Why does my Ryzen R7 1700 not maintain max boost on single core?

[CuriousMike]

Never throw away stuff from a build. Always tuck everything away into your computer case box and put it away in the attic or garage.

Tell me this 6 months ago -- you will have saved me $9 amazon dollars... 🙂

 

[CuriousMike]

Ryzen 1700 comes w/ Wraith Spire RGB.
This is why the Ryzen 1700 sold so many units. Excellent value. I have it at 3.7Ghz @ 1.20V w/ memory OC to 2933.

That is exactly what I have ( don't know what voltage I'm running chip at... whatever BIOS default is.) If I run it at 3.7ghz and i'm running BOINC with a blend of PrimeGrid and Rosetta@Home (75% cores, 100% CPU time) in a similar 78-80' room, I'm hitting 80-81'C... that's probably a safe temperature, but it makes me nervous.

I've knocked it back down to stock (so it's running all cores at 3.2) and I see 64-75'.

Amazon is set to deliver the AM4 brackets (that I threw away) on Sunday, so I'll be able to install the BeQuiet 120W cooler and see if that doesn't keep me < 80'.

 

[ub4ty]

That is exactly what I have ( don't know what voltage I'm running chip at... whatever BIOS default is.) If I run it at 3.7ghz and i'm running BOINC with a blend of PrimeGrid and Rosetta@Home (75% cores, 100% CPU time) in a similar 78-80' room, I'm hitting 80-81'C... that's probably a safe temperature, but it makes me nervous.

I've knocked it back down to stock (so it's running all cores at 3.2) and I see 64-75'.

Amazon is set to deliver the AM4 brackets (that I threw away) on Sunday, so I'll be able to install the BeQuiet 120W cooler and see if that doesn't keep me < 80'.

WEW LAD .. 80-81C ... that's too hot.
64-75C stock is absurd.
You need to take that stock cooler off and make sure you get a good application of artic silver or something on there?
Someone help this guy out. Seems like he has a bad mate w/ his cooler. No way you're supposed to be seeing 64-75C at stock settings with a wraith spire cooler. I never see above high 60s on an overclock of 3.7Ghz @ 1.20V. Stock, I don't crack high 50s at 100% CPU load. Are you doing the 20*C adjustment? Update your bios and make sure you have everything setup. Your temps are way too high. Something has to be wrong w/ your settings or cooler mating.

Also, show us what kind of case you have and your airflow [fan inlet/outlet].
You better not have one of those tiny meme cases w/ poor airflow.

 

[CuriousMike]

LOL, my case is good - Corsair 300R with plenty plenty of fans.

If I'm just gaming, I rarely hit above 60'.

BOINC is just like a non-stop stress test.

 

[CuriousMike]

Got the brackets, set it all up, went to put the HSF on and... it seems the orientation is wrong?

That got me googling if this is an issue, and evidently some HSF's can come oriented in vertical vs. horizontal.

It seems unoptimal to me, since my graphics card would be blocking where the HSF is trying to suck air from.

I went ahead and assembled it and I'm stressing it now, and so far it seems to have knocked ~9' C off ( running a high of 71'C at 3.7ghz ), so it is an improvement. I have two fans on the top of my case blowing air out, so maybe the config works. It just seems like it'd be more better if it blew air out the back of the case.

Me, discovering the block doesn't fit in the circle.
hsfv2 by CuriousMike, on Flickr

 

[Topweasel]

1700's top cooler was made for 95W my dude which is why you can OC it to 1800x's 95W TDP performance and it handles it.
I'm not sure who told you otherwise. It's not a 65W cooler.

The one sold with the Ryzen 5's is a 65W cooler which is the wraith stealth

Ryzen 1700 comes w/ Wraith Spire RGB.
This is why the Ryzen 1700 sold so many units. Excellent value. I have it at 3.7Ghz @ 1.20V w/ memory OC to 2933. Memtest 4 pass no errors. Prime95/Aidi64 full system stress test. Absolutely no issues. Never gets above 65-67*C in a toasty 78*F room. No big advantage to Ryzen 2k processors. I'll upgrade at 7nm.

