Testing does FX-83xx seem to flat line after 4.3GHz

[keny]

Strange things them cpu's...... Its good to get stuck into something now and again isn't it

 

[RGone]

Yes it is I guess

...I guess but scratching my head is dangerous to the hair I still have. Hehehe. I got too fixated on the target I was looking at and forgot where I was and drove off the road into the ditch.

Strange things them cpu's...... Its good to get stuck into something now and again isn't it

Okay 'keny' here is what I believe you asked me about and with a little added spice for good measure. I put most of the information at the bottom of the chart itself. That way I could actually link the chart and not have to type the information again. Info would follow. I could have gone way on 'back' to the 4.1-ish speeds I was testing before but you seemed to ask about the area most of us could reach on good water out to about 5.3-ish. So that is what the chart includes.
RGone...



DO I need to go back and spoiler only a 'single' image? For only a single image I thought not to spoiler so it does not take one more ''click'' to see what is up. Open to suggestions if you guys think I should.
RGone...Again.

 

[keny]

Awesome !

What were the scores in relation to cine bench @ 5.3 @ 2600nb and 5.3 @2400nb, sorry for the zillions of questions but the tables are a little confusing for me, I'm old, and scores I understand lol

 

[I.M.O.G.]

Rgone it sure seems that the HT Link sp/NB Freq need to go up in scale when you get to those high CPU mhz.

I haven't fully parsed this thread, however htlink has no effect on performance in my experience for piledriver between 1600-2000mhz.

Cpu-nb is very important for memory tasks. But scores don't scale at all by pushing higher that link.

RGone, have you tried different htlink? Do your results coincide with my experience?

 

[keny]

Cpu-nb is very important for memory tasks. But scores don't scale at all by pushing higher that link.

This was my experience with bulldozer, scores seem diminish slightly in comparison, once the CPU NB went above 2400, (unless cold) but that was bulldozer and just wondering if pile driver was smiler. RGone has done a great job with these tests and there a very interesting read.

 

[RGone]

For "keny" and "I.M.O.G."

@ "keny":
Cinebench 11.5 @ 5.3Ghz CPU_NB & HT Frequency @ 2585Mhz = 9.12 pts CPU.
Cinebench 11.5 @ 5.3Ghz CPU_NB & HT Frequency @ 2370Mhz = 9.03 pts CPU.

@ I.M.O.G.

however htlink has no effect on performance in my experience for piledriver between 1600-2000mhz

I have never lowered the HT Frequency that low on my earlier 8120 nor the newer 8350. 2600Mhz was the default on my 8120 and the 8350 so I have kept HT Frequency up close to that, except for this foray back down towards 2400Mhz HT Frequency for "keny".

A buddy of mine was having trouble getting his FX-4170 to run high CPU_NB and we had to up CPU_NB voltage. During the course of dealing with that I remembered that with AMD it was always CPU_NB, equal to or greater than HT Frequency, 'until' BD and then PD apparently since AMD defaulted the HT Freq to 2600Mhz. Because of my previous experience and thoughts on CPU_NB and HT Freq, I did test CPU_NB ahead of and behind HT Freq.

That would be CPU_NB = 2400Mhz-ish and HT Freq = 2600Mhz-ish (default)
and then:
that would be CPU_NB = 2600Mhz-ish and HT Freq = 2400-ish and the second setting was about 5.5% faster or 0.5 points more CPU Score in Cinebench 11.5.

I expect that the CPU_NB is the real power producer not HT Freq. but since AMD spec'd the HT Frq. at 2600Mhz, I have just started to use that with BD/PD.

My normal everyday speed is 236x19.5 = ~4.6Ghz with Windows Power Mode set to Balanced Performance. My new Video Editting speed is now 236x20.5 = ~4.8Ghz with Windows Power Mode set to Balanced Performance after moving to the FX-8350 and looking at the CPU performance increase from 4.6Ghz to 4.8Ghz. Windows balanced mode does make a big difference in temps after going beyond 4.5Ghz so the balanced mode dropping the Cpu speed to about 1600mhz is nice.

Anyway before I get off track as I can do. I run everyday where I can keep CPU_NB and HT Freq. running at the same speed and between 2550 and 2610 and this keeps my DDR1600 kit and my DDR1866 kit totalling 16gigs in the range of DDR1850 to DDR1900 which is just about the limit with a DDR1600 kit in the slots with DDR1866 kit.

