Adventures in 2700X Overclocking

i'm finding undervolting the 2700x may lead to better performance. I under volted with offset - 0.1000 and instead of hitting 1.45v+ its running at max 1.344 temps are better less power draw but iam still running 4ghz on all cores and its hitting its 4.3ghz on low workloads. I want to try undervoltaging a little more see if i can get the power consumption and temps down.
 
os2wiz said:
I have a 2700X running on an X370 Tutanium. I keep getting thermal shutdowns when benchmarking or stress testing. The cpu temp will shoot up to 100 Celcius in cinebench 15 or Real Bench. I can not go above 3.8 GHZ when running cinebench 15 without the thermal shutdown. My Alphacool Eisbaer was ultracool before pooping in this cpu never going above 65 celcius under full load. My pump speed and fans on the cooler are working fine so I believe either the latest MSI Titaniujm bios is faulty in reading cpu sensors and thus causing shutdowns OR the cpu utself is faulty.
Click to expand...

This sounds bad, you can check if the cpu is making contact with the HSF? Maybe try the stock cooler?
 
Have you guys tried bclk o/c? Apparently as long as you stay under 105 bclk, it works in tandem with the PB2 o/d and xfr2 boosting single core up to 4.5ghz and all core to 4.2ghz, all while retaining the power gating and idle ability. That's pretty incredible, tbh. Makes me want to get an x370 or 470 and an AIO and go nuts...
 
bobzdar said:
Have you guys tried bclk o/c? Apparently as long as you stay under 105 bclk, it works in tandem with the PB2 o/d and xfr2 boosting single core up to 4.5ghz and all core to 4.2ghz, all while retaining the power gating and idle ability. That's pretty incredible, tbh. Makes me want to get an x370 or 470 and an AIO and go nuts...
Click to expand...

My board doesn't have a clock gen. So no bclk OC for me.
 
bobzdar said:
Have you guys tried bclk o/c? Apparently as long as you stay under 105 bclk, it works in tandem with the PB2 o/d and xfr2 boosting single core up to 4.5ghz and all core to 4.2ghz, all while retaining the power gating and idle ability. That's pretty incredible, tbh. Makes me want to get an x370 or 470 and an AIO and go nuts...
Click to expand...

Very interesting observation. Although you would have to make sure the Ram speed is stable too since overclocking the bclk will also increase memory frequency. INTERESTING......makes me wonder if you can get 4.5ghz on more than just 1 core.
 
im on 2700x running stock settings and - 0.1000 offset and my cpu will hit 4733mhz on a core for a breif moment when doing like light office work or running browser playing youtube in the background. I'm using triple monitors so i multple stuff running on each screen
 
Pieter3dnow said:
This sounds bad, you can check if the cpu is making contact with the HSF? Maybe try the stock cooler?
Click to expand...

Had absolutely nothing to do wwith the cpu. Bad Pinnacle Ridge bioses on many boards are causing this. As soon as it went into my new MSI X470 Gaming M7 AC there were no problems. Agesa code for X370 bios updates was flawed.
 
123Lanoix said:
im on 2700x running stock settings and - 0.1000 offset and my cpu will hit 4733mhz on a core for a breif moment when doing like light office work or running browser playing youtube in the background. I'm using triple monitors so i multple stuff running on each screen
Click to expand...

Mine were all clocking to 4.2GHz or so right out the box fluctuating between 1.4.1.45v - with the stock Wraith Prism cooler. Waiting for an h150i to come in stock as my old h100i didn't come with brackets for AM4.

Just seemed crazy to me that it's running voltages that are considered unsafe by many, straight out of the box.

Couldn't get it to boot with Offset of any lower than -.1v so I have it set manually; with Offset -.1v it was still hitting up to 1.45v.

It was doing 3.8-4GHz at 1.225v which I need to tinker with as it was fine then went unstable and wouldn't take it.

It's now at 1.25v manual and doing 4-4.2GHz fluctuations.
 
For me, typical usage falls between 3975 at the low end and 4225 at the high end, on all cores. Boosts to 4350 are occasional, and super quick, and never involve more than 2 cores - usually only 1.
 
i have a samsung b-die 3200 CL 14 g,skill x-flare kit. I changed it to 3600mhz 1.35v 16-16-16-36-52 T1 timings wow got more performance out of the chip i hit over 4800 on the multi test in cpu-z and 470 in single
 
i think you would benefit more from just getting the fastest ram kit you can and use that with a 2700x than trying to overclock it memory speed to really open this chip up alot !
 
