4.0 limit on 2700x ?s

[dowhopdedodo]

Having trouble getting past 4.0 ghz on my new build using the Asus detailed guide (for 1800x, slower DRAM and older Asus bios GUI but suggested for Ryzen 2) as reference. https://youtu.be/pBmVf0S4UDs

I keep thinking it's the offset voltage, but after extensive testing (for me) I doubt that's it now. System is stable at .02500 offset at 4k, but unstable at 4.1 with even a much larger offset.

Maybe it's just my luck in the chip lotto, although when I saw the low voltage at stock I thought I'd have room to get to 4.2 without pushing voltage too much. CPU voltage shows 1.2125 on Ryzenmaster.

Temp at idle is 32-33C at 4k. I have been able to get to 4.1 on manual using 1.375v without excessive heat, but won't boot 4.2 even at 1.4v manual.

Not sure any of it matters. I've kind of set an arbitrary baseline of 4.2ghz at 3200 with a 1900 +/- Cinebench score, but realistically want to keep an offset OC without risking expensive CPU.

Am I missing something? Apparently Asus has a new, manual OC guide coming for Ryzen 2. Maybe I should just wait to see that before trying to go farther than the 4k I'm at. Thanks in advance.

AMD 2700x
Asus Prime 470 0222 x64 bios
G.SKILL Flare X 32GB (4 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200 (PC4 25600) F4-3200C14Q-32GFX
Corsair H60 2018 liquid
Nvidea Geforce GTX1060 Founders Ed
Seasonic Focus 650 gold
Samsung Evo 970 M.2
Corsair 270R box

 

[formulav8]

Honestly, people are getting better performance not overclocking via the multiplier and just upping the Bus a few Mhz. Some are getting almost 4.45ghz single/double core boost speeds. I would try that if your board has the Bus overclock option. You need to be careful though.

 

[SickBeast]

There seems to be some nonsense with the Ryzen CPUs where the reviewers are given golden sample chips that can hit higher than realistic overclocks. I have an 1800X and it's a horrible overclocker. I can't even get 3.9ghz out of it. But according to AMD, 4.2ghz should have been easy. It's good that we're getting some actual user experiences online. I was interested in the 2700X if it could hit 4.3ghz. I'm starting to realize that it can't unless you get a golden chip.

 

[Markfw]

I have 2 2700x's. I can't figure out the ASUS bios on one, so its stock. The other can run 4.2 ghz on my Taichi motherboard, but will lockup after 24 hours of 100% load, so its at 4.1 now.

 

[Hitman928]

There seems to be some nonsense with the Ryzen CPUs where the reviewers are given golden sample chips that can hit higher than realistic overclocks. I have an 1800X and it's a horrible overclocker. I can't even get 3.9ghz out of it. But according to AMD, 4.2ghz should have been easy. It's good that we're getting some actual user experiences online. I was interested in the 2700X if it could hit 4.3ghz. I'm starting to realize that it can't unless you get a golden chip.

Where did AMD say 4.2 GHz should be easy with 1st gen Ryzen?

As for reviewers, I don't think it's a matter of golden samples, but more a matter of very few reviewers properly test their overclocks before calling it good. Most of the time their reported overclocks would certainly eventually throw errors if they actually used the chips at those speeds 24/7. A lot of times, reviewers will also overclock on open test benches so their temps stay lower than if you actually put the system in a case which means they can usually use a little higher voltages before their temps go too high compared to the average overclocker.

For Ryzen, the limits are pretty clear and have been for a while. For 1st gen you're looking at 3.9 GHz +- 200 MHz depending on luck but most fall right about 3.9 GHz. Second gen Ryzen basically just shifts that up by 200 MHz.

As for OP, I agree with forumulav8, manual overclocking is actually causing most people to lose performance due to losing the higher boost clock. Unless you're doing something consistently heavily threaded like Markfw, it's better to leave at stock or slightly overclock the bus if possible.

 

[Markfw]

Where did AMD say 4.2 GHz should be easy with 1st gen Ryzen?