I normally only test for myself or for or with some close friends. I saw SO many users come into the AMD Cpu Forum section with less than CHV or 990FX Sabertooth motherboards and all those starry-eyed 5.0Ghz numbers the review sites put up when BD first came out and then these PDs, that I just began to see what CPU speed would give the most bang for the Heat and Vcore. It was 4.3Ghz with BD and seems to be about 4.4Ghz or for me 4.6Ghz and I am next going to try it with air-cooling instead of this monster water-cooling setup I normally run on the bench.

I don't expect to see any plus 5.0Ghz runs on air. But I have already settled in daily at 4.6Ghz anyway and I bought the #3 ranked Air-cooler over at FrostyTech and will make it push-pull not just push so I feel I can run my daily numbers on $80.00 air just to see if it can be done. Yes in my mind I think the 4.8Ghz will be doable on air and to do so I just bought one of the 'greater than average' clamping force hold-downs so I can vary the clamping pressure on the air-cooled HSF. You just never know.

Since this thread can go all over without me minding, I can just add such air-cooled testing and feel no issue.
RGone...

 

[Mandrake4565]

I bought the #3 ranked Air-cooler over at FrostyTech and will make it push-pull not just push

It be interesting to see how the 8350 does compared to your H20 setup. Which heatsink did you purchase if you don't mind me asking?

 

[RGone]

@ "Mandrake4565", what I hit or missed on is described below.


Aegir SD128264


Xigmatek S128264 Aegir Stacked Heatpipe Heatsink Review


Well the list a SS Phase as #1. Hehehe. The Noctua NH D-14 is a shaded quieter but not quite as good in cooling.

Would be splitting hairs to decide which would be better. I have had a lesser and cheaper 90MM Xigmatek and it performed on-par and often better than higher priced coolers a few years ago so I just decided to stick with a brand I knew and it having a good rating.

I knew better than to think a CM 212 EVO would handle the heat at what was 4.5Ghz and now 4.6 and 4.8Ghz so had to go with one of the big boys.

The Aegir SD128264 and the Noctua NH D14 are both tested with a push fan only. I bought a Xigmatek fan rated slightly 'higher' than the push fan to use for the pull fan. We will just have to see.
RGone...

 

[kyfire]

I have read this thread and wil likely reread it as I venture into the world of OC'ing my new FX 6300. The time that has been spend by RGone and other in this thread to provide the info contain shows thier intrest in now only improving thier machines but to help others improve thiers. Sadly, I think that RGone is right that most that come here are looking for the quick down and dirty way to boost thier CPUs and don't care or ever relize that there is a better way to get the most from thier computers. Myself, I have a large investment in my new system (will be getting it Tuesday the 22nd) and I want to be able to obtain the best preformace. I'm not looking to set records, nor am I looking to have extreme FPS in games. I want a stable system that will provide the preformace i need on a 24/7 basis.
It is also my hope that this old dog, after learning new tricks, will be able to help others learn.
A huge pat on the back to RGone and all the others that have put forth so much time and effort and have been willing to share thier knowledge with the rest of us.

 

[Mandrake4565]

I was looking for a new Air cooler for my 955, that is the same cooler I was thinking of buying. I have a Hyper 212 + on my I5-2500K so I wanted to try something different and maybe compare the two. Thanks.

 

[mtdewcowboy]

I need to know what cooling are you using Rgone?
I know its water cause there is noway on air
I am reading up on water cooling still have a ways to go but its nice see that you can get pretty high up on the clocks

 

[Mandrake4565]

mtdewcowboy, I don't know what RGone's setup is but I can tell you at 4.7 on mine with a 1.44375 v core and 1.25 Cpu/Nb my temps on 2 hours of prime were max 58 cpu socket and 49 on the package temp. My system is in my signature. I'm going to modify my right hand side case door and add a fan on the backside of the motherboard to see if it helps the socket temp.

 

[mtdewcowboy]

I am guessing a custom setup

 

[I.M.O.G.]

Thanks for the reply RGone about your experience with HT Link frequency.

