I’m currently running Linux on a cheap pre-built PC with a Ryzen 7 1700 CPU. There is no facility for overclocking in the BIOS, but I have been able to use Zenstates and P-state overclocking to get an all-core boost of 3.8GHz. (The cheap power supply won’t permit more.) Doing this, I retain the full range of frequencies from 1.55GHz idle to 3.8GHz max.
I am now going to build a new computer with an OC capable x570 motherboard and a Ryzen 5000 series CPU. My question is, should I attempt to overclock the CPU using the features of the motherboard, or should I leave it alone and stick with P-state overclocking? I’ve never done “real” overclocking before. This PC will be for productivity and software development. I’m not a gamer, and I don’t need the CPU running at 5.0GHz all the time wasting electricity. I will try tightening the memory timings, but as for the CPU, what do you think?
Generally with Ryzen if you go overclocking you give up PBO which can potentially enable you to boost higher than an all core overclock will get you. This can help with lightly threaded workloads.
I mean if you’re doing it for fun, go for it, just don’t expect to gain much without exotic cooling. AMD really has done a pretty good job of optimizing what is available out of the box IMHO.
If the computer does what it should at a reasonable speed, I see no point in OC.
Sure, OC is awesome, but raw CPU speed is only going to get you that far and these days, it’s usually more interesting to lower the wattage than overclock in either case. At 8 cores+, the bottleneck will not really be your CPU.
The native boosting would be already good left alone, especially when addressing low thread count loads. Forcing a bigger OC [additional cores / all-core engagement / constant higher base freq. / etc], would certainly need more aggressive cooling [Liquid-AIO / BIGG HS] and higher native/peak power draws [where MHz dividends, I’d question being worth while, over baseline operations]… Also assuring that the motherboard picked, can allow and put up, with that kinda lingering torturing
Memory seems to be Zen3s 12/10 stimulant - esp. hearing Wendell getting some real dumb timing exploits, out of his various bits of S C I E N C E - E R Y
I’ve had the new PC up and running for a couple of weeks now, and after much messing around, I found that enabling PBO for the processor and running the RAM at XMP settings gives me the best performance. Two cores boost to 5050MHz, so I have no complaints. I still want to look into tweaking the RAM timings, but there’s no rush.
If you’re concerned about power usage and leaving it on for long periods of time, look into undervolting. It keeps lower heat and noise, and ryzen still gives the same or better peak boost.
no real benefit to be honest.
tldr: just get some fast low latency ram.
but for them that dont know or have watched to many youtube videos from salesmen…
for online games you want a system that can display 128 frames if you have a 144 hz screen and not much more.
why?
packet data all online games have a hard limit of how many packets they can send and receive from the server typically 40-128* per second. they limit packet rates so they can sync your incoming data with the game on the server correlate all the data for the players and spit the packets back to you. giving you an updated game experience ideally at 1 packet per frame.
for this reason you can limit your fps in most online games to 128 and suffer no reduction in in-game accuracy caused by desyncing your frames on screen with the packet data received.
being that each frame of a game is rendered on a per packet basis IDEALY!.
sure you can buy faster kit and run it balls to the wall but i guarantee on the server at the other end. all your spitting out is a max of 128 packets like every one else. but suffering an accuracy reduction due to frame and packets desync and likely screen tearing due to running 300 fps on a 60/144/265 hz screen.
what am i saying with all this?..
you dont need to run your system balls to the wall for gaming.
there is such a thing as enough gaming power, and thats decided by the gaming server regardless of your EXTREME! build.
*there was 1 exception… fortnight ran an experiment allowing unlimited packet data. the end results were people running systems that could generate 265 fps were 8 times more accurate than some one running a 60fps rig… needless to say it was a duck shoot for the high end gamers.
but as i said it was an experiment. the amount of complaints they got of cheating went up over night. so they limited the packets again. evening out the battle field.
THE END!.
oh you read this far… HERO!
as for other performance gains its give and take with over clocking for productivity. the best thing to do is just build a stable system with fast low latency ram and let the cpu handle the clocks.
amd really did a decent job of squeezing the performance of there chips and historically never really left room for overclocking at the higher tiers.
so yeah its a good chip but no real reason to run it hot with an oc.
