI's forums are, putting it mildly, less than helpful. I tried to find out why the board is trying to push 1.49v and your post explains that's to 1 core (if needed...not necessarily what it's actually getting 24/7) Makes sense. I can set my voltage back to Auto (from 1.35v) and be done with it. I'm okay with that.
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
You're assumptions on voltage on Zen is wrong. And the stilt is not some random guy lmao.
Based on the explanation by bobzdar you're correct that my assumptions were wrong. :)

bobzdar had a better "explanation" of why my assumption was wrong, though. :)

Thanks to both of you I have a better understanding of the all-core versus single core power issue. I appreciate it :)
 

blade52x

2[H]4U
Joined
Sep 9, 2006
Messages
3,021
No - you'll burn your chip up doing that. It only hits 1.47 volts on a single core, 1.325V is the maximum for all core. Above that and you will impact silicon reliability - it's not the voltage that's the issue but the amount of current and localized heat the chip has to handle when all cores are running at that voltage. With a single core, it's not taking much current even at 1.47V and not generating a lot of heat. With all cores loaded, each CCX is taking a considerable amount of current and localized heat, even at 1.325V. You don't seem to fully understand how these chips run on auto, so you may want to do some more research before you start manually pushing that much voltage through all cores...Just seeing max voltage of 1.48 in hwinfo64 and thinking that's the max it can or what it puts out in all loads is wrong and using the flawed assumption to base your o/c voltage on will burn it up in short order.
That only partially explains it for me. Here's on example as to why:

I'm focused on all core loading. I'm going to be hitting the CPU hard with video encoding, parallel Python algorithms, and machine learning. Running the R15 or R20 benchmarks results in an all core load. So even if 1.45v+ is only for one of the cores, Ryzen Master reports 300W+ during load with PBO on, versus 170-180W with a manual overclock of 4.2ghz at 1.3v. In these two cases benchmark scores are comparable yet the temperatures with PBO do get higher with the sustained load. So how exactly is running the former actually better?

I understand that reducing voltage can increase the current load, but that's if you stay at a constant power draw. Reducing 120W is a massive drop in a power draw so I'd assume current stays in check. I'll do more research myself, but right I still have a hard time believing a generalized algorithm to fit all chips in existence is the best solution to a single chip. The only benefit I see to using PBO/XFR2 is for high ST performance, and lower idle power usage.
 
Last edited:

bobzdar

[H]ard|Gawd
Joined
Jun 6, 2003
Messages
1,736
That only partially explains it for me. Here's on example as to why:

I'm focused on all core loading. I'm going to be hitting the CPU hard with video encoding, parallel Python algorithms, and machine learning. Running the R15 or R20 benchmarks results in an all core load. So even if 1.45v+ is only for one of the cores, Ryzen Master reports 300W+ during load with PBO on, versus 170-180W with a manual overclock of 4.2ghz at 1.3v. In these two cases benchmark scores are comparable yet the temperatures with PBO do get higher with the sustained load. So how exactly is running the former actually better?

I understand that reducing voltage can increase the current load, but that's if you stay at a constant power draw. Reducing 120W is a massive drop in a power draw so I'd assume current stays in check. I'll do more research myself, but right I still have a hard time believing a generalized algorithm to fit all chips in existence is the best solution to a single chip. The only benefit I see to using PBO/XFR2 is for high ST performance, and lower idle power usage.
It's hard to say without knowing what pbo settings you're using and what the resultant clocks are. If the scalar isn't touched it should be safe. I burnt up a 2700X with pbo, so I know it's not always safe...
 

thesmokingman

[H]ardness Supreme
Joined
Nov 22, 2008
Messages
5,699
Based on the explanation by bobzdar you're correct that my assumptions were wrong. :)

bobzdar had a better "explanation" of why my assumption was wrong, though. :)

Thanks to both of you I have a better understanding of the all-core versus single core power issue. I appreciate it :)
Here's a reddit post that collected some of the results of the degradation. The big issue is there's no way for end users to know how much their silicon can handle or how much latitude there is before degradation. That said in most cases its best to just run PB2, ie. stock with PBO disabled.

Regarding blades question, check your settings. Using PBO obviously raises power limits and whatever static voltage you set is not raising power limits.
 

notarat

2[H]4U
Joined
Mar 28, 2010
Messages
2,100
So...my system is coming apart again after I order some coolant and drain what's in there....sigh...

I forgot how inconvenient hard lines can be...making even simple build changes requires complete disassembly a lot of the time, depending on what needs to be changed...or where that tiny part to replace is located...

The left-hand LED in the XSPC Raystorm Pro is dead. In order to replace it I have to remove the top Radiator...

Oh...and it's b fandanna time. I got tired of making hats for fans...now I'm into fandannas...
 

Zarathustra[H]

Official Forum Curmudgeon
Joined
Oct 29, 2000
Messages
29,461
So...my system is coming apart again after I order some coolant and drain what's in there....sigh...

I forgot how inconvenient hard lines can be...making even simple build changes requires complete disassembly a lot of the time, depending on what needs to be changed...or where that tiny part to replace is located...

The left-hand LED in the XSPC Raystorm Pro is dead. In order to replace it I have to remove the top Radiator...

Oh...and it's b fandanna time. I got tired of making hats for fans...now I'm into fandannas...
Lol.

I would have just left the LED dead.

I never look inside my case except for when I'm building it :p
 
Top