If I might add to this,
ht link is of little use to us, it effect OUR performance little.
cpu speed is king and cpu/nb is queen.
what I have found to work best is get the cpu speed first then cpu/nb second then ram timings,
this yeilds the best results for me.
...did a little more looking at cpu speed brings awesome video performance increases.
I was just thinking of some other methods of comparison of CPU speed to Video Performance. This comparison is only relative to me as I see so many first time DIY system configurators struggle to get the high cpu speed that many review sites seem to attain when they do their reviews of AMD FX-series components.
Most DIY system configurators, don't seem to realize that most review sites have at their disposal, components that are n0t entry level parts and pieces. Then DIYer goes to his friendly place of parts procurement and buys his stuff most often based on 'price'.
Nothing wrong with being cost/price concious. I do it. Although I will say I wasted my money on a 6850 video card because it 'fit' my price point. Oh it is more powerful than my 5770 card but not in real life enough to justify buying it. I digress.
So DIYer gets his $80.00 to $100.00 motherboard and at some point comes up into the OCF AMD CPU forum section wanting to overclock his stuff; knowing in his mind that he can do just what the review sites have done. NOT. I would speculate +95% n0t.
So I do not have cooling issues. They are handled. I can run any water-cooled speed within reason with my FX-8350 on my CHV Asus motherboard which is certainly not priced as entry-level by any stretch of one's imagination.
So easing back into this post's real emphasis. I have previously tested "my" own FX-8350 cpu and found that for any given cpu speed increase there 'seem' to be cpu speeds that give the most bang for the buck. What is that buck? Performance increase in video performance for one thing and that peformance balanced against Vcore needed for X increase in speed that also has the result of increased heat. Heat is without a doubt a factor that must be considered when overclocking an AMD FX-series processor be at an earlier Bulldozer series or the later Piledrive processor series. Because I have the temps in check, I can test a reasonable cpu speed and give myself an others a look at whether or not it seems prudent to try and run high cpu speeds on FX processors. I am mainly focusing on the Big Boys, the FX 6 and 8 core cpus.
I typed into g00gle, free + game + benchmark. One of the first 'hits' to appear was a link to Heaven DX-11 Benchmark in Version 3.0. I am not a gamer so no idea what such a game might involve or look like or any of that, but I read what the link seemed to say.
I could download Heaven DX-11 Benchmark and running it would 'not' assure that I could run the game but would show me something about how the Unigine Engine would react to my system components. At the completion of the 26 'steps' of the bench it would give some sort of result. From "my" earlier testing in this long thread, I felt that 4.4Ghz, 4.6Ghz, 4.8Ghz and finally 5.0Ghz were the peak speeds that gave me a performance boost compared to a lesser cpu speed.
So sight unseen and no experience at all with the bench, I downloaded the Unigine Benchmark and began to run tests with it at the 4.4, 4.6, 4.8 and 5.0Ghz speeds. I benched with bench in it's base/default settings. NO adjustment was made to the bench at all. I mean we have a newbie to the bench, why dork with stuff I am not familiar with? Didn't.
I had two ATI video cards for use. A stock 5770 and a stock 6850 that like a numb-nutts, I bought being into; "fits my price point". I used both in the testing. Of course the more powerful video card gave more FPS. I think that would be a given. But did I gain way more performance with either video card, if I increased the Cpu speed? I have the answer in my own mind. The bench results are for others to make their own decision.
I am grateful to 'ssjwizard' for guiding me toward use of the 'spoiler' to reduce clutter when scrolling a thread between post. Thanks man. I will have the 4.4Ghz capture showing for both the 5770 video card and the 6850 video card but the other speed tests for each card will be 'within' a spoiler. Click the spoilers to see 6 more results total.
One last thought and it relates to memory and CPU_NB and HT Frequency speeds. I made an effort to keep all three within the range which should have little to no bearing on the bench's result. Note that I said I made the effort. Plenty of CPU_NB for whatever it brings to the table and ram speed in the DDR1850 range which I find is good for the FX-series processor. Glad I don't have a ton of DDR1600 ram since I find it rather less than as good as DDR1866-ish speed ram.
On with bench results.
RGone...___________
Testing 5770 Video Card. Four cpu speeds.
CLICK Spoiler for three more test results at greater cpu speeds.