As for reviewers, I don't think it's a matter of golden samples, but more a matter of very few reviewers properly test their overclocks before calling it good. Most of the time their reported overclocks would certainly eventually throw errors if they actually used the chips at those speeds 24/7. A lot of times, reviewers will also overclock on open test benches so their temps stay lower than if you actually put the system in a case which means they can usually use a little higher voltages before their temps go too high compared to the average overclocker.

For Ryzen, the limits are pretty clear and have been for a while. For 1st gen you're looking at 3.9 GHz +- 200 MHz depending on luck but most fall right about 3.9 GHz. Second gen Ryzen basically just shifts that up by 200 MHz.

As for OP, I agree with forumulav8, manual overclocking is actually causing most people to lose performance due to losing the higher boost clock. Unless you're doing something consistently heavily threaded like Markfw, it's better to leave at stock or slightly overclock the bus if possible.

Yes, if I had the time to mess with ASUS bios and learn it, I would get the other one to 4.2. I firmly believe (I have 2 now to judge) that 4.2 is a reasonable OC, but since I run mine 100%load 24/7/365, I like lower temps, so I go with the "easy" OC of 4.1

 

[dowhopdedodo]

Honestly, people are getting better performance not overclocking via the multiplier and just upping the Bus a few Mhz. Some are getting almost 4.45ghz single/double core boost speeds. I would try that if your board has the Bus overclock option. You need to be careful though.

Would love to read/watch a tutorial on this if you happen to know of one. Thanks.

 

[dowhopdedodo]

Thanks for the replies. I feel a better being stable at 4k. Seems like most everything I was reading left the impression that 4.2ghz was standard OC.

My temps actually at 28c at idle and 58c under load, which is good now that 4k appears the standard give or take 100mhz. I'm not a bencher, at least not beyond my one week replacement build every three to five years. Just trying to get the most bang for the buck like everyone else. Haven't installed the apps I work with which, at least in some cases, is where the performance difference from my old 4.2ghz I5 Ivy Bridge will really show.

Would love to learn a bit more regards bus speed vs multiplier. Just not clear on what the means as a practical matter, avoiding errors and pitfalls. Any links would be appreciated.

Thanks again.

 

[formulav8]

Would love to read/watch a tutorial on this if you happen to know of one. Thanks.

Person's are simply raising the so-call "FSB" to 104-105mhz. That means, that when the Precision Boost multipliers kick in, the clock speeds will be higher.

For instance, instead of PB2 on 2700x being 43.5 x 100mhz, It's 43.5x X 104 mhz which ='s 4524mhz on 1-2 cores. The multi core boost will increase as well.

But you will need to see what your setup can handle. AMD improved Precision Boost with 2nd gen.

But by raising the Bus settings to 105 Mhz or so, some are not getting their M.2 storage devices detected. So it's a fine line that you need to troubleshoot and figure out for your own setup. But could be worth trying out for max single threaded performance if that's what you're wanting.

 

[Markfw]

Person's are simply raising the so-call "FSB" to 104-105mhz. That means, that when the Precision Boost multipliers kick in, the clock speeds will be higher.

For instance, instead of PB2 on 2700x being 43.5 x 100mhz, It's 43.5x X 104 mhz which ='s 4524mhz on 1-2 cores. The multi core boost will increase as well.

But you will need to see what your setup can handle. AMD improved Precision Boost with 2nd gen.

But by raising the Bus settings to 105 Mhz or so, some are not getting their M.2 storage devices detected. So it's a fine line that you need to troubleshoot and figure out for your own setup. But could be worth trying out for max single threaded performance if that's what you're wanting.

Speaking ONLY for the Taichi, its easy. Set vcore to 1.25, set speed to 4100, and set memory to equal or LOWER than its rated speed. Mine is 3600 and 4266, and both only run at 3466 reliably. Save and enjoy.

When I get time, I will mess with the ASUS, not so easy to figure out IMO

 

[formulav8]

I'm talking more about CPU clockspeed for lower-ish thread count uses like many games and tasks. I have neither a 2700x nor a board capable of changing the fsb frequency to test personally. But in regards to the 2700x if you plan on ocing, you can most likely get better single-ish core performance by only raising the 'fsb' to 104mhz or so and letting Precision Boost do the rest.