123Lanoix said:
i think you would benefit more from just getting the fastest ram kit you can and use that with a 2700x than trying to overclock it memory speed to really open this chip up alot !
Click to expand...
Glad to see you are seeing what it is capable of! Looking forward to see how far you can crank things. I imagine 3600 would be the sweet spot for you, but I'd be curious if 3800 or 4000 speeds are possible with looser timings.
 
yeah i think i might have to up the voltage to 1.38v or 1.4 for those frequencies on the ram
 
iDShaDoW said:
Mine were all clocking to 4.2GHz or so right out the box fluctuating between 1.4.1.45v - with the stock Wraith Prism cooler. Waiting for an h150i to come in stock as my old h100i didn't come with brackets for AM4.

Just seemed crazy to me that it's running voltages that are considered unsafe by many, straight out of the box.

Couldn't get it to boot with Offset of any lower than -.1v so I have it set manually; with Offset -.1v it was still hitting up to 1.45v.

It was doing 3.8-4GHz at 1.225v which I need to tinker with as it was fine then went unstable and wouldn't take it.

It's now at 1.25v manual and doing 4-4.2GHz fluctuations.
Click to expand...


I have gone back to Stock auto as it appears to give me the best performance. my temps are about the same. I think my ambient temp were high when i initial recorded the temps
 
123Lanoix said:
yeah i think i might have to up the voltage to 1.38v or 1.4 for those frequencies on the ram
Click to expand...
You'll want to give the soc voltage a little nudge. That may help more for stability than the extra dram voltage. Not too much though. They say 1.2 is fine. Some say the absolute safe max is 1.3, but I keep it to 1.2 in case of voltage reporting errors and what not.
 
The 14/12 LP nodes were already at their peak with the 1800X they squeezed some extra out but it comes at a low reward gain and more voltage which I don't trust modern motherboards with, I have overclocked my 4970K like a handful of times to the point I was like meh, the real world gains are not noticeable on a CPU overclock, RAM still shows notable realworld improvement and SSD performance are notable. When I gen my 2700X it will run stock with the stock fan in my ITX setup.
 
123Lanoix said:
i have a samsung b-die 3200 CL 14 g,skill x-flare kit. I changed it to 3600mhz 1.35v 16-16-16-36-52 T1 timings wow got more performance out of the chip i hit over 4800 on the multi test in cpu-z and 470 in single
Click to expand...
Seen a few now at 3600, they really fly!
As soon as I saw an early leak with higher stable ram speeds, I realized quite a large amount of the advantage of this spin can be found in memory changes. Even though you lose a bit of latency on ram, the IF speed increase in the CPU seems to help in most workloads.
These chips remind me a little of the Athlon 64s.
 
Neapolitan6th said:
You'll want to give the soc voltage a little nudge. That may help more for stability than the extra dram voltage. Not too much though. They say 1.2 is fine. Some say the absolute safe max is 1.3, but I keep it to 1.2 in case of voltage reporting errors and what not.
Click to expand...

According to the stilt, 1.1v max for best ram speeds. Above that and stability decreased in all of the chips he tested. 88% got 3433 speed, 67% got 3566. Max speed achieved at <=1.1vsoc for all of them. I don't think it'll hurt anything to go over that, but I don't think it'll help get max speeds.
 
  • Like
Reactions: N4CR
like this
i went back to 3200@CL14 Memtest kept throwing errors at 3600 CL16


3200 CL14 is actually faster than 3600 CL16 you would need to run 3600 CL15 to get lower latency
 
OrangeKhrush said:
The 14/12 LP nodes were already at their peak with the 1800X they squeezed some extra out but it comes at a low reward gain and more voltage which I don't trust modern motherboards with
Click to expand...

One has to remember the high performance node is 14HP. The 14LPE/LPP nodes were designed for efficiency. Glofo already got rather high clocks from the 14LPP node and up to ~250MHz more from 12LP.
 
I like this thread. I'll have to try some of these things when I get home!
 
