Adventures in Ryzen 1700X Overclocking

D

DuronBurgerMan

[H]ard|Gawd
Joined
Mar 13, 2017
Messages
1,340
Now that the system has been up and running for a while, it is time to experiment with overclocking.

To recap, these are the system specs:

Ryzen 1700X
Noctua 12S cooler
Asus X370 Prime mobo
G.Skill Ripjaws 32GB (2 DIMMs) running at 2933
Plextor MP8e 1TB NVMe SSD
Geforce 1080 Ti
And some other sh*t

So far, here are my results:
  • 3800 MHz across all cores stable @ 1.4 vcore. Passes Intel Burn Test (10 passes) & Cinebench.
  • 3850 MHz across all cores appears stable @ 1.41 vcore. Passes Intel Burn Test (10 passes) & Cinebench.
  • 3900 MHz across all cores a bit iffy @ 1.42 vcore. Failed after 6 passes of Intel Burn Test. Will try adjusting vcore to 1.43.
  • 3900 MHz across all cores stable, but hot @ 1.43 vcore. Passes Intel Burn Test (10 passes) but temps close to 90C. Not sure if this includes the +20C Ryzen bug or not. Will have to investigate.
  • 4000 MHz across all cores @ 1.45 vcore. POSTs and boots to Windows. Temps too high on air cooler. This vcore is too much for Noctua 12S
Will continue to update as I investigate further.
 
Last edited:
Update: 3850 MHz fully stable at 1.41 vcore. Successfully passed Intel Burn Test (10 Passes) and Cinebench.

Update: 3900 MHz stable @ 1.43 vcore, but very hot. Anybody know for sure if Ryzen Master displays the proper temps or if it suffers from the +20C temp reading bug?
 
DuronBurgerMan said:
Update: 3850 MHz fully stable at 1.41 vcore. Successfully passed Intel Burn Test (10 Passes) and Cinebench.

Update: 3900 MHz stable @ 1.43 vcore, but very hot. Anybody know for sure if Ryzen Master displays the proper temps or if it suffers from the +20C temp reading bug?
Click to expand...

The 1700X and the 1800X both have the 20C offset.
1700 Does not.

And I'm pretty sure Ryzen Master does have that bug unless they updated it.
 
Last edited:
CAD4466HK said:
The 1700X and the 1800X both have the 20C offset.
1700 Does not.

And I'm pretty sure Ryzen Master does have that bug unless they updated it.
Click to expand...

It definitely makes a difference here. 3900 MHz is stable at 1.43 vcore... but temps in Intel Burn Test hit 91C. That's too high. But if +20C offset is skewing the result... and it's really 71C, then I'm golden, and I can try for 3950. Is there an application known to be correct with the 20C offset?
 
DuronBurgerMan said:
It definitely makes a difference here. 3900 MHz is stable at 1.43 vcore... but temps in Intel Burn Test hit 91C. That's too high. But if +20C offset is skewing the result... and it's really 71C, then I'm golden, and I can try for 3950. Is there an application known to be correct with the 20C offset?
Click to expand...
The only one I've heard of to be "close" to accurate is the Corsair link software.
 
DuronBurgerMan said:
It definitely makes a difference here. 3900 MHz is stable at 1.43 vcore... but temps in Intel Burn Test hit 91C. That's too high. But if +20C offset is skewing the result... and it's really 71C, then I'm golden, and I can try for 3950. Is there an application known to be correct with the 20C offset?
Click to expand...

Yeah that temp looks good.
AMDMatt had this to say about temps.

Overclocking voids warranty, remember that. Cannot provide you with an approximate guideline for temp vs voltage. Best practice would be to stay below the maximum recommended temperature.

The press deck applies and is a good guideline to follow. That said voltages up to 1.55v can be used if you have a robust liquid cooling solution to keep things cool.

Maximum temperature for the 1700 is 75C. 95C for the 1700X and 1800X with the -20c offset.
Click to expand...
 