But IIRC with your personal uses, you are big on doing Distributed Computing tasks which work well with higher clocked multi-core clockspeeds and not single core so much.

If you have the desire, it may be worth a look at changing the bus speed only and letting precision boost set the multiplier. Would be interesting to know how it goes with your use case.

 

[dowhopdedodo]

If you have the desire, it may be worth a look at changing the bus speed only and letting precision boost set the multiplier. Would be interesting to know how it goes with your use case.

Helpful and intriguing. I'd enjoy testing this option. The issue that gives me pause is my low level of knowledge as well as a dearth of information using the current Asus uefi. As an example, both the printed and PDF manuals reference BCLK control (FSB) via the initial Ai Tweaker Auto/Manual control in the GUI. The current bios GUI display only offers Auto/D.O.C.P, however; which then allows you to select your RAM frequency/settings automatically but no BCLK access/control that I can discern. I could be missing something, however.

I'll keep an eye out with any future bios updates if this is the current state of the uefi

 

[mjdupuis]

I'd start from scratch at a manual voltage overclock. You've hit a wall which doesn't seem to make sense unless we assume it's just a really poor chip.
Leave the RAM at stock (not XMP) or even manually set at 2133, and then start trying for a high OC.
Your board may not be able to handle 4.2+ GHz AND 32GB at 3200. I could be wrong, but that's my hunch.

 

[cfenton]

Is BCLK OCing less annoying that it used to be? I used in on my Ivy Bridge system for a while, but it broke some USB devices (not permanently, they just didn't work while the base clock was higher). To me, it wasn't worth doing since I could never be sure if something wasn't working because of the BCLK OC.

 

[lightmanek]

My not so stellar sample of 2700X is doing fine @4.2GHz all core OC. It needs 1.395V to achieve daily stability, going to 4.3GHz is a no go with required voltage around 1.45V. I can't simply cool it enough to stay stable. It loads Windows and runs benchmarks in short bursts at that freq. but heat quickly climbs to above 85C and then system freezes.

I know few lucky ones can do 4.4GHz all core, but for that you really need a Golden Sample. My 1700 could do 4.05GHz pushed to the max. current 2700X can do 4.25GHz pushed to the max, so net gain of 200MHz, right around AMD estimate, or one may say I was equally unlucky with my CPU lottery

Edit: all above on the same ASUS X370-Pro Prime motherboard.

 

[dowhopdedodo]

Your board may not be able to handle 4.2+ GHz AND 32GB at 3200. I could be wrong, but that's my hunch.

I tend to agree, and I didn't really follow that sound advice when beginning my OC testing. I can't recall now if I backtracked later by doing only the manual CPU stepping. I'm sure I did a couple of tests that way because I have a saved profile, but probably did those with too little voltage to make the OC work.

I happened across this little clip on YT while looking for some info on BCLK OCing and given the content creation side of my excuse for buying a 2700X minimally applies nowadays, the reasoning in the video below seemed to make perfect sense, i. e. save on heat & energy costs by running the CPU at stock and running the ram at the rated clock rather than at 1.35v + for 200mhz. https://youtu.be/MS7FuMNJvGs

That said, my one or two hours of MMO gaming at night could benefit from a higher clock on single core as the app is single threaded and CPU reliant as I recall. But, absent access to BCLK on the latest bios GUI I'm sure I'll survive, more or less, better than I would playing on my 4.2 OC'd 2012 Ivy Bridge.

Speaking of videos, and I realize you guys are all pros here, this second, older and more in-depth analysis seemed relatively supportive of the notion that the 2700x's real improvement is the far lower voltage required for essentially the same speed. Admittedly, I was quite surprised to see 1.212v at stock on my chip (1.275 at 4ghz when pressed). https://www.youtube.com/watch?v=1A2yatfyLoo

Thanks for all the support and info.