123Lanoix said:
i went back to 3200@CL14 Memtest kept throwing errors at 3600 CL16


3200 CL14 is actually faster than 3600 CL16 you would need to run 3600 CL15 to get lower latency
Click to expand...
Well it's a give and take...

At 3600 CL16, your bandwidth is going to be higher, so for things that need a lot of throughput speed to accomplish their task, will obviously end up finishing faster.
At 3200 CL14, as you said your latency is what is superior by being lower, so jobs that are doing lots and lots f small things, they will knock them out quicker thanks to the quicker access time.

From there, the question becomes what are you doing with your computer, and which is going to yield higher performance for you? If I recall correctly, games actually kinda like both, so 3400 CL15 would no doubt be the sweet spot. [someone please correct me if this is wrong!]

Since I believe you said to be using a G.Skill 3200 CL14 AMD kit, there's a fair chance that you could get away with one lower CL, so in your case CL13.
At least, that's what I was able to achieve on my Trident Z kit (non-AMD), which is 3200 CL15, which I'm running perfectly fine (sig system) at 3200 CL14. My SoC voltage is like 0.975V and DRAM is 1.36V. (I've gotten CL13 into Windows and through AIDA CacheMEM tests, but it's not useably stable as I recall)
Memtest 200% 14-14-14-34 Tight Subs.png
 
Will be building a rig for my bro's girl, have a question if anyone can help out, save me some time?
(work mostly, got busy)

Can i OC the RAM while having PB2 enabled? Before you say yes, i mean manually OC the RAM, change SoC (non Auto value), maybe change impedances if needed (again non Auto).
Haven't had time to read as much as i'd have liked to, so would appreciate the pointers.
 
Aenra said:
Will be building a rig for my bro's girl, have a question if anyone can help out, save me some time?
(work mostly, got busy)

Can i OC the RAM while having PB2 enabled? Before you say yes, i mean manually OC the RAM, change SoC (non Auto value), maybe change impedances if needed (again non Auto).
Haven't had time to read as much as i'd have liked to, so would appreciate the pointers.
Click to expand...

Why do that ? Use Samsung b die and decide if you want low latency or higher frequency... Other than that, stay stock for this CPU as it will manage better that way.
Also, building a computer for someone else you should always go the reliability route and not the [H] route, my 2 cents... save hassle down the road !
 
MMitch said:
Why do that ?
Click to expand...

You probably misunderstood me; why do what, OC the RAM? Why wouldn't i? A lot better than any "XMP", and 99% chance with lower voltages to boot.
Minimum stress on the chip, a hell of a benefit given IF's nature. Bought her a Flare-X 2x8gigs, 3200-14-14-34 kit for just that reason.

Anyway, not my first rodeo, or the second; just haven't had enough time to read up on all the changes, initial plan was proper OC, but given the nature of PB2, i'm thinking i can afford to be lazy, if i can do a manual on the RAM simultaneously :)

Now back on topic, stock can be a lot of things, it's why i was being specific. I can see a scenario where PB2 goes bye bye the minute it sees an altered SoC; i can easily foresee an opposite scenario where it continues to function no matter what, as long as i have it on 'enabled'.
(if you're wondering 'why not try it out and let us know', if i had the time, i already would have, lol. Only thing i haven't bought is a cooling solution, hence the enquiry here)

* Worst case scenario i'll stay up some more after i'm done tonight, lol
 
Well, I got my 2700X and currently wait for the 32GB FlareX along the Asus ROG x470 so I can't comment on actual real world result BUT... seems that every reviewer says PB2 is godlike and since it isn't your rig, I would recommend sticking with XMP + PB2. Remember you'll be the first contacted and/or blamed for any failure.... also it's painful to manage RMA for someone else...

There's so little to gain on a 3rd party build doing OC on this platform that I can't understand why you're asking ? I would understand if that was only to "test" then stock.. but OC Ryzen+ on a family computer is looking for trouble and phone calls lol
 
Formula.350 said:
Well it's a give and take...

At 3600 CL16, your bandwidth is going to be higher, so for things that need a lot of throughput speed to accomplish their task, will obviously end up finishing faster.
At 3200 CL14, as you said your latency is what is superior by being lower, so jobs that are doing lots and lots f small things, they will knock them out quicker thanks to the quicker access time.

