FX-8320 vs. FX-8350 vs. FX-8320e Real-World Power Consumption

[SlowSpyder]

Yea, as long as you have CnQ enabled idle power will be still very low. My CPU drops as low as 1.4GHz and .88v at idle.

 

[DrMrLordX]

Hmmmm, explain a bit more if you can.

Are you saying the theory is that I could get 4.3GHz at the same or equal TDP as the 8350? Would I also get the benefit of lower idle consumption than the 8320/8350?

I'm not familiar with editing pstates. If it's too technical then I'm not sure I'm comfortable with it. I can follow a guide for most things and I have overclocked in the past so I know what changing multipliers and FSB along with voltages entails. I worry that will mess with the ECC process, too, though.

Well since you have the 8350 coming, you can do it with that instead of the 8320e.

But yeah, basically, you can hit 4.3 GHz (or 4.2 on the 8350) without chewing up that much more than the listed TDP for the chip. CnQ *should* work for you regardless of what is the base clockspeed of the processor. All you're doing is disabling turbo and picking the max turbo state as your base clockspeed.

Editing pstates is another way to accomplish this, especially for processors that don't behave properly when the base clockspeed is altered in the BIOS (such as my Kaveri; CnQ doesn't work in Windows when I set the base clock to anything but auto, but it will in Linux, go figure). The pstates represent all the different clockspeeds the chip can reach depending on usage level. So, it has all the low-power states for being idle and all the turbo states. Vishera has . . . 5 pstates I think. It's just a command-line tool where you enter the name of the pstate (p0, p1, etc) then plug in the CPU multiplier and then the voltage and you're done.

You can probably get your Vishera to go into low-power states during idle with an altered base clockspeed, so no worries.

The only reason I recommended the 8320e is that you can probably hit 4.3 GHz with an 8320e at around 1.3v vcore. The 8350 - assuming all the 8350s out there are still coming from before wk29 of 2014 - will probably take 1.36-1.38v for 4.2 GHz (your mileage may vary).

 

[soccerballtux]

just turn off the TDP limiting

on my 8310 at least, the turbo is horribly implemented-- it never turbos to 4.3ghz, and spends most of the time at 3.7ghz, and then downclocks occasionally to 2.9ghz

 

[soccerballtux]

Well since you have the 8350 coming, you can do it with that instead of the 8320e.

But yeah, basically, you can hit 4.3 GHz (or 4.2 on the 8350) without chewing up that much more than the listed TDP for the chip. CnQ *should* work for you regardless of what is the base clockspeed of the processor. All you're doing is disabling turbo and picking the max turbo state as your base clockspeed.

Editing pstates is another way to accomplish this, especially for processors that don't behave properly when the base clockspeed is altered in the BIOS (such as my Kaveri; CnQ doesn't work in Windows when I set the base clock to anything but auto, but it will in Linux, go figure). The pstates represent all the different clockspeeds the chip can reach depending on usage level. So, it has all the low-power states for being idle and all the turbo states. Vishera has . . . 5 pstates I think. It's just a command-line tool where you enter the name of the pstate (p0, p1, etc) then plug in the CPU multiplier and then the voltage and you're done.

You can probably get your Vishera to go into low-power states during idle with an altered base clockspeed, so no worries.

The only reason I recommended the 8320e is that you can probably hit 4.3 GHz with an 8320e at around 1.3v vcore. The 8350 - assuming all the 8350s out there are still coming from before wk29 of 2014 - will probably take 1.36-1.38v for 4.2 GHz (your mileage may vary).

1.33v for my 4.3ghz, and that's pretty high compared to what Anandtech saw

4.3-4.5ghz is where the voltage runaway starts on this chip.

 

[DrMrLordX]

The 8310 doesn't seem to do quite as well as the 8320e, but it is still a step up from a lot of 8350s out there. 1.33v for 4.3 GHz is better than what a lot of Visheras pre wk292014 can do.

 

[sandorski]

Not scientific at all, but my 8320 at approx 50% use on all cores and OC to 4ghz uses max 99watts, according to the Asus Monitor software. I'm not sure I trust that figure though.

 

[LTC8K6]

http://www.tigerdirect.com/applications/SearchTools/item-details.asp?EdpNo=9494388

Go with an 8310.

95W , 7700 passmark, $119.

 

[smitbret]

http://www.tigerdirect.com/applications/SearchTools/item-details.asp?EdpNo=9494388

Go with an 8310.

95W , 7700 passmark, $119.

That was my plan for the last 6 months, but as of Thursday and for the previous month or so the 8310 was "In Store Only". I had been following to see if it would come back in stock and finally gave up. I checked before I ordered the 8350, otherwise that's the way I woulda gone.

Son of a........

Oh well, with the 8350 I won't have to deal with the wondering if I should have gotten the faster CPU for just a few bucks more and by the time S&H from TigerDirect are included, the 8350 was only about $25 more.

 

[dave_the_nerd]

We need, like, a site banner telling people to ignore cpuboss links.

Maybe we should have the forum auto-filter them out like they do certain image hosts.

 

[smitbret]

We need, like, a site banner telling people to ignore cpuboss links.

Maybe we should have the forum auto-filter them out like they do certain image hosts.

Not trustworthy, huh?

 

[LTC8K6]

That was my plan for the last 6 months, but as of Thursday and for the previous month or so the 8310 was "In Store Only". I had been following to see if it would come back in stock and finally gave up. I checked before I ordered the 8350, otherwise that's the way I woulda gone.

Son of a........

