FX Bulldozer VS Piledriver, 8120 VS 8320, Gaming comparrison

[RGone]

Thanks for inputting...

A little sidetrack here but I've noticed the same thing on my 6100. At 4.0Ghz the chip comes alive and then at 4.2 it becomes a beast but anything after that the excitement is gone.

From my Cinebench and 3DMark 11 tests I've seen the numbers start to go down around 4.4ghz and the diminishing returns begin. The plateau is there between 4.2 and 4.4 for sure.

I don't have 3Dmark11 but the one before that so glad you let us know what you are seeing and it on a 6 core processor.

This information that is seldom available as concisely as it seems to be coming in here is the type of thing that I have used for years to determine my purchases.

In this situation where the hastily (gawd they hyped forever) finished BD takes so much power and puts off so much heat, it is a very good thing to know it is not at all necessary to push the dang thing so hard and makes it what should be a great deal easier to clock these cheaper boards that give up when that huge voltage wall comes after about 4.3Ghz.

Save the pain and frustration. Stop where the thing is able to really flex its' muscles instead of on the downhill side fraught with power hunger and heat.

Again thanks "gnusounduave", input from those we have confidence in is a good thing.

RGone...ster.

 

[juane414]

RGone alerted me to this thread the other day and I figured that I would test out this plateau phenomenon on my Pildedriver FX-8320. I need to run some more benchmarks yet, but initial testing up to 4.5GHz has not revealed the sort of plateau in performance that was observed with Bulldozer. I have had to increase vcore to achieve stability with Prime 95 beyond 4.3GHz, but my performance has continued to increase as well in Cinibench and SuperPi.

I'll report back with the results tomorrow.

 

[RGone]

I don't think that "ssjwizard" had found a plateau for sure with his FX-8320 either. Or maybe not nearly as prominent as there was with the BullDozer cpus. I really have not looked over his 8320 stuff but I felt he would have mentioned it likely if he had seen one but then I think the holidays ended and he was chasing the CPU-PLL anomoly with this Gigabyte mobo also

I got bored about 11:30 last night and began to run Cinebench R 11.5, 3D06 CPU score only and an earlier 3D Mark Vantage starting at 5.2Ghz and working downward or backwarrds until I will get to 4.1Ghz. A lot of tests and then a bunch of datat to sort. and captures since no sense in running such without documentation. So I have at least another 2.5 hours of tests to run and then I don't even want to guess what to sort the data. I will get it done though.

Good luck with your testing "juane414". RGone...

 

[ssjwizard]

I hit a wall of what Im comfortable with pushing the CPU to before I found a plateau on the Vishera CPU. As with all chips the voltage required to go further deifintely starts to climb steeply but the performance kept going up.

I have a handful of additional benches to add to this thread later, but as RGone pointed out the holidays were around, and I was also looking into the CPU-PLL effect.

Feelfree to add any your results to this thread.

 

[juane414]

@ 4.2GHz
vcore: 1.4
max socket temp: 49c
max core temp: 37c
cinibench: 7.06
superpi1m: 21.559s

@ 4.3GHz
vcore: 1.4
max socket temp: 49c
max core temp: 37c
cinibench: 7.06
superpi1m: 21.123s

@ 4.4GHz
vcore: 1.425
max socket temp: 51c
max core temp: 39c
cinibench: 7.18
superpi1m: 20.639s

@ 4.5GHz
vcore: 1.475
max socket temp: 55c
max core temp: 43c
cinibench: 7.23
superpi1m: 20.186s

@ 4.6GHz
vcore: 1.5
max socket temp: 59c
max core temp: 47c
cinibench: 7.29
superpi1m: 19.750s

@ 4.7GHz
vcore: 1.5
max socket temp: (did not stress test)
max core temp: (did not stress test)
cinibench: 8.04
superpi1m: 19.344s

@ 4.8GHz
vcore: 1.525
max socket temp: (did not stress test)
max core temp: (did not stress test)
cinibench: 8.20
superpi1m: 19.407