From there, the question becomes what are you doing with your computer, and which is going to yield higher performance for you? If I recall correctly, games actually kinda like both, so 3400 CL15 would no doubt be the sweet spot. [someone please correct me if this is wrong!]

Since I believe you said to be using a G.Skill 3200 CL14 AMD kit, there's a fair chance that you could get away with one lower CL, so in your case CL13.
At least, that's what I was able to achieve on my Trident Z kit (non-AMD), which is 3200 CL15, which I'm running perfectly fine (sig system) at 3200 CL14. My SoC voltage is like 0.975V and DRAM is 1.36V. (I've gotten CL13 into Windows and through AIDA CacheMEM tests, but it's not useably stable as I recall)
View attachment 69333
Click to expand...



my 3200 CL 14 did boot into windows and stuff @ 3400 CL 14 stock voltage 1.35 but i didn't test it I know i have samsung b-die maybe i try this now
 
MMitch said:
Well, I got my 2700X and currently wait for the 32GB FlareX along the Asus ROG x470 so I can't comment on actual real world result BUT... seems that every reviewer says PB2 is godlike and since it isn't your rig, I would recommend sticking with XMP + PB2. Remember you'll be the first contacted and/or blamed for any failure.... also it's painful to manage RMA for someone else...

There's so little to gain on a 3rd party build doing OC on this platform that I can't understand why you're asking ? I would understand if that was only to "test" then stock.. but OC Ryzen+ on a family computer is looking for trouble and phone calls lol
Click to expand...
You're still misunderstanding him lol

He is asking purely about RAM clock speeds. On Zen+ anything over 2933MHz for RAM speed is considered an "Overclock", because it's outside the rated specifications.
Even then, if he had DDR4-3200 and wanted to try for DR4-3333, that would also be considered an "Overclock" to the RAM, since it's beyond the rated speed.
So any way you look at it, he is using the proper terminology in this situation :p

Thus, he is curious about: if he wants to run his at say DDR4-3400 (which is overclocked), along with changed timings, SoC voltage and/or RAM voltages, will any of that cause PB2 to not operate?


To answer his question, based solely on my understanding and usage of my 1700X: Yes, you can fool around with your RAM and SoC while still retaining Boost functions. I believe the only thing that causes the boost clocks to stop functioning is by changing the CPU Multiplier to anything other than Auto. You can play around with the P-State multipliers though, and still retain PB functionality. Which even if changing the CPU Core Voltage was needed by you, most motherboards (except MSI, for certain) seem to offer an "Offset" option, which basically tricks the CPU into thinking it's running at stock voltage when in reality it's being fed Default Voltage +/- the Offset amount.

Hope that helps, Aenra!
 
  • Like
Reactions: Aenra
like this
I tried 3400 CL 14 but it must throws errors in memtest so i guess i'll just stick to 3200 CL 14
 
123Lanoix said:
I tried 3400 CL 14 but it must throws errors in memtest so i guess i'll just stick to 3200 CL 14
Click to expand...
What's wrong with 3400 CL15?
And are you only changing the CAS timing, or are you changing all the primaries? IE 14-14-14 to being 15-15-15?
 
i haven't tried 3400 CL15 yet. do thing its even worth it? yes i changed all the timings so i would change it to 15-15-15-36-52
 
It's looking like 3400 @ 15-15-15-35-51 1.35v stable running memtest86. I'm gonna let it continues running while I sleep and check for errors in the morning. If it doesn't have bc any errors then it's good to go.
 
Not quite sure what "do thing its even worth it" was meant to say, but if you were asking is it worth it, or will it work, then of course :) Early Ryzen had issues with Odd number timings, but that was fixed. The numbers we use correspond to a bunch of complicated things behind the scenes, but what it boils down to is it's a reference for the BIOS to determine a time in milliseconds to refresh the RAM at, which just like CPU cycles that happen at an interval of time... when calculated with a reference clock, equates to a frequency.

Hmm alright, I may not have really worded that well enough (it's 12:33am, cut me some slack lmao) but basically 15 is nothing more than another divsor the RAM utilizes from a massive list of them.

