hmm maybe the graphics card or other PCI-E devices are crapping out. BCLK clock of 103 although not a lot, still allows for RAM speed of 3000mhz instead of 2933 for those with DDR4 3000 kits. Would love to see the tomahawk beating lower end X370 boards for tweakability.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
SOME RYZEN 7 1700 OVERCLOCKING!
- Thread starter Gibbo
- Start date
never been a fan of overclocking with software but i tried out ryzen master and didnt think it was too bad at all, i set up three profiles, one at stock, one at 3.8 ghz and one at 4ghz.
used the stock settings for desktop, the 3.8 for gaming and the 4ghz for benching and all done with the click of the mouse.
i also tried out the asus crosshair oc presets, the 3.8 crashed instantly but the 4ghz worked fine, i think this was down to only the 4ghz profile set my ram at the correct speed.
i will have a play about with the bios overclocking later as im too busy reinstalling everything at the mo
used the stock settings for desktop, the 3.8 for gaming and the 4ghz for benching and all done with the click of the mouse.
i also tried out the asus crosshair oc presets, the 3.8 crashed instantly but the 4ghz worked fine, i think this was down to only the 4ghz profile set my ram at the correct speed.
i will have a play about with the bios overclocking later as im too busy reinstalling everything at the mo
You probably don't need this many repeats in IBT nor this much RAM.
What you want is enough RAM to get to near the peak GFLOPS (at least 80% but 90% is better), then 3-5 repeats is usually enough.
I searched around but could not find any info on how many runs to call it stable. I then ran 20
I'm happy to say that this is 100% solid.
Why do you need them? You only have to look at the OC thread to establish that nobody is achieving reliably more than 4.0ghz~ (1700/X - 4.1ghz 1800X), versus Intel the chips all appear to be cut from very similar cloth and with very similar limitations.
Also, why are people bothering to run IBT, so far as I'm aware it isn't even aware of the Ryzen instruction sets and like P95 (only worse) isn't a good method, use Linpack or HyperPi if you want to use something similar, or much better something like x264 encode tests.
Some good info here:
http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics
Some good info here:
http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics
Because it saves me the hassle of playing silicone lottery and trying to overclock myself, much better if I can just plug and play a new system.
I'd rather buy pre-built that I know is guaranteed to work.
I guess it depends on the board, on the Asrock Taichi is the opposite, level 1 gives zero vdroop, level 5 has little to no effect.
It has already been shown that it has picked up on issues with stability so that is why.
IBT uses linpack.....
IBT was abandoned in 2012, it uses an old compile of Linpack. IBT also isn't aware of current instruction sets like AVX - it's unrealistic and obsolete.
You can download a 2017 version of Linpack if you want to force an unrealistic load on your machine. Just like Furmark for GPU loading, it's a poor tool IMO.
It's proven valuable to me and others on here for finding instabilities which p95 and realbench have missed. The people over at overclocked.net have also had success with it.
As for adding unrealistic loads, yes it does but why would you want a system that cannot do it. Just because it's good enough for gaming doesn't make it stable.
Last edited:
Nobody mentioned gaming.
Does the fact I can't drive my car at top speed for 600 miles with melting the engine make it 'unstable', does it also make my tyres 'unsafe'? Or would they be working perfectly based on typical usage, or specifically my usage? Everything is designed to work around tolerances.
I get what you are saying and agree. However what IBT has done is resolve the issue for instability in a number of games. This could be contributed to the fact that it very well could be loading the CPU in a specific way with certain instructions that the game also does and thus by testing with this and making sure it runs then a few have now got their rigs gaming stable where Prime and similar had not.
Further to that if gavin has the time etc it would actually be good for him to revert back to the old settings pre IBT and confirm it is crashing again. Then load it with the 2017 version of Linpack and do the same test. Now if that test then also crashes and the same is resolved with it then we can say that both IBT or 2017 Linpack work, however at moment we have no idea if 2017 Linpack woudl have picked up the issue that gavin had. Of course he may just want to game on it now but also anyone else who was having similar issues could do the same to firm that up.
