- Joined
- Apr 20, 2004
- Location
- JAX, Mississauna
Up front Note. There were just too too many pics to look at when scrolling down. So I cut the pics by over 80% with the help of "ssjwizard". Thanks man.
I put the pics in a "Spoiler" with this so you can find the hidden pics.
Other Benches and BIOS Screenshots are in the Spoiler.
At the last of this page I indicated if I found other information, that I would come back and update the first post to say there is other information on further in the thread. There IS.
Okay, I am going to try and set the stage if I can remember all I have thought about during the benching time and when I was looking over all the results. I know two things that come to mind pretty fast. 1.) It takes a major out-lay of time and energy and resolve to run all the benches and keep track of the captures and try to do them orderly. 2.) Graphs or charts, may not be for us ole fahts. I needed to graph or is that graft somebody else's skills with spread sheets into my brain. But learned a thing or three.
What was I trying to do? I wanted to determine if the FX-8350 8 core processor seemed to sort of level off in adding cpu power after about 4.3 Ghz cpu speed with all the other busses and timings and stuff pumped-up and attended to as the FX-8120 I personally had did and others seemed to see the same in their own testing. Long statement of intent.
Okay some constaints from testing FX-8120 to testing of the FX-8350.
1. > Same motherboard and same bios. CHV non-Z with bios 1102.
2. > Identical cooling from FX-8120 to the FX-8350.
3. > Same installed operating system and drivers.
4. > Same 5770 ATI video card. Aging I know.
5. > There is one thing that shocked me during the entire testing. Don't ask me why but I wound up with two dissimilar kits of ram in the board. 1x 8 gig in two sticks Kit of Gskil Ares DDR1600. 1x 8 gig in two sticks Kit of Gskil Ares DDR1866. I had made a memory timing change to DRAM READ to WRITE Delay and it really stepped my ram up. Tightened the timings on all sticks to the numbers you will see later and also set 1T and the dissimilar sticks played happily together from about DDR1850 to DDR1900. Yes, the 16gigs has remained in the motherboard through-out the testing and remains in the slots still. Got to love that Gskil stuff and CHV and memory adjustments.
6. > Wish I had thought to do testing on just multiplier clocks since it is tough to land EXACTLY on a specific cpu speed and juggle the other busses for close speeds etc. Just using the multiplier would have been easier but I doubt I would have reached 5.4Ghz with only the upping of the multiplier.
7. > I made a strong effort to keep the cpu speeds falling close to the 4.1Ghz thru 5.4Ghz as I upped the cpu speed incrementally. Also made an effort to keep all the other busses within a reasonable range of each other incrementally as the cpu speed rose. I was able to keep the ram in the DDR1850 to DDR1900 range and as we know ram speed in that range has little to no effect on performance.
8. > When I spoke of keeping all the busses in range of each other in #7 above, I managed to keep CPU_NB and HT Frequency equal and the range was 2550 Mhz to 2610 Mhz through-out testing. I have done testing before and keeping everything to that tight range has never really affected performance good or bad. I did though keep the CPU-NB and HT Frequency up there to gain any performance boost that is within those two parameters.
9. > In my mind a lot of benching is just the ole cpu Mhz moon-shots we used to do. They tell little about the day to day opeation of the system. I tested/benched to 5.4Ghz because (Mr. Nameless), oh well...C_D would not leave me alone. I was going to stop at 5.2Ghz which I had P95'd and make the tests from 4.1Ghz to 5.2Ghz. Start at the same point I started with my FX-8120 and go as far as I had stability. But I kept hearing push it. Push it man push it. So crap I did. However I don't claim full stability for 5.3Ghz and 5.4Ghz. I know that benching is not 25% as tough as P95ing. So that is how that lays out. He got his 2 speeds over 5.2Ghz and glad I did really. When you see what the results are up that fast, then you may be able to leave the big numbers to DICE and LN2.
10. > I have used Cinebench R 11.5 in both the GPU and the CPU test and made performance graphs using the CPU performance only results for all speeds. I will put the graph of my FX-8120 and my newer FX-8350 at the beginning of the bench captures. Then I will post the actual capture of the test of Cinebench R 11.5. Some say that overall Cinebench can be a cross-platform performance test. I am not yet willing to go that far. But it does test the CPU and GPU.
11. > I have used 3D Mark 06 for CPU tests only. No video since it is old but I see some sites use it as a CPU performance predictor. Since I tested with it before, I did so again. 3D Mark 06 results should follow next.
12. > Then I ran 3D Mark Vantage and captured the results of the tests and will post them as well.
13. > Since I do not consider myself a gamer, I did not look for any games that had their own speed tests with them for test. Free or otherwise. I was far more interested in seeing if the CPU continued to produce more performance for each 100Mhz jump in speed from 4.1Ghz to 5.4Ghz. Or would the performance increase per 100Mhz cpu speed increase tend to flat-line after 4.3Ghz as the BullDozer/Zambizies tended to do. Early on the charts should tell that story and the screen captures are there from which I gathered the data.
14. > I will also post FULL F12 captures of my bios at each speed. This of course will be from the CHV mobo. There will be only 3 captures of the bios up until 5.0Ghz. At 5.0Ghz I swapped the CPU_LLC from HIGH to Ultra-High and captured that extra page of the bios to show CPU_LLC on Ultra High. So 5.0, 5.1, 5.2, 5.3 and 5.4Ghz were all run on the Ultra-High CPU_LLC setting. The extra voltage and load at the elevated speed, created the need for Ultra-High LLC. Thank goodness for superb cooling since the HEAT is still there even with PileDriver/Vishera Cpus.
By the way I will just add this note here about CPU_LLC since I have made mention of it. Asus says the setting of High is considered 50% CPU_LLC by them and that Ultra-High is considered to be 75% CPU_LLC by them.
15. > I put borders around places of emphasis on the screen captures. Text is added into the captures hoping for clarity. There will be one set of screen captures that includes HWMonitor at the speed I generally run 24/7. That speed should be in the neighborhood of 4.53Ghz, but I am steadily inching toward the 4.6Ghz mark to stay just on the rise side of the graphed performance leading up to 4.6Ghz.
16. > Okay that is just about all the background information I can think to put here. If I find anything else, then I will come back in and edit this first Post.
RGone...
I put the pics in a "Spoiler" with this so you can find the hidden pics.
Other Benches and BIOS Screenshots are in the Spoiler.
At the last of this page I indicated if I found other information, that I would come back and update the first post to say there is other information on further in the thread. There IS.
Okay, I am going to try and set the stage if I can remember all I have thought about during the benching time and when I was looking over all the results. I know two things that come to mind pretty fast. 1.) It takes a major out-lay of time and energy and resolve to run all the benches and keep track of the captures and try to do them orderly. 2.) Graphs or charts, may not be for us ole fahts. I needed to graph or is that graft somebody else's skills with spread sheets into my brain. But learned a thing or three.
What was I trying to do? I wanted to determine if the FX-8350 8 core processor seemed to sort of level off in adding cpu power after about 4.3 Ghz cpu speed with all the other busses and timings and stuff pumped-up and attended to as the FX-8120 I personally had did and others seemed to see the same in their own testing. Long statement of intent.
Okay some constaints from testing FX-8120 to testing of the FX-8350.
1. > Same motherboard and same bios. CHV non-Z with bios 1102.
2. > Identical cooling from FX-8120 to the FX-8350.
3. > Same installed operating system and drivers.
4. > Same