Some additional tweaking yields the following:

  • 3850 stable at 1.4 vcore. Intel Burn Test and Cinebench passed. This will probably be my daily/normal OC. Temps very good.
  • 3950 stable at 1.44 vcore. Intel Burn Test and Cinebench passed. VERY hot. Ryzen Master says 95C. Even if this is a case of the +20C offset, it's too hot. I can probably get away with this for occasional benchmarking fun, but not daily use.
  • 4000 would probably be stable under water cooling. It POSTs and boots into Windows just fine.
The CPU has more overclocking legs in it, but temps becoming a problem. Air cooling on this chip tops out around 3950. 4000 or maybe even higher would be possible on better cooling. If I had better cooling, I'd try for 4100 at a more aggressive vcore.
 
Here's the 3.85 Cinebench result... the irritating thing is, I can't seem to beat 1699 without bumping up a notch. I wanted an even 1700 on one!

cinebench.jpg
 
DuronBurgerMan said:
Some additional tweaking yields the following:

  • 3850 stable at 1.4 vcore. Intel Burn Test and Cinebench passed. This will probably be my daily/normal OC. Temps very good.
  • 3950 stable at 1.44 vcore. Intel Burn Test and Cinebench passed. VERY hot. Ryzen Master says 95C. Even if this is a case of the +20C offset, it's too hot. I can probably get away with this for occasional benchmarking fun, but not daily use.
  • 4000 would probably be stable under water cooling. It POSTs and boots into Windows just fine.
The CPU has more overclocking legs in it, but temps becoming a problem. Air cooling on this chip tops out around 3950. 4000 or maybe even higher would be possible on better cooling. If I had better cooling, I'd try for 4100 at a more aggressive vcore.
Click to expand...
Looks like your 1700x behaves much like mine in the OCing department. I am surprised you got to DDR 4 2933 with 32gb, most seems to be fall short there with 2x 16gb sticks. Are you using any LLC for Vdroop? That can take make temperatures go high but can give stability. The problem with LLC is that it can spike your voltage rather high when the cpu load suddenly drops, especially if using an aggressive LLC setting.

Also IBT is the test that pushes Ryzen over the edge quickly if you have any instabilities. Great info!
 
noko said:
Looks like your 1700x behaves much like mine in the OCing department. I am surprised you got to DDR 4 2933 with 32gb, most seems to be fall short there with 2x 16gb sticks. Are you using any LLC for Vdroop? That can take make temperatures go high but can give stability. The problem with LLC is that it can spike your voltage rather high when the cpu load suddenly drops, especially if using an aggressive LLC setting.

Also IBT is the test that pushes Ryzen over the edge quickly if you have any instabilities. Great info!
Click to expand...

Yeah... it seems 1700X runs up against a thermal wall somewhere in the 3.9GHz range. For the RAM, I combed the motherboard QVL, and found exactly *one* configuration that would net me 32GB in two DIMMs at 2666. However, it was a 3000 part. I loaded the DOCP 3000 profile, and it worked (albeit at 2933). That was a nice surprise, as I was only expecting the 2666 from the QVL.

I am not using LLC. Temps seem to be the biggest problem I'm facing. I can keep stability at 3950 with vcore at 1.44, but it's too hot for my liking. 4000 would probably be worse, of course. But I imagine LLC would increase temps, not reduce them.

However, I'm not sure water cooling would have been worth it anyway. From what I've read, these CPUs top out around 4.1-4.2, even with great water cooling and aggressive vcore. So somewhere around 3.9 on air is pretty good. Not sure it's worth the extra cost to push the last 200 MHz or so out of the chip.
 
DuronBurgerMan said:
Yeah... it seems 1700X runs up against a thermal wall somewhere in the 3.9GHz range. For the RAM, I combed the motherboard QVL, and found exactly *one* configuration that would net me 32GB in two DIMMs at 2666. However, it was a 3000 part. I loaded the DOCP 3000 profile, and it worked (albeit at 2933). That was a nice surprise, as I was only expecting the 2666 from the QVL.

I am not using LLC. Temps seem to be the biggest problem I'm facing. I can keep stability at 3950 with vcore at 1.44, but it's too hot for my liking. 4000 would probably be worse, of course. But I imagine LLC would increase temps, not reduce them.