I stopped stress testing for stability after 4.6GHz because I was more interested in the benchmark scores anyway and I was beginning to get uncomfortable with the amount of vcore and high socket temps. If there is a plateau with Vishera/Piledriver, I'm guessing that it would come somewhere around 5.0GHz on the FX-8320. Starting at 4.6GHz I was having to increase vcore by one step (.025) for each 100MHz increase and my temps started going up considerably. However, my cinibench and superpi scores were improving each time, with the exception of my final superpi score at 4.8GHz. For some reason it was slightly slower than 4.7GHz. That could be a sign that I was nearing some sort of plateau, but the cinibench score continued to improve. Obviously my tests showed nothing conclusive. If there is a plateau it is far beyond what I'm comfortable with for 24/7 usage.

 

[RGone]

Good show "juane414". There may not be a "plateau" as we call it but there certainly is information to be gleaned from your testing up to 4.8Ghz. Here is how I read your results and they are what peeps need to know really. Especially one with zero knowledge of high current consuming FX- type processors.

1. In a well cooled case a CM 212 plus or EVO would do well on a 'newer' Piledriver 8 core cpu up to about 4.5Ghz.

2. With your level of air-cooling and only 1.475 Vcore, the 4.5Ghz is easily held in check temp wise and should stay pretty darn cool in all respects as you have your case and cpu cooler setup.

3. By 4.7ghz you were reluctant to stress test. That tells me you were getting to a point that your good case air-flow and your air-cooled Cpu were getting out of hand for 24/7 use and generally we should find that happy spot for 24/7 use. Over the moon shots and let me push it to the max are not 24/7 happy spots for most.

4. The Vcore increases also point to some similar steps/walls or whatever we might call it when coming up to certain levels and seeking to pass the one we are at now and moving to higher CPU Mhz.

5. In my mind the jump from 4.5Ghz to 4.6Ghz did not return a significant increase in Cinebench cpu scoring. Increase enough to warrant moving on into terrritory you were reluctant to stress test in, most likely due to heat generated by having to up the Vcore.

Now you have to realize that what I write is 'my' assessment of the results you are seeing and graciously post for our viewing. Percentage wise it looks like the largest increase in reported Cinebench cpu power is in the area of 4.4Ghz when you move there from 4.3Ghz. 7.06 CineB score to 7.18 CineB score is percentage wise a pretty good increase and is coming at 100Mhz higher than my FX-8120 cpu was showing.

So we could perhaps understand that what I saw myself as a sweet spot on the Bulldozer 8 core cpu at 4.3Ghz is now 100Mhz higher and that is good to know from my perspective. I see in my minds-eye some good information in your testing results.

Good job man and thanks. RGone...

 

[juane414]

Thanks for your assessment RGone. I agree that the FX-8320 sweet spot is somewhere in the 4.4-4.5GHz range. The vcore and temperature increase at those speeds are pretty minimal even with my budget-friendly cooling. I do have another 140mm fan on the way which should improve the airflow in my case which is already pretty decent. I'm also thinking about replacing the stock Cooler Master TIM with some AS5. Currently I've been running 24/7 on stock vcore but if a new fan and new TIM can shave off a few degrees then I will have no reservations about running 24/7 at the 4.4-4.5GHz sweet spot. I could run at that speed with my setup the way it is, but I really have no need for that much power at this point.

 

[Anonaru]

I don't think that "ssjwizard" had found a plateau for sure with his FX-8320 either. Or maybe not nearly as prominent as there was with the BullDozer cpus. I really have not looked over his 8320 stuff but I felt he would have mentioned it likely if he had seen one but then I think the holidays ended and he was chasing the CPU-PLL anomoly with this Gigabyte mobo also

I got bored about 11:30 last night and began to run Cinebench R 11.5, 3D06 CPU score only and an earlier 3D Mark Vantage starting at 5.2Ghz and working downward or backwarrds until I will get to 4.1Ghz. A lot of tests and then a bunch of datat to sort. and captures since no sense in running such without documentation. So I have at least another 2.5 hours of tests to run and then I don't even want to guess what to sort the data. I will get it done though.

