SOME RYZEN 7 1700 OVERCLOCKING!

[gavinh87]

Was using realbench, but still got black screen issues with that (its some luxmark issue that comes up with nvidia), so had enough of that and just sticking to Aida64 now.

There are two ways realbench dies with black screen though - one is hard system crash, where it requires shutting down. The other is luxmark/gpu related where I can just unplug the hdmi cable and plug it back in to see the luxmark crash message.

it's easy enough to tell which by whether my keyboard lights are on or off.

Though, I was going down that road of too much tinkering again - all I wanted from this was 4Ghz, so that is all I am aiming for now without pushing anything to what I consider mental levels - I do want this to last a while

I recommend OCCT for stability mate. Or intel burn test, both use linpack. Realbench has been a complete **** for me too nvidia related as you are aware of.
I found AIDA not so good for CPU but really good for memory.

 

[roysters]

what if you're not bothered about overclocking and just wan't the best performing chip, wouldnt the 1800x at stock be the best option?

 

[Smudge]

what if you're not bothered about overclocking and just wan't the best performing chip, wouldnt the 1800x at stock be the best option?

They all overclock to basically 1800X levels so yes, if money is no object then an 1800X is the way to go for guaranteed top performance. It's very easy to clock the 1700/1600 to similar clocks as the 1800X though with a few bios settings and save some money. With the 1800X you may want to also set the bios settings for it to boost all cores to 4GHZ, so why not just get a 1700 and do the same.

 

[roysters]

Thanks for the info

 

[gupsterg]

what if you're not bothered about overclocking and just wan't the best performing chip, wouldnt the 1800x at stock be the best option?

My R7 1700 was ~£300 delivered, after flogging the Wraith Spire (net me £43) = £257. Now would I pay for a 1800X? nope. 3.7GHz ACB is not out of reach of CPU OC results I have seen.

A spreadsheet I have been collating data for R7.

The other thing is in stock mode all the Ryzen CPUs tend to go high on voltage when XFR/PB occurs. Seen lot's of posts of members querying is it normal? is it ok for CPU? you would think so as it's stock operation. When members OC the CPU, it doesn't seem to hit as high a voltage, as XFR is gone. I prefer that then stock operation TBH.

 

[gavinh87]

Well I've pretty much confirmed I have a runt of a 1700 :/
3.9 still isn't stable at 1.45! It copes with prime and realbench aida but OCCT and IBT kills it.
3.8 at 1.37 looks like the most I'm going to achieve. I've already posted the cinebench results, the difference is not going to be noticeable in real use.
How 8pack can guarantee these 3.9 chips are stable is a bit of a worry. Mine is realbench stable at 4.0 but it cannot handle linpack, this had brought me down to 3.8.

 

[eyetrip]

What is linpak representative of in real world usage?

Its been a challenge finding the right stability tester.

I agree aida64 isn't the best when pushing cpu heavy. Have had that pass to fail encoding in handbrake.

But is linpak/ibt like the furmark of the cpu world?

 

[gavinh87]

What is linpak representative of in real world usage?

Its been a challenge finding the right stability tester.

I agree aida64 isn't the best when pushing cpu heavy. Have had that pass to fail encoding in handbrake.

But is linpak/ibt like the furmark of the cpu world?

It could very well be but if a stock CPU pass it then why shouldn't an overclocked one?
I'm doing this purely to eliminate any future issue, that saves me from troubleshooting any issues further down the line. As I've said before realbench could pass but i was getting crashes in overwatch and bf1.
Piece of mind that it can handle anything i can throw at it i suppose. If it means i lose 30 points in fire strike and 2 FPS in gaming benchmarks then so be it

 

[eyetrip]

I get what you mean about wanting to know it's all. Good to go. It's why I'm annoyed with myself for trying to push over 4ghz.

I just want to know it'll handle things all round without issue.

I'll use occt if it really does mean it'll be fine for everything else. It's just I want to be sure that's the case

 

[gavinh87]

I get what you mean about wanting to know it's all. Good to go. It's why I'm annoyed with myself for trying to push over 4ghz.

I just want to know it'll handle things all round without issue.

I'll use occt if it really does mean it'll be fine for everything else. It's just I want to be sure that's the case

I'm not suggesting it's the be all and end all but nothing hits my cpu harder than occt or IBT. Realbench doesn't cut it for proper gaming at least in my experience. It might be worth running a mix of everything but from now on occt will be my first test then everything else can follow.
Oh and make sure you tick avx

 

[eyetrip]

whats a recommended run time?