Testing 6850 Video Card. Four cpu speeds.
CLICK Spoiler for three more test results at greater cpu speeds.
and i've noticed this same phenomenon on 3dmark 11.
I've benched at 4.0, 4.3 with AOD and at 4.5ghz with AOD noting about the same as you had projected with the only difference being in the combined tests that 3dmark11 offers. (combined CPU GPU tests) of course less overclock on the Cpu would yeild a lowered Combined or single physics result.
in the Racing simulation that I use CPU overclocking is going to be very valuable as it's a CPU intense product and less reliant on a powerful GPU like Battlefield and other similarly marketed games.
May I post a few of my benchmarks? Some are not stable overclocks on the CPU and some have utilized GPU tweak, but I feel they will illustrate similar characteristics to the info you've already provided.
I have benchmarks for Asus 7970 Direct CUii
Asus 680Gtx 4gb Direct CUii
and lastly Asus 7970 Direct CUii and Gigabyte 7970 in crossfire.
The most comparable to your results would come from the single Asus 7970 and benchmarks i've run with overclocks going up the ladder as I learned things from this forum!
Okay. If i'm reading the charts right, there is no significant increase in Frames when using a CPU overclock. So a monster clock isn't going to help your gaming video performance.
But, on the other hand I have these results from Benchmarks I did while learning more about the overclocking process.
*Disclaimer. I'm learning and these overclocks were produced with AMD Overdrive. What i've learned is that AMD Overdrive is not a suitable OC'ing method and to attain true stability more traditional methods and patience will provide better results.
So, on to my results from 3dmark11. Some graphics tests actually gained 10+ frames yet with 0.2ghz more CPU power the physics test only gained by a couple of frames (which when viewed on a large scale against higher clocked CPUs a few frames is actually a large margin on the benchmark.)
Now what i'm wondering is how a proper stable OC affects the benchmark. I have my CPU at 4.4ghz which is right in the middle. I can fire up the benchmark and go for it one more time to get a third result.
I can tell one what happened to me when I was not stable. I was just pushing up the cpu speed and stab and slap the Vcore and CPU_NB voltage up as I 'thought' I remembered them. That was okay until the clocks got up in the 4.8Ghz range and the bench began to hang. Had to go and get my real values for stable overclock to ensure that Unigine did not hang up at high cpu speed. YMMV
RGone...
Have seen that before. Low Vcore influencing the score. Think back a number of posts I showed just such a thing happening with the same cpu speed; the only change was to Vcore to get a higher benching score.
RGone...
Have seen that before. Low Vcore influencing the score. Think back a number of posts I showed just such a thing happening with the same cpu speed; the only change was to Vcore to get a higher benching score.
RGone...
Ok so these tests had me thinking, so I did a few 3d Mark 11's myself. I was curious if I would have the same result as Fugu with the large jump in FPS on the graphics tests. I'm wondering if, because he has such a high end GPU that at 4.3ghz his GPU is being bottle-necked by the CPU. As you'll will see I did not have the same results. I hope my first foray into graphs will be easy to understand. What I did was set the gpu at the stated speeds in the title above it and then ran at 4.4,4.6 and 4.8 on the cpu.
Ok so these tests had me thinking, so I did a few 3d Mark 11's myself. I was curious if I would have the same result as Fugu with the large jump in FPS on the graphics tests. I'm wondering if, because he has such a high end GPU that at 4.3ghz his GPU is being bottle-necked by the CPU. As you'll will see I did not have the same results. I hope my first foray into graphs will be easy to understand. What I did was set the gpu at the stated speeds in the title above it and then ran at 4.4,4.6 and 4.8 on the cpu.
Well while you were gone testing, Fugu did another set of benches and he in fact got a better score at his lower stable speed. Similar to what your graphs show. So his first post of benches had to be off in some manner. His later bench seems to mirror your results and in fact just about what we have discussed. More cpu speed may not be worth the effort and certainly not if the configuration is not up to "poooshing-it".
RGone...
Well while you were gone testing, Fugu did another set of benches and he in fact got a better score at his lower stable speed. Similar to what your graphs show. So his first post of benches had to be off in some manner. His later bench seems to mirror your results and in fact just about what we have discussed. More cpu speed may not be worth the effort and certainly not if the configuration is not up to "poooshing-it".
RGone...
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