However, I'm not sure water cooling would have been worth it anyway. From what I've read, these CPUs top out around 4.1-4.2, even with great water cooling and aggressive vcore. So somewhere around 3.9 on air is pretty good. Not sure it's worth the extra cost to push the last 200 MHz or so out of the chip.
Click to expand...
4100mhz to 3900mhz is only 5% (4000mhz to 3900mhz is 3% :meh:) - I agree that is a heavy price for 5% which you would never notice except some minor differences in a benchmark. Sounds like your cpu is a better OCer then min - 3.9 is pretty much my ceiling with stability but with some heavy voltages. It will go into windows higher but that would only be for show and not to lean on. I am settling out around 3.8ghz for the time being with 3200 ram.
 
It's worth noting that @ 3.9GHz, this CPU outperforms the 1800X in almost every benchmark, sometimes by a pretty hefty margin. Ryzen isn't a bad overclocker by any stretch... it's just that the single core boost + XFR speed is likely to be close to the top for *ALL* cores after OCing. Benefits are tangible. After all, this is +500 MHz over base frequency. Not bad at all.

Trying an all day workload @ 3.9GHz/1.43 vcore to see if fully stable and temps okay. If it passes the test, that will be my normal operating overclock.
 
DuronBurgerMan said:
It's worth noting that @ 3.9GHz, this CPU outperforms the 1800X in almost every benchmark, sometimes by a pretty hefty margin. Ryzen isn't a bad overclocker by any stretch... it's just that the single core boost + XFR speed is likely to be close to the top for *ALL* cores after OCing. Benefits are tangible. After all, this is +500 MHz over base frequency. Not bad at all.

Trying an all day workload @ 3.9GHz/1.43 vcore to see if fully stable and temps okay. If it passes the test, that will be my normal operating overclock.
Click to expand...

Really the biggest knock on Ryzen overclocking is they all seem to be more or less the same the same chip, with perhaps a tiny bit better binning for the upper models. So that makes the 1700 a pretty great overclocker, the 1700x a decent overclocker, and the 1800x a potato overclocker.
 
DuronBurgerMan said:
Some additional tweaking yields the following:

  • 3850 stable at 1.4 vcore. Intel Burn Test and Cinebench passed. This will probably be my daily/normal OC. Temps very good.
  • 3950 stable at 1.44 vcore. Intel Burn Test and Cinebench passed. VERY hot. Ryzen Master says 95C. Even if this is a case of the +20C offset, it's too hot. I can probably get away with this for occasional benchmarking fun, but not daily use.
  • 4000 would probably be stable under water cooling. It POSTs and boots into Windows just fine.
The CPU has more overclocking legs in it, but temps becoming a problem. Air cooling on this chip tops out around 3950. 4000 or maybe even higher would be possible on better cooling. If I had better cooling, I'd try for 4100 at a more aggressive vcore.
Click to expand...

What I found is that you can run easily run Noctua fans (at least mine) full speed. It does make a difference for me on my FX 8320E overclock (low profile Noctua NH-C12P)
 
FrozenSun said:
So that makes the 1700 a pretty great overclocker, the 1700x a decent overclocker, and the 1800x a potato overclocker.
Click to expand...
I got my 1800X to 4.05GHz, which is something (though not the crazy gains from other CPUs). The temps were in the 75C range (under load), which seemed too high to me so I set back to stock. However, now I remember the 20C offset, meaning it was only really at 55C which could be doable.

CInebench saw a tangible score increase but all my games are GPU limited (at 4K) so there isn't much reason to overclock at the moment.
 
Ryzen Master on 1700X and 1800X *includes* the +20C. So whatever you see on there, subtract 20C and you have your temp.

All of the chips have a 75C thermal cutoff. It just "looks" like it's 95C on 1700X and 1800X due to the offset.

If you're ever in doubt, just crank it up and watch the temps. When you trip the thermal shutdown, whatever temp you see = 75C lol

I also feel as though the offset was intentional, and not exactly to mislead anyone either, just was the easiest (albeit hackiest) way to accomplish getting "high end chips" out that have a higher TDP. Ryzen wasn't really intended from the start to be this high of a clocked product and dollars to donuts is that the 1700 WAS the top end part initially. When they found out they needed a bit more legroom, and had the clock headroom as well, they went and created the offset to trick the systems since it'd take too long for the motherboard manufacturers to code a BIOS that accomplished the same thing, thereby artificially widening that thermal window. I suspect if the Ryzen 5 X chips are on the same stepping, they'll also feature it. If they're on a new stepping and have normal temp readout (no offset), then it may actually point to a leakage problem that they just had to deal with in order to get Ryzen shipped. [/speculative opinion]

All I can say is my lil Scythe Katana III I have on my 1700X, when it says 70C on my onboard readout, nothing actually feels like 70C, it feels like 50C. AIDA64's Stress Test when running at Stock everything, reads that my "CPU Package" is at 104W. What that is referring to, I dunno. Could be Power Draw, could be Thermal Output *shrug*.
 