This can go back and forth. AMD doesn't offer the HSF with a 90w CPU. It may be rated at 90w, but only 90w which means it won't work well for a 90w CPU on a hot day. It's the overprevisioning you want when offering a 60w CPU with turboing functions. Doesn't mean it's good enough in all circumstances to allow XFR to max out. It did a standup job for me before I got the bracket for my cooler. But anyone complaining about turboing and oc'ing with a stock cooler even one with the extra room of the spire, doesn't have a basis in reality. Nor does anyone using their specific cpu, with their specific setup, in their specific location proof that something should work in a completely different scenario.

 

[ub4ty]

This can go back and forth. AMD doesn't offer the HSF with a 90w CPU. It may be rated at 90w, but only 90w which means it won't work well for a 90w CPU on a hot day. It's the overprevisioning you want when offering a 60w CPU with turboing functions. Doesn't mean it's good enough in all circumstances to allow XFR to max out. It did a standup job for me before I got the bracket for my cooler. But anyone complaining about turboing and oc'ing with a stock cooler even one with the extra room of the spire, doesn't have a basis in reality. Nor does anyone using their specific cpu, with their specific setup, in their specific location proof that something should work in a completely different scenario.

Got the brackets, set it all up, went to put the HSF on and... it seems the orientation is wrong?

That got me googling if this is an issue, and evidently some HSF's can come oriented in vertical vs. horizontal.

It seems unoptimal to me, since my graphics card would be blocking where the HSF is trying to suck air from.

I went ahead and assembled it and I'm stressing it now, and so far it seems to have knocked ~9' C off ( running a high of 71'C at 3.7ghz ), so it is an improvement. I have two fans on the top of my case blowing air out, so maybe the config works. It just seems like it'd be more better if it blew air out the back of the case.

Me, discovering the block doesn't fit in the circle.
hsfv2 by CuriousMike, on Flickr

This is why I asked him for his info. He provided it. It is the same as mine w/ around the same room temp. He has the same CPU/Stock cooler at same room temp So, he should not be seeing literally 10*C-15*C hotter..... Unless
Wait, I just saw his mobo pic : B350.
GG OP... VRMs on B350s with an 8 core OC'd are what's killing you.

Also, I just took note of that Bronze Power supply.
Your power delivery quality to the CPU and subsequent increased heat is likely what's causing your issue. You'll also want to check what voltage you have set on your CPU. I doubt you can do a stock voltage setting and get a stable OC to 3.7Ghz? Stock is 1.18V I think?

You're doing high consistent loads + OC on a B350... Your VRMs are likely cooking...

Post up VRM temps if you can.

 

[CuriousMike]

You'll also want to check what voltage you have set on your CPU. I doubt you can do a stock voltage setting and get a stable OC to 3.7Ghz?

Running BOINC Primegrid (mentioned in post #27) which would be similar to running Prime95, I _was_ hitting 80-81.
With the new HSF I'm hitting a max of 73'C, so that has improved and I'm more comfortable. When running other workloads (such as Rosetta@Home), it's rarely hitting over 65' C.

I'm running all this at whatever the BIOS sets the CPU voltage to, and it's been stable and on 24/7 since installing the new HSF @ 3.7ghz.

 

[ub4ty]

Running BOINC Primegrid (mentioned in post #27) which would be similar to running Prime95, I _was_ hitting 80-81.
With the new HSF I'm hitting a max of 73'C, so that has improved and I'm more comfortable. When running other workloads (such as Rosetta@Home), it's rarely hitting over 65' C.

I'm running all this at whatever the BIOS sets the CPU voltage to, and it's been stable and on 24/7 since installing the new HSF @ 3.7ghz.

Welp, you're set then. Outside of this discussion #@!(* these summer temps.
I run my major loads at night.

