FX 8120 to 8350?

[Joni]

This 8350 has VID 1.375 V, would be perfect for extreme clocks with watercooling, or was it just the opposite?

4.4 goes with 1.375 V, 4.6 takes 1.450 but sadly that's the limit for my NH-D14. Did blend only couple of hours though.

I usually dont push over 60c core just because i have no need to push it that hard. How far you can go on air really depends on the chip and how thirsty it is. I can do around 4.9 give or take when i had my noctua on it, at 1.52vcore and i was running 62-64c after an hour with 18.3c room temp and good case airflow. With my 8320 i couldnt do over 4.5 without dumping 1.55 into it.

As for is it an upgrade? Yes

Is it worth it? No, unless the 8120 isnt providing the performance you need. Piledriver will give you a bit better performance but isnt worth the price to upgrade. Wait for steamroller to come and see some benches on that to see how much better it is..

In the mean time, to keep the itch away, start playing with some benches and see what that chip can do! Whenever I feel like building a new rig or swapping to something different, I just load up Superpi or something and take it out on the poor 8350. If it goes up in flames its time to upgrade! I think it likes it though as it always begs for more.

gotta try that. Someone offered me 130 € for my 8120 so the price of upgrade isn't really that bad.

 

[RGone]

4.4 goes with 1.375 V, 4.6 takes 1.450 = That happens to a lot of FX processors and one of the reason most never see those hyped up easy 4.9 and 5.0Ghz numbers that are thrown around in forums.

130 € for my 8120 = That sounds like a good price to your favor for that cpu. Just remember that is a very good cpu that takes low Vcore, but now you have already fallen to the lure of the FX-8350 and opened and tried it.
RGone...

 

[Spliffstarr]

Welcome to the dark side.

 

[Joni]

RGone, On your thread "Testing does FX-83xx seem to flat line after 4.3GHz" were those voltages tested on prime or just for benching?

Also for example on 4.4 GHz you had LLC 50% (medium) and manual voltage set as 1.381250 V. Do you remember if the voltage on CPU-Z (1.380 V) dropped on load when running Cinebench or Prime?

On my M5A99X EVO, 4.4 GHz, LLC 50% (medium) and manual voltage 1.381250 V. CPU-Z reports 1.380 V on idle, under load it goes to 1.344 V.

I was wondering if my 6+2 power phase is not enough for 8350.

Also what VID is your 8350?

 

[Mandrake4565]

Joni, I had my 8350 on a M5A99X Evo, I do recall getting V Droop at medium settings while running prime. I can also tell you I couldn't get the 8350 stable above 4.4 on that board. I would get hard locks within minutes or trying to run prime95. Where as I have my 8350 stable at 5.1 2 hours on my CHV. I really like my M5A99X Evo, it is a great board and I have my 955be on it now running 4.1 24/7 on air, I just don't feel it has the guts to run the FX 8xxx series chips at high overclocks.

 

[RGone]

RGone, On your thread "Testing does FX-83xx seem to flat line after 4.3GHz" were those voltages tested on prime or just for benching?

My Answer >> I am not at that computer right now but going thru that post and one for the Fatal1ty 990FX, I would say they are just the voltages needed to run the tests. I do remember posting about what I thought was an odd happening in the the FX8350 flatline thread that I saw scores in one of my Cinebench tests rise when I raised the Vcore. So I am pretty sure I would not call those voltages out as P95 stable but not the very very least to get by with Bench testing.

Also for example on 4.4 GHz you had LLC 50% (medium) and manual voltage set as 1.381250 V. Do you remember if the voltage on CPU-Z (1.380 V) dropped on load when running Cinebench or Prime?

My answer >> here is mostly answered above, except that I don't remember me using any medium LLC. I don't think I used that since I generally use High or Ultra High.

On my M5A99X EVO, 4.4 GHz, LLC 50% (medium) and manual voltage 1.381250 V. CPU-Z reports 1.380 V on idle, under load it goes to 1.344 V.

My Answer >> I would think a medium CPU LLC would n0t be enough on the FX-8350. OR you would have to raise the initial Vcore a good bit.

I was wondering if my 6+2 power phase is not enough for 8350.

My answer >> I said earlier that if your FX-8120 really only used 1.272 Vcore for the speed you were shoiwing, then you were going to likely be put off when an FX-8350 was used since MOST of the FX processors would not begin to run 4.4Ghz at 1.272 Vcore. Now if you were not prime 95 stable before and you are now testing real stability then I can understand your line of questioning. We have more than one user that have moved to the more top of the line boards to fully utilize as much of the FX-8350 speed as they could. Above 4.5Ghz generally is what they felt they were missing out on. They moved to better boards and got the extra speed but with VERY good water cooling.

Also what VID is your 8350?

My answer >> 1.375VID if I am remembering correctly.
RGone...

 

[Woomack]

1 tip that may work if you have weaker power section or it's overheating. If you have options to set it then change LLC to low/regular, CPU/CPU-NB switching frequency to lowest possible, every voltage swiching ratio ( depends from board there is memory, NB etc ) from auto or x2 to x1. Enable power saving features but disable AMP.
At least on my FX8320 vrm temps dropped by about 10*C even though I had to set CPU voltage manually +0.05V. Also socket temps are a bit lower and overall stability is the same.
FX8320/50 cpus have stock voltage up to 1.425V ( depends from used cores ) but most will run at about 1.35-1.375 at auto. My FX8120 could run at about 1.375V @4.7GHz but FX8320 needs 1.42V to make 4.5GHz ( it's running in F@H right now at these settings ).

 

[Joni]

Here is mostly answered above, except that I don't remember me using any medium LLC. I don't think I used that since I generally use High or Ultra High.