 

[gavinh87]

1 hour was enough to prove errors for my 3.9 run. In fact it failed in less than 10 mins!
I've ran 3.8 for 1 hour and 8 hours so I'm sure it will find it fairly quickly.

 

[eyetrip]

Well, i certainly made it past 10 min at 4ghz, but it did shut down some time after that - was gaming in living room so not sure.

Did see temps get to 70c before I left it.

That was around 1.45v.

Upped the voltage ever so slightly, and dropped to 3.975

Will see how that goes.

 

[Scramz]

I recommend OCCT for stability mate. Or intel burn test, both use linpack. Realbench has been a complete **** for me too nvidia related as you are aware of.
I found AIDA not so good for CPU but really good for memory.

Surely if realbench black screens, that is a sign of an unstable overclock? This happens to me unless you raise the volts then it's stable. All regardless of NV.

 

[gavinh87]

Surely if realbench black screens, that is a sign of an unstable overclock? This happens to me unless you raise the volts then it's stable. All regardless of NV.

For me it was more a case of luxmark has stopped working. Rolling back nvidia drivers sorted this but this was way back, theres a thread about it somewhere as it was even happening @ stock. My point here is realbench isn't all what its cracked up to be. 8pack and silicon lottery are selling chips that can run at x speed but I bet neither can handle linpack or prime.

 

[gavinh87]

Well, i certainly made it past 10 min at 4ghz, but it did shut down some time after that - was gaming in living room so not sure.

Did see temps get to 70c before I left it.

That was around 1.45v.

Upped the voltage ever so slightly, and dropped to 3.975

Will see how that goes.

Good luck, I take it that is with a low fan speed?

 

[Scramz]

For me it was more a case of luxmark has stopped working. Rolling back nvidia drivers sorted this but this was way back, theres a thread about it somewhere as it was even happening @ stock. My point here is realbench isn't all what its cracked up to be. 8pack and silicon lottery are selling chips that can run at x speed but I bet neither can handle linpack or prime.

With my CPU, Realbench, linpack and prime all fail and pass at the same voltages/speed. Saying that, for years now on every CPU I have owned, I can pass Intel Burn Test with ease but will fall over with 30 runs of Realbench H264 encoding only (Non stress test).

People only want to put a certain loads on their CPUs because they use them for different things, horses for courses and all that.

 

[gavinh87]

With my CPU, Realbench, linpack and prime all fail and pass at the same voltages/speed. Saying that, for years now on every CPU I have owned, I can pass Intel Burn Test with ease but will fall over with 30 runs of Realbench H264 encoding only (Non stress test).

People only want to put a certain loads on their CPUs because they use them for different things, horses for courses and all that.

I can only go from my own experience but IBT stresses the CPU much more than realbench, the temp readings show this. Given I can pass RB at 4.0 but bf1 and overwatch give me issues suggest RB isn't all that. When I tried to find the issue OCCT hard crashed my pc within 10 mins, IBT did the same. I did a 4 hour run of OCCT earlier at 3.8 no issues, but 3.9 even at 1.45 is a no go. I either have the worst 1700 on the internet or people aren't properly stressing their CPU's and are posting overclocks just because it managed a run of cinebench.
This is what I'm trying to get to the bottom of

 

[Scramz]

I can only go from my own experience but IBT stresses the CPU much more than realbench, the temp readings show this. Given I can pass RB at 4.0 but bf1 and overwatch give me issues suggest RB isn't all that. When I tried to find the issue OCCT hard crashed my pc within 10 mins, IBT did the same. I did a 4 hour run of OCCT earlier at 3.8 no issues, but 3.9 even at 1.45 is a no go. I either have the worst 1700 on the internet or people aren't properly stressing their CPU's and are posting overclocks just because it managed a run of cinebench.
This is what I'm trying to get to the bottom of

Cinebench doesn't confirm a stable overclock as we all know, I think the same for AIDA.

As for the others, yes their do pull out instabilities.

I settled 3.9Ghz at 1.31v, passes every test under the sun and has been a rock for heavy VM loads from day one, even now with 3200Mhz RAM. However, looking around I seem to have been a bit lucky with the CPU.

I am not sure of your loads but rather than spending a lot of time tinkering with stress tests which may put way more stress than you really need, set clocks and use your normal work load, like gaming for example, as your stress tools. And have fun doing so.

You may have issues over 3.8Ghz with stress tools, but maybe perfectly fine for your loads at higher speeds and never fail.

 

[Scramz]

I either have the worst 1700 on the internet or people aren't properly stressing their CPU's and are posting overclocks just because it managed a run of cinebench.
This is what I'm trying to get to the bottom of

What Intel burn test settings you using to fail at 3.9? I will give it a run.