As I mentioned earlier, my kit is actually advertised as 15-15-15-35. Origially, Ryzen couldn't do Odd CAS, so I either had to use either 14-15-15 or 16-15-15. It was later fixed, but I just ended up figuring out the needed subtimings to get 14-14-14 stable :D

While these days we work in whole values, once upon a time we had half-values. Like with DDR1 there could be RAM with timings of 2.5-3-3-4.


Anyways yea, good luck, I hope it pans out! I think there's a 3466 divisor as well, so if 3400 CL15 is good, given 3466 a shot next.
 
Last edited:
my motherboard will only do 16-15-15-15-35-51 1.35v when i set the first timing to 15 it automatically changes it to 16 and it won't post with 14-15-15-15-35-51 3400mhz 1.35v will there be a performance difference?
 
Formula.350 said:
Yes, you can fool around with your RAM and SoC while still retaining Boost functions!
Click to expand...

Finally! A person that can speak English!
Thank you oh savior! Jeesus, lol

And guys? Honestly? When someone asks a very specific, utterly technical question? For some very specific reasons, which he even goes on to state?
If you don't know the answer, just don't reply, lol
Don't be that guy!
 
Aenra said:
Finally! A person that can speak English!
Thank you oh savior! Jeesus, lol

And guys? Honestly? When someone asks a very specific, utterly technical question? For some very specific reasons, which he even goes on to state?
If you don't know the answer, just don't reply, lol
Don't be that guy!
Click to expand...

Amen, brother.
 
  • Like
Reactions: Aenra
like this
ok so i ran memtest86 and got 3466mhz 16-15-15-15-35-51 1.35v working no errors for about 1 1/2 hr i will run a longer test tonight but the face errors didn't immedtately pop up seems like its stable.
 
I'm considering my first AMD build in well over a decade and am a little confused as well. It sounds like with Ryzen 2700x the way to go is to leave the chip multiplier alone, rely on PB2 (even though it hertz the [H] in me), and instead run the memory as fast as possible. So I'm thinking that the Gigabyte (a brand I like anyway) Gaming 7 X470 is the way to go for me because it supports up to DDR4-3600. It also has the fancy onboard audio I want.

Looking at the QVL for the board, they list F4-3600C17Q-32GTZR (a 4x8GB TridentZ kit) for 3600MHz. I only want 16GB, because I don't need more and because LOL ram prices at the moment. But the 16GB kit isn't on the list.

Would it be safe to assume that if I picked up F4-3600C17D-16GTZR (the equivalent 2x8GB TridentZ) it would work? Also, this is G.Skill's memory for Intel and they have announced AMD memory for later, which ends in 'X' instead of 'Z'. Does that matter? I've never bought AMD/Intel specific memory before!

Thanks for any insight.
 
enc0re said:
I'm considering my first AMD build in well over a decade and am a little confused as well. It sounds like with Ryzen 2700x the way to go is to leave the chip multiplier alone, rely on PB2 (even though it hertz the [H] in me), and instead run the memory as fast as possible. So I'm thinking that the Gigabyte (a brand I like anyway) Gaming 7 X470 is the way to go for me because it supports up to DDR4-3600. It also has the fancy onboard audio I want.

Looking at the QVL for the board, they list F4-3600C17Q-32GTZR (a 4x8GB TridentZ kit) for 3600MHz. I only want 16GB, because I don't need more and because LOL ram prices at the moment. But the 16GB kit isn't on the list.

Would it be safe to assume that if I picked up F4-3600C17D-16GTZR (the equivalent 2x8GB TridentZ) it would work? Also, this is G.Skill's memory for Intel and they have announced AMD memory for later, which ends in 'X' instead of 'Z'. Does that matter? I've never bought AMD/Intel specific memory before!

Thanks for any insight.
Click to expand...

That sounds odd. I'd double check with Gigabyte on that. On the Asus QVL, you could generally run LESS memory of a specific SKU at the highest rated speeds (I'm assuming this is the exact same SKU, just in a two pack, not four), but NOT more (they indicated this with a dot under each number of populated DIMMS, 1, 2, or 4). But I am not sure if Gigabyte does that or not. I'd say it will probably work - almost definitely - but I would double-check with Gigabyte first.
 
You must log in or register to reply here.
Back
Top