 

[Topweasel]

Even then different silicon matters as well. This goes way back, but when I built my first computer. I saved up money had a friend that was building a computer at the exact same time as me order all the parts. We basically got clones of each others system. Same board, same retail Celeron 366A. Different cases but not by much. Me a less memory. He got a SB Live and I got an SB Awe64. That was it. We got them set up and on posted, left them on overnight in the Bios. Middle of the night my system hit a temp alarm limit. His was fine. We both were using retail CPU's with OEM heatsinks with OEM thermal pads.

Fast forward a couple of we both got better coolers and actual thermal paste. Both are chips are comfortable at 550Mhz. His caps out at ~600MHz, mine stops at about 590mhz or so. But several people I know get OEM CeleronA's. Only one of them can even hit 550 with a decent cooler.

Tolerances are a lot better nowadays but the point still stands. Even with identical setups there shouldn't be an expectation of matching performance.

 

[ub4ty]

Even then different silicon matters as well. This goes way back, but when I built my first computer. I saved up money had a friend that was building a computer at the exact same time as me order all the parts. We basically got clones of each others system. Same board, same retail Celeron 366A. Different cases but not by much. Me a less memory. He got a SB Live and I got an SB Awe64. That was it. We got them set up and on posted, left them on overnight in the Bios. Middle of the night my system hit a temp alarm limit. His was fine. We both were using retail CPU's with OEM heatsinks with OEM thermal pads.

Fast forward a couple of we both got better coolers and actual thermal paste. Both are chips are comfortable at 550Mhz. His caps out at ~600MHz, mine stops at about 590mhz or so. But several people I know get OEM CeleronA's. Only one of them can even hit 550 with a decent cooler.

Tolerances are a lot better nowadays but the point still stands. Even with identical setups there shouldn't be an expectation of matching performance.

I'll never forget this slotted form factor :

haha.

You're 100% correct. I am aware of the variances in 1700 given that it is a lower-binned part. I was just curious as to why there was a 10-15*C difference. That's alot ! This could be a simple issue with thermal paste or a whole host of other things. B350 has cut down VRMs and cooling. That's one factor and a difference. Also, there have been bios updates which I'm not sure he's on. There was a big change in what temp reading was reported. I know because I just recently did an identical build and the temps were off by that much. It made no sense as they were exactly the same parts. After some head scratching, I checked the bios revision and then did stepped tests w/ them. The temp reading difference was due to a change in how they did temp reporting in Ryzen across microcode updates. Problem solved. The temps were the same. So, when someone is seeing a 10-15*C difference, you definitely want to try to find out if its something far more simpler.

 

[CuriousMike]

You keep going on this drastic 10-15'C temp difference. I was essentially running Prime95 non-stop, which would be drastically hotter than a normal load. If you can say you were 15' cooler running Prime95 at 100%, then you have a point.

I've kept the BIOS up to date.

 

[Topweasel]

Also keep in mind that the 1700 never had the temp reading issue only the R7 X series CPU's, maybe the 1600x as well but I might be wrong with that.

 

[Malogeek]

On a related note to 1700 issues, I got tired of repeated failures trying to get any decent performance out of my 1700 over the last year. I couldn't even get it stable at 3.6Ghz properly trying all sorts of settings. Boost clock was useless and base 3.2Ghz was barely sustained ever, on a Noctua cooler. No end of frustrations for me.

Decided to sell it on ebay for $140 and picked up a 2600X from Microcenter for $190 and I'm sure glad I did. This thing out-of-the-box is rock stable at 4.2Ghz in gaming and around 3950Mhz in handbrake and prime95, no custom settings whatsoever in BIOS for the CPU. I basically got a 1Ghz gain for 90% of my desktop use (gaming) for $60. Lost 2 cores? Irrelevant for me really for my purposes. Even in handbrake the clockspeed different makes up for the 2 less cores.

I'm very happy with the Ryzen 2xxx refresh.