My Titanium motherboard does seem to show the correct temp (maybe from it's own sensors) as it will read around 35C (I've seen as low as 32C) just idle on the BIOS. The Corsair H110i is really solid.
 
Ryzen 7 1700X
Cryorig A40 Ultimate
Asus Crosshair VI Bios 1001
G.Skill Ripjaws F4-3200C14D-16GVK 16GB 14-14-14-14-34 1T

Achieved 4GHz max temp of 76C (As reported in HWiNFO Beta)
@ Offset +.10625v and LLC Level 4 (HWiNFO reports 1.44v under load, haven't checked with a DMM for accuracy)
Bsclk 109
Target DRAM Frequency 2933, actual 3197MHz
 
RaptormanUSMC said:
Ryzen 7 1700X
Cryorig A40 Ultimate
Asus Crosshair VI Bios 1001
G.Skill Ripjaws F4-3200C14D-16GVK 16GB 14-14-14-14-34 1T

Achieved 4GHz max temp of 76C (As reported in HWiNFO Beta)
@ Offset +.10625v and LLC Level 4 (HWiNFO reports 1.44v under load, haven't checked with a DMM for accuracy)
Bsclk 109
Target DRAM Frequency 2933, actual 3197MHz
Click to expand...

Very nice. Confirmation that with water cooling, 4+ GHz is definitely possible. 3.9-3.95GHz appears to be my maximum on air cooling. But it boots into windows just fine at 4 GHz & 4.1GHz.... and I suspect it would be stable (at least at 4) if I could just get rid of the heat more effectively.
 
cybereality said:
My Titanium motherboard does seem to show the correct temp (maybe from it's own sensors) as it will read around 35C (I've seen as low as 32C) just idle on the BIOS. The Corsair H110i is really solid.
Click to expand...

See, I get idle temps around 55C on the AI Tweaker and in the BIOS. So it's gotta be the +20C offset. The heatsink doesn't even feel warm to the touch. It's room temperature.

Still, this offset issue is a bit irritating. It's been my only real complaint with the CPU thus far. AMD says "oh, this is easy, just subtract 20C." Sure, sure... but you don't know for sure if the BIOS and software are showing true value, or false one. So kind of annoying.
 
DuronBurgerMan said:
Very nice. Confirmation that with water cooling, 4+ GHz is definitely possible. 3.9-3.95GHz appears to be my maximum on air cooling. But it boots into windows just fine at 4 GHz & 4.1GHz.... and I suspect it would be stable (at least at 4) if I could just get rid of the heat more effectively.
Click to expand...

With my current voltage I tried 4.1GHz last night, made it to windows, but it crashed shortly after booting. Tried backing it down to 4.05GHz but that resulted in a boot loop and an overclocking failure. Seems my max is right at 4GHz at current voltage, and until we really know what voltages the Crosshair VI is pushing to the CPU (Lots of controversy on actual voltage being sent to the CPU. DMM is suppose to be close, but even that is skewed slightly for some) I'm not going to push it further.
 
RaptormanUSMC said:
With my current voltage I tried 4.1GHz last night, made it to windows, but it crashed shortly after booting. Tried backing it down to 4.05GHz but that resulted in a boot loop and an overclocking failure. Seems my max is right at 4GHz at current voltage, and until we really know what voltages the Crosshair VI is pushing to the CPU (Lots of controversy on actual voltage being sent to the CPU. DMM is suppose to be close, but even that is skewed slightly for some) I'm not going to push it further.
Click to expand...