Oh well, with the 8350 I won't have to deal with the wondering if I should have gotten the faster CPU for just a few bucks more and by the time S&H from TigerDirect are included, the 8350 was only about $25 more.

No problem, they went back out of stock again anyway.

 

[smitbret]

Wowza!

Everything about this CPU is a lot higher than the FX-6100.

Where the system used to idle at about 62-65W this one really wants to lock in around 81-89W. I ran the Passmark CPU tests on them both and the FX-6100 peaked at around 174W but the 8350 peaked at 226W.

That's the bad news, but what did I expect? It's 125W vs. 95W TDP.

The good news blows me away, though. When I ran the CPU Test (6100/8350):

CPU Mark - 4745 vs. 9166 93% improvement
Integer Math - 9473 vs. 23987 153%
Floating Point Math - 4528 vs. 8321 84%
Prime Numbers - 16.2 vs. 27 67%
Extended Instructions (SSE) - 17.3 vs. 34.4 99%
Compression - 5998 vs. 13741 129%
Encryption - 833 vs. 1764 111%
Physics - 271.1 vs. 567 109%
Sorting - 3592 vs. 8211 129%
Single Threaded - 1177 vs. 1570 33%

I expected less, much less. Keep in mind that these tests were run as the system ran normally. I didn't turn off any services so the Media Server(s), A/V, disk monitoring, etc. were all running at the time although none appeared to be actively engaged in any tasks.

Gonna encode a BD Rip and see how well it stacks up.

 

[LTC8K6]

http://www.cpu-world.com/Compare/302/AMD_FX-Series_FX-6100_vs_AMD_FX-Series_FX-8350.html

2 more threads, 21% more clock, handles more instructions...

 

[smitbret]

http://www.cpu-world.com/Compare/302/AMD_FX-Series_FX-6100_vs_AMD_FX-Series_FX-8350.html

2 more threads, 21% more clock, handles more instructions...

Which is kind of what I expected, about a 50-60% increase. All of these benchmarks are way above that except for the single core test which was still 33% higher. I am 2:21 into the Handbrake Re-Encode of a BD Rip and it is cruising along at 2x (if not more) than what I normally got with the FX-6100.

After just a few hours, I am very pleased.

 

[LTC8K6]

There are other improvements in there between generations, too.

 

[Abwx]

he FX-6100.

Where the system used to idle at about 62-65W this one really wants to lock in around 81-89W. I ran the Passmark CPU tests on them both and the FX-6100 peaked at around 174W but the 8350 peaked at 226W.

Idling power shouldnt be different from the FX6100, there s some power savings management disabled apparently.

 

[DrMrLordX]

Wowza!

Everything about this CPU is a lot higher than the FX-6100.

Where the system used to idle at about 62-65W this one really wants to lock in around 81-89W. I ran the Passmark CPU tests on them both and the FX-6100 peaked at around 174W but the 8350 peaked at 226W.

That's the bad news, but what did I expect? It's 125W vs. 95W TDP.

You can tame that with some undervolting. What's the default vcore on it right now, and have you messed with voltages or clockspeeds yet?

 

[smitbret]

You can tame that with some undervolting. What's the default vcore on it right now, and have you messed with voltages or clockspeeds yet?

No, I haven't done anything other than swap the CPUs, start it up, reboot when Windows recognizes the new CPU and then benched and encoded.

I will have to hook up a monitor and check the power-saving and voltages in the BIOS. Probably won't happen till Friday or Sunday, though.

 

[LTC8K6]

Is it clocking down at idle?

Is Windows set to Balanced power mode?

 

[soccerballtux]

You can tame that with some undervolting. What's the default vcore on it right now, and have you messed with voltages or clockspeeds yet?

I cna't believe people are complaining about 95w. At full load 4.3ghz in Prime I'm using 350w. At idle it's 130. At 4.6ghz it's 420w.

 

[DrMrLordX]

I cna't believe people are complaining about 95w. At full load 4.3ghz in Prime I'm using 350w. At idle it's 130. At 4.6ghz it's 420w.

Woooeee that's some serious power draw right there. You got a leaky chip, no doubt about it.

 

[soccerballtux]

Woooeee that's some serious power draw right there. You got a leaky chip, no doubt about it.

maybe. that's 130 at the wall, so less inside the case (PSU inefficiency).

Also, a 7850 driving 2 monitors-- https://www.techpowerup.com/reviews/AMD/HD_7850_HD_7870/24.html -- with the memory running at 1200mhz because of the monitor overclocking that locks out the VRAM frequency throttling/CnQ

and the fans seem to be about 20w no matter what.

it's really little worse than running a 100w bulb all the time, since most of the time it's idle (don't game anymore)

 

[SlowSpyder]

I cna't believe people are complaining about 95w. At full load 4.3ghz in Prime I'm using 350w. At idle it's 130. At 4.6ghz it's 420w.

Yikes, what is you voltage set to? I don't start seeing those kinds of numbers until I'm up into 5GHz+ range.

Dual graphics cards are a big draw, of course. But what is the change you see between idle and full CPU load (leave GPU's at idle)?

 

[LTC8K6]

Yikes, what is you voltage set to? I don't start seeing those kinds of numbers until I'm up into 5GHz+ range.

Dual graphics cards are a big draw, of course. But what is the change you see between idle and full CPU load (leave GPU's at idle)?

Well, the rest of us were talking about the CPU power, and he chimed in with system power for some reason.

 

[DrMrLordX]

It's still a high delta W, given that the task is CPU-only (no GPU stress from Prime95). That's delta W of 220W at the wall @ 4.3 GHz. I don't know his PSU efficiency, so it's hard for me to tell more than that.