Good luck with your testing "juane414". RGone...

I can now confirm that single-core performance on PD scales best until 4.74GHz, where it takes a small hit (Its very small, but still measurable). However, I pushed a single core (One integer core on one module) on my shop 6300 all the way up to 5.45GHz, and saw no appreciable drop in how well scores scale per MHz added, it was almost a perfectly linear increase-- I'm not sure if this is the case with a fully saturated module-- I haven't been able to push one past 5.1GHz without LN2, and huge scores, frequencies, vCores, and the sort aren't going to reflect quite properly on the real world.. I may attempt to do a 3.2-5.5Ghz run full module (Or perhaps full 6 cores) on my benching table, but I'm going to have to use a different mobo for that

Conclusion: Best to shoot for the sky with PD, best to hang out at 4.3 ish on BD. Best not to go past 4.5 or 4.6 unless you have some creative cooling solutions on either

 

[juane414]

Conclusion: Best to shoot for the sky with PD, best to hang out at 4.3 ish on BD. Best not to go past 4.5 or 4.6 unless you have some creative cooling solutions on either

+1 to that conclusion

Well said!

 

[Anonaru]

+1 to that conclusion

Well said!

I did get my UD3's mosfets up to 140*C when pushing the 6300 beyond 5.0GHz without active cooling ^_^

Board is a bit tougher than I thought, but it now smells funny (And mosfets are now being actively cooled) Had a lot of fun during my little hiatus from these forums-- Got an old 920 rig running, got my oil submersion computer back on the shelf for people to see (With sexy dark-green LEDs, mind you), and been torturing my poor FX-6300 again. I'm taking this chip home with me if it survives a few more days, what a soldier (Even took LN2 and a push to 7.13GHz, woo)

 

[Mandrake4565]

got my oil submersion computer back on the shelf for people to see (With sexy dark-green LEDs,

where's the pics?

 

[Anonaru]

where's the pics?

Ask and ye shall receive + dimmed the lights just for you

This is in my office, before it was put back out on display. You likey?

Spoiler

 

[RGone]

Anonaru, that is class man. Just real class. RGone...

 

[Mandrake4565]

Nice pics Neo I like. I wanted to do a submerged rig for my son, but I'm married, she thinks it's going to catch on fire. Do I need to say more.

 

[RGone]

Three words fire, married, son. Nuff said man. RGone...

 

[Anonaru]

Nice pics Neo I like. I wanted to do a submerged rig for my son, but I'm married, she thinks it's going to catch on fire. Do I need to say more.

I must be lucky then. I'm married, and my wife said "That looks so cool!" when she saw it

 

[Mandrake4565]

I must be lucky then. I'm married, and my wife said "That looks so cool!" when she saw it

Indeed.

 

[Anonaru]

Indeed.

I'm in the process of updating the components in this immersion rig-- I have a huge rad a guy gave me with a leak (I patched!) and a honkin' pump. I'm thinking about running the oil though this to give it increased cooling power

I could totally showcase a gaming rig in oil I'll post a video if I get there

Got my shop 6300 to 5.4GHz on full 6 cores. Passed 20 minutes of Prime, passed 1 hour of prime, couldn't complete cinebench, no matter what I did. Daheq?

I'm going to try getting it fully stable with a single module, then work back up to 6 cores. Nyerking thing

 

[Mandrake4565]

I have a huge rad a guy gave me with a leak (I patched!) and a honkin' pump.

must be a hella pump if your going to be pumping oil.

 

[Anonaru]

must be a hella pump if your going to be pumping oil.

Its in the testing phase-- I don't think its a pump designed for water cooling, looks like a very small pond pump Maybe for in-ground decorative ponds? The feed line lead is 7/8ths of an inch I believe, so this is a very high-volume pump compared to what I usually work with

So far, its pumped a line out and back into the tank since 11 this morning (It is now 9:30 this evening / 21:37, and its still alive