At 4.1 GHz, mine boots into Windows fine, at 1.44 vcore. It *seems* stable, like I can go into applications and games, and run them okay. But it's so hot! Temps reported of around 90C under even a moderate load, which means probably 70C. When I run the Intel Burn Test, it quickly hits the thermal wall and crashes. Cinebench crashes as well, though not nearly so quickly as the Intel Burn Test. 4.0 GHz does similarly, though I can usually get a pass or two out of the Intel Burn Test before failure, and Cinebench just errors out rather than crashes the whole machine. So 4GHz is *almost* stable, and probably would be if I could get rid of the heat, and maybe up the vcore just a hair more. 3.95 is perfectly stable, runs all tests, but still runs hotter than I'd like for long term use. So 3.85 or 3.9 will probably be best, depending on my long term testing of 3.9.
 
Need liquid cooling to get anywhere above 4Ghz stable. Performance is capped by the Low Power process. Supposedly it will have a newer generation 14nm to improve that. Looks like they are targeting 7nm for EUVL for 2018, but the tools are not there yet.
 
Here are my 1700 results using Arctic Liquid Freezer 120:

3.9/2400:
3.9 GHz @ 1.325v
1200 MHz RAM
Max CPU Power: 127 watts
Max Temp Arctic Liquid 120: 60.8 C
Handbrake Re-encode Test: 47.5 fps
Cinebench: 1696

4.0/2400:
4.0 GHz @ 1.45v
1200 MHz RAM
Max CPU Power: 161 watts
Max Temp Arctic Liquid 120: 70.8 C
Handbrake Re-encode Test: 48.8 fps
Cinebench: 1731


The power needed to get to 4.0 kinda turned me off so I have been running 3.9. 4.05 and 4.1 will boot, but crashes shortly after any stress test starts.
 
starmax said:
Here are my 1700 results using Arctic Liquid Freezer 120:

3.9/2400:
3.9 GHz @ 1.325v
1200 MHz RAM
Max CPU Power: 127 watts
Max Temp Arctic Liquid 120: 60.8 C
Handbrake Re-encode Test: 47.5 fps
Cinebench: 1696

4.0/2400:
4.0 GHz @ 1.45v
1200 MHz RAM
Max CPU Power: 161 watts
Max Temp Arctic Liquid 120: 70.8 C
Handbrake Re-encode Test: 48.8 fps
Cinebench: 1731


The power needed to get to 4.0 kinda turned me off so I have been running 3.9. 4.05 and 4.1 will boot, but crashes shortly after any stress test starts.
Click to expand...

That's an amazingly low vcore for 3.9. Mine won't do anywhere near that low at 3.9.
 
starmax said:
Here are my 1700 results using Arctic Liquid Freezer 120:

3.9/2400:
3.9 GHz @ 1.325v
1200 MHz RAM
Max CPU Power: 127 watts
Max Temp Arctic Liquid 120: 60.8 C
Handbrake Re-encode Test: 47.5 fps
Cinebench: 1696

4.0/2400:
4.0 GHz @ 1.45v
1200 MHz RAM
Max CPU Power: 161 watts
Max Temp Arctic Liquid 120: 70.8 C
Handbrake Re-encode Test: 48.8 fps
Cinebench: 1731


The power needed to get to 4.0 kinda turned me off so I have been running 3.9. 4.05 and 4.1 will boot, but crashes shortly after any stress test starts.
Click to expand...

Are you not able to get your ram beyond 2400? Ryzen benefits much more in performance with faster ram clocks versus more cpu speed. If you're having a hard time getting beyond 2400 I may be able to help you.
 
My 1700x will do 3.9 at 1.35 but needs well in excess of 1.4 to do 3.95 stable. Not comfortable with that, so running at 3.9.
 
RaptormanUSMC said:
Are you not able to get your ram beyond 2400? Ryzen benefits much more in performance with faster ram clocks versus more cpu speed. If you're having a hard time getting beyond 2400 I may be able to help you.
Click to expand...

No, I have this crappy ASUS B350M-A (because I needed a microATX board). I have tried lots of settings but cannot get anything higher than 2400 to post. Frustrated....
 
starmax said:
No, I have this crappy ASUS B350M-A (because I needed a microATX board). I have tried lots of settings but cannot get anything higher than 2400 to post. Frustrated....
Click to expand...

If you can change your dram vboot set it to 1.40v and try faster speeds. Also, using bsclk to up ram timings has helped me.
 