I would think a medium CPU LLC would n0t be enough on the FX-8350. OR you would have to raise the initial Vcore a good bit.

Sorry, I meant 50% high not medium. On 4.4 GHz I have set 1.381250 V, LLC high. CPU-Z reports 1.380 V on idle, under load it goes to 1.344 V, what I was wondering if your top MB had the same amount of vdroop?

I said earlier that if your FX-8120 really only used 1.272 Vcore for the speed you were shoiwing, then you were going to likely be put off when an FX-8350 was used since MOST of the FX processors would not begin to run 4.4Ghz at 1.272 Vcore. Now if you were not prime 95 stable before and you are now testing real stability then I can understand your line of questioning. We have more than one user that have moved to the more top of the line boards to fully utilize as much of the FX-8350 speed as they could. Above 4.5Ghz generally is what they felt they were missing out on. They moved to better boards and got the extra speed but with VERY good water cooling.

I always run Blend on Prime overnight to see if it's steady, I'm kind of a perfectionist and even though OC that might fail on long Prime run would be stable on normal use 24/7, it's just not enough for me. That 8120 has VID 1.2XX thus the low voltage it needs to be stable.

1 tip that may work if you have weaker power section or it's overheating. If you have options to set it then change LLC to low/regular, CPU/CPU-NB switching frequency to lowest possible, every voltage swiching ratio ( depends from board there is memory, NB etc ) from auto or x2 to x1. Enable power saving features but disable AMP.
At least on my FX8320 vrm temps dropped by about 10*C even though I had to set CPU voltage manually +0.05V. Also socket temps are a bit lower and overall stability is the same.
FX8320/50 cpus have stock voltage up to 1.425V ( depends from used cores ) but most will run at about 1.35-1.375 at auto. My FX8120 could run at about 1.375V @4.7GHz but FX8320 needs 1.42V to make 4.5GHz ( it's running in F@H right now at these settings ).

Thanks, I'll try these

 

[Mandrake4565]

Sorry, I meant 50% high not medium. On 4.4 GHz I have set 1.381250 V, LLC high. CPU-Z reports 1.380 V on idle, under load it goes to 1.344 V, what I was wondering if your top MB had the same amount of vdroop?

My board set on high doesn't droop until I'm really pushing it around 4.9-5.0, I usually use Ultra High at that point. I never use extreme, it just overshoots my voltage settings by way too much for my liking.

 

[Joni]

My board set on high doesn't droop until I'm really pushing it around 4.9-5.0, I usually use Ultra High at that point. I never use extreme, it just overshoots my voltage settings by way too much for my liking.

Thanks, that's helpful information I somehow missed your previous post above about the M5A99X. I'm at 4.4GHz and just like you said, can't seem to get past it.

I think I have a wall on 4.6GHz on air anyway, so maybe CHV would be a waste of money without proper watercooling. I wonder if Sabertooth would get this to 4.6?

 

[RGone]

waste of money without proper watercooling. = Is the overall situation where you are now. Period.

 

[Joni]

waste of money without proper watercooling. = Is the overall situation where you are now. Period.

Yeah, I guess 4.6 is not worth a new mb.

Does LLC itself create more heat?

I mean running without LLC, idle for example 1.440 and load with vdroop 1.404 vs LLC High and steady 1.404 on load and idle.

 

[RGone]

Based on what you write it would seem using CPU_LLC would be slightly cooler. But the real truth is in your own testing. HWMonitor logs Current/Min/Max of temps so should be easy enough to tell for yourself at your end.
RGone...

 

[Joni]

What kind of 24/7 clocks are people running 83xx with watercooling, also how high are the fans needed to be to cool that water?

 

[Joni]

Yay, I got into 4.5GHz! LLC Ultra High (75%), 1.4280 V. 65 C socket and 55 C cores after 2 hours blend on Prime.

 

[manu2b]

What kind of 24/7 clocks are people running 83xx with watercooling, also how high are the fans needed to be to cool that water?

4.8/4.9GHz is realistic. You might reach 5 if you have a good chip.

Regarding fan speed, it depends on the radage, but with a 360, you set them on 100% when stress testing/folding and can keep them on the minimum under normal use.

 

[RGone]

4.8/4.9GHz is realistic. You might reach 5 if you have a good chip.

Regarding fan speed, it depends on the radage, but with a 360, you set them on 100% when stress testing/folding and can keep them on the minimum under normal use.

The reaching of an extended use, 5.0Ghz is certainly only with what most of us consider a golden piece of silicon.

The 4.8/4.9Ghz is going to need some serious water though. Not these $100 buck mini-water cooling loops. Not from what I have seen in the forums anyway.

Bassnut has been very busy lately and not posting nearly like he has done previously but I remember this shall we say extensive (and expensive) expedition he took into water cooling for his FX processors. He was serious and got serious results as I remember. I think you can look in his signature and see his FX-8350 is now at 4.9Ghz daily but his FX-8150 was only at 4.72Ghz.

LINK >> http://www.overclockers.com/forums/showthread.php?t=715371&highlight=cooling
Build FX Tamer - Water Cooled Upgrade

 

[manu2b]

^Sure.

Was talking about a custom loop with, at least, a 240 for the CPU only.

 

[Johan45]

The 4.8/4.9Ghz is going to need some serious water though. Not these $100 buck mini-water cooling loops. Not from what I have seen in the forums anyway.

Ya you said it RGone I have a AIO cooler and 4.6 is good for me. I can reach 4.8 but I'm pushiing my limits. These things are like a mini reactor

 

[Joni]

What do you consider stable? For example yesterday I got 12 hours of blend in Prime95 without errors, left it on for the night. But doing it again when watching a BD at the same time and using some other apps it might freeze or give illegal sumout error in two hours.