Why does everyone use cars lol.
The fact is a stock CPU could pass IBT no issues, there is no reason an overclocked one shouldn't be able to do the same. I'm not talking about IBT failing it completely caused my system to power off, a small increase in voltage later and it's passing 20 runs no problem. This proved instability. This crashed within 40 minutes whereas prime was running for 8 hours. I had realbench running for 8 hours too and Aida without a problem.
I stand by what i said earlier, I urge anyone to run IBT for 10 runs on max then tell me it didn't crash. If it passes then it's stable.
I searched around but could not find any info on how many runs to call it stable. I then ran 20
I'm happy to say that this is 100% solid.
Your peak GFLOPS there are 80 ish, but the CPU should be capable of 250 ish (3.8*8*8) with FMA. Perhaps that version of IBT can only do 4 FLOPS/cycle on Ryzen? In that case you'd still only be getting 65% of peak (125 ish), which is NOT a good stress test. You need 80%+.
How do you get there? I don't know. You obviously can't use any more memory. It might be paging, which you can fix by using a little less memory. Also I would try running 8 threads instead of 16.
As if the newer versions of Linpack are radically different to the old ones.
Some of these linalg codes date back to the 80s and are still relevant.The instructions issue is an interesting one though as mentioned above. The last IBT (2.54) behaves fine on Intel CPUs as far as I know, maybe a Ryzen quirk.
As for 'unrealistic', others have addressed this above. You do the stress tests you want. If you think using a CPU to solve mathematical equations is 'unrealistic' that's up to you.
Your peak GFLOPS there are 80 ish, but the CPU should be capable of 250 ish (3.8*8*8) with FMA. Perhaps that version of IBT can only do 4 FLOPS/cycle on Ryzen? In that case you'd still only be getting 65% of peak (125 ish), which is NOT a good stress test. You need 80%+.
How do you get there? I don't know. You obviously can't use any more memory. It might be paging, which you can fix by using a little less memory. Also I would try running 8 threads instead of 16.
As if the newer versions of Linpack are radically different to the old ones.
The instructions issue is an interesting one though as mentioned above. The last IBT (2.54) behaves fine on Intel CPUs as far as I know, maybe a Ryzen quirk.
As for 'unrealistic', others have addressed this above. You do the stress tests you want. If you think using a CPU to solve mathematical equations is 'unrealistic' that's up to you.
Others on oc. net have shown similar scores so I'll try tweaking over the weekend.
Linpack has still shown to be a greater stress than p95 and realbench as previously said they were passing without issue. Upon linpack event viewer reported errors followed by a complete shut down. Now that is fixed i really don't know what else to throw at it.
Lots of mathematians here trying to find new prime numbers i see
If you want a proper burn test in that manner, go run Firestarter. Check the way it loads your CPU versus IBT.
I might drag out my abacus tomorrow, obsolete tools seem to be all the rage here
. It doesn't have the capability to push the CPU correctly, using an imperfect tool to crash your machine is exactly that.
There are many better tools.
If you want a proper burn test in that manner, go run Firestarter. Check the way it loads your CPU versus IBT.
I might drag out my abacus tomorrow, obsolete tools seem to be all the rage here
. It doesn't have the capability to push the CPU correctly, using an imperfect tool to crash your machine is exactly that.There are many better tools.
What's the harm in using all tools at your disposal?
As mentioned above I've had errors appear in IBT/LinX that were never picked up in other stress tests.
I've had games crash after doing a 9hr P95 test...
It's best to do a mix of stress tests until you are happy that your PC is stable for your needs.
As mentioned above I've had errors appear in IBT/LinX that were never picked up in other stress tests.
I've had games crash after doing a 9hr P95 test...
It's best to do a mix of stress tests until you are happy that your PC is stable for your needs.