I got my waterblock adaptor today from XSPC (Performance PCS). I am going to put it on tonight. I would be happy to get 3.8ghz if that on water with this chip from the look of it. I can get 3.7ghz on air at good temps but nothing more really.
 
tangoseal said:
I got my waterblock adaptor today from XSPC (Performance PCS). I am going to put it on tonight. I would be happy to get 3.8ghz if that on water with this chip from the look of it. I can get 3.7ghz on air at good temps but nothing more really.
Click to expand...

That seems pretty low. 3.8 and 3.85 are both very attainable at good temps on air for me. Do you have the vcore set too high, perhaps?
 
DuronBurgerMan said:
That seems pretty low. 3.8 and 3.85 are both very attainable at good temps on air for me. Do you have the vcore set too high, perhaps?
Click to expand...

No I have been afraid to push it higher. I haven't air cooled a CPU in 7 years literally until this chip. I guess I have zero faith in air coolers to do anything other than make noise. I have always ran a custom loop with massive amounts of radiator surface area through 7 x 120 mm worth of radiator space.
 
tangoseal said:
I got my waterblock adaptor today from XSPC (Performance PCS). I am going to put it on tonight. I would be happy to get 3.8ghz if that on water with this chip from the look of it. I can get 3.7ghz on air at good temps but nothing more really.
Click to expand...
Kinda sound like what I got :shame:. 3.8ghz, 1.4v and it can't pass IntelBurnTest on Maximum. I am going to test like what Kyle does, if it passes that kind of test it is most likely stable. So in limbo.
 
tangoseal said:
No I have been afraid to push it higher. I haven't air cooled a CPU in 7 years literally until this chip. I guess I have zero faith in air coolers to do anything other than make noise. I have always ran a custom loop with massive amounts of radiator surface area through 7 x 120 mm worth of radiator space.
Click to expand...

Pretty happy with the Noctua 12S cooler I put in this rig. It really is a quality unit. But it's still air cooling, and so I don't think I'm getting 4GHz out of the chip with air. 3.9 appears to be the limit.
 
Not 1700X but this is what I'm running currently:

Ryzen 1800X @ 4.0ghz 1.4v
Noctua NH-D15S cooler
ASRock Taichi X370
2x8gb G.Skill TridentZ 3200 CL14 DDR4 (Running XMP profile @ these settings)

Worst temperatures I've seen so far are low 70's when the +20C offset is taken into account. That was after temps leveled off in handbrake and Prime95 type testing after 20-30 minutes.
 
Magic Hate Ball said:
Not 1700X but this is what I'm running currently:

Ryzen 1800X @ 4.0ghz 1.4v
Noctua NH-D15S cooler
ASRock Taichi X370
2x8gb G.Skill TridentZ 3200 CL14 DDR4 (Running XMP profile @ these settings)

Worst temperatures I've seen so far are low 70's when the +20C offset is taken into account. That was after temps leveled off in handbrake and Prime95 type testing after 20-30 minutes.
Click to expand...

AMD said they binned the Ryzen 7s based more on voltage than clock speed, and I think we're starting to see the difference here. My 1700X will do that same 4 GHz, no problem (except heat), but it wants a lot more vcore than yours to do it. That D15S cooler is nice -- basically two of my 12S's in one tower -- but there was no way I was fitting that in the case I wanted!
 
DuronBurgerMan said:
AMD said they binned the Ryzen 7s based more on voltage than clock speed, and I think we're starting to see the difference here. My 1700X will do that same 4 GHz, no problem (except heat), but it wants a lot more vcore than yours to do it. That D15S cooler is nice -- basically two of my 12S's in one tower -- but there was no way I was fitting that in the case I wanted!
Click to expand...

Yeah, I have my D12's fan attached for a second fan currently. I have an NZXT x62 in the box awaiting the AM4 bracket to be shipped. They contacted me and said next week they should start going out.
 
Any tips?

Crosshair VI
It will go 4ghz on water (I blocked it last night).

It will not do ram using XMP at 3200 or 2999 or 2666 at that clock speed. I even bumled ddr voltage to 1.37v and it wont post. Im using LPX Corsair 16 18 18 36. I can get ram to clock to 3000 with stock cpu speed. This chip is absolutely frustrating. TPU is useless.

Any OC tips? Id want at least 2666 ram at 3.9ghz. Doable? Im true water cooled. Not one of those boxed somutions.
 
You must log in or register to reply here.
Back
Top