I've always run Bulldozer at close to 2000MHz HTLink... Just because keeping it right there was easier, while paying more attention to CPUNB and CPU frequency. Easy to forget, and when cranking FSB, easy to pull things further out of spec than intended... Using a really low multi on HTLink hasn't ever hurt my benchmarks so long as CPU/CPUNB/Memory settings are comparable to my competition.

FWIW, this is the highest frequency Superpi 32M run ever on record for AMD:
http://hwbot.org/submission/2230669_i.m.o.g._superpi___32m_fx_8120_10min_53sec_125ms

My score is competitive with scores of similar frequency, though efficiency suffered a bit because my memory settings were not that strong compared to others - those were rare GSkill Flare RAMs, god's own ram pretty much, and I didn't want to push them hard when Bulldozer already sucks at SP32M (they ended up dieing shortly after that despite never having abused them). CPU-NB wasn't stable for SP32M any higher than that 3.73GHz frequency, regardless of what voltage I gave it (This was run at 1.65V CPU-NB). HTlink isn't shown in the screenshots, but it was set at an 8X multi so was running just over 2.1GHz...

Dunno if that means anything to you as it isn't a practical application, but my point is, I have tested it pretty extensively, and even at ABSURD CPU frequency and extreme CPU-NB frequency, low HT Link around 2GHz does not produce observable drawbacks to performance (at least on bulldozer, I assume the same is true of piledriver).

I'll stop in case this pulls things off topic, but I wanted to chime in on this point in case it helped fill out the picture you are looking at.

 

[ivanlabrie]

@ "Mandrake4565", what I hit or missed on is described below.


Aegir SD128264


Xigmatek S128264 Aegir Stacked Heatpipe Heatsink Review


Well the list a SS Phase as #1. Hehehe. The Noctua NH D-14 is a shaded quieter but not quite as good in cooling.

Would be splitting hairs to decide which would be better. I have had a lesser and cheaper 90MM Xigmatek and it performed on-par and often better than higher priced coolers a few years ago so I just decided to stick with a brand I knew and it having a good rating.

I knew better than to think a CM 212 EVO would handle the heat at what was 4.5Ghz and now 4.6 and 4.8Ghz so had to go with one of the big boys.

The Aegir SD128264 and the Noctua NH D14 are both tested with a push fan only. I bought a Xigmatek fan rated slightly 'higher' than the push fan to use for the pull fan. We will just have to see.
RGone...

I gotta say one thing: don't trust frosty tech's testing procedure.
An NH-D14 is better, Silver Arrow SB-E Extreme would be best...
Phanteks is pretty close too.

 

[RGone]

I.M.O.G., man there is no off-topic all in this thread really. I have told them we are beating the bushes. I don't like to get off when helping a user myself, but I started this wandering thing. No harm. No foul.

I imagine the HT Freq as a huge adder of Ooomph to the scores is not going to happen. I have found that the faster I try and push the Cpu clock after getting close to max, that lowering both CPU_NB and HT Frequency makes it easier to get that last few Cpu Mhz. Of course that is me and my crap. I try to make it up front that everyone's stuff can act entirely different. One's own testing if really pushing is the only definitive way to know where one's own stuff will really go and do the job.

I understand using the FSB can ****-up everything else when going up the ladder. I have just trained my ole slow self to watch all the busses. AMD gave out 2600Mhz HT Freq and I am just using it because I can I guess.

Thanks for the heads-up. If I ever decide to DICE, that no loss from lower HT Freq. will be good to know.
RGone...

 

[RGone]

I need to know what cooling are you using Rgone?
I know its water cause there is noway on air
I am reading up on water cooling still have a ways to go but its nice see that you can get pretty high up on the clocks

Water? Yes. 6 year-old 700gal per hour pump. 1/2inch insulated hoses. 6 year-old D-tek copper block 1/2 inch fittings. 55-60F water. Ghetto, red-nake, off the wall, down and dirty? Yep.
RGone...

 

[keny]

Thanks for the heads-up. If I ever decide to DICE, that no loss from lower HT Freq. will be good to know.
RGone...

You know you want to

 

[RGone]

You know you want to

Hehehe. You getting as bad as caddi daddi, bout wantin me to poosh shett.
RGone...______________

 

[azanimefan]

awesome thread. and thank you for that last post RGone; found that bit of "behind the scenes" talk to be very interesting.

there is some incredible work in this thread.