 

[Yuriman]

Maybe I'm in the minority here but I find it pretty hard to get excited about a 5% overclock. Less than 15-20% and I just don't bother anymore, because the only place I can tell a difference is in synthetic benchmarks (and at the wall, because power consumption goes up so quickly).

 

[LTC8K6]

Maybe I'm in the minority here but I find it pretty hard to get excited about a 5% overclock. Less than 15-20% and I just don't bother anymore, because the only place I can tell a difference is in synthetic benchmarks (and at the wall, because power consumption goes up so quickly).

Yes, Ryzen seems to automatically overclock about as much as it can, so it seems pointless to work for that extra 100-200mhz.

 

[PG]

1 - Update the bios. You are on the original bios and there have been 2 updates since then. The latest has some important updates.

2 - The Prime is the entry level X470 board. It has weaker mosfets and worse heatsinks than the upper level boards. The mosfet heatsinks need some airflow, but you have a water cooler which provides none. This is not ideal. The stock cooler would push air over the correct heatsink but you aren't using it.
Make sure you have good case airflow, and in particular some air moving of the mosfet heatsinks, especially the one to the left of the socket. That is the side with the mosfets for the cores.
It's not a bad board or anything, but you need airflow.

 

[rbk123]

Maybe I'm in the minority here but I find it pretty hard to get excited about a 5% overclock. Less than 15-20% and I just don't bother anymore, because the only place I can tell a difference is in synthetic benchmarks (and at the wall, because power consumption goes up so quickly).

Yes, Ryzen seems to automatically overclock about as much as it can, so it seems pointless to work for that extra 100-200mhz.

Agree with both. It's a very nice chip but overclocking isn't one of it's strengths.

 

[mjdupuis]

That said, my one or two hours of MMO gaming at night could benefit from a higher clock on single core as the app is single threaded and CPU reliant as I recall. But, absent access to BCLK on the latest bios GUI I'm sure I'll survive, more or less, better than I would playing on my 4.2 OC'd 2012 Ivy Bridge.

From my experience with gaming on my former 2600X, you may get better results with letting it automatically boost since your single core boost speed will be much higher on default settings than an all-core OC.

In fact, I ended up going with heavily massaged RAM speeds and timings and just leaving my CPU to boost on it's own. My board and/or chip didn't want to go past 4.1 without the voltage seeming unreasonable, but was happy to boost past 4.2 all day long while leaving things at stock.

 

[dowhopdedodo]

Make sure you have good case airflow, and in particular some air moving of the mosfet heatsinks, especially the one to the left of the socket. That is the side with the mosfets for the cores.
It's not a bad board or anything, but you need airflow.

Yeah, 3 120mms atm, pulling from front pushing from back. Cooler radiator fan pulls air down from above. Thanks for the info. I'd seen a review with similar complaint.

 

[dowhopdedodo]

Yes, Ryzen seems to automatically overclock about as much as it can, so it seems pointless to work for that extra 100-200mhz.

Yup, just don't need that extra power consumption for a small boost and all cores at all times for what I'm doing. If I was doing more rendering maybe, but not so much these days.

Funny you mention auto because I've taken opportunity to keep CPU-Z up during day to day activities and noticed it was boosting to 4.2 at times under load, which was a huge surprise (1.42v +\- as well). Was very pleased seeing that. Happy camper here.

 

[dowhopdedodo]

In fact, I ended up going with heavily massaged RAM speeds and timings and just leaving my CPU to boost on it's own. My board and/or chip didn't want to go past 4.1 without the voltage seeming unreasonable, but was happy to boost past 4.2 all day long while leaving things at stock.

I've seen one review/example where there is some room to massage RAM clocks although I've not attempted that (yet, or perhaps not).

 

[ancient pedant]

Maybe I'm in the minority here but I find it pretty hard to get excited about a 5% overclock. Less than 15-20% and I just don't bother anymore, because the only place I can tell a difference is in synthetic benchmarks (and at the wall, because power consumption goes up so quickly).

This^

My 2700x is running quite nicely on Precision Boost, some cores reaching up past 4.1 on occasion, and doing it very quietly and economically, staying cool, on air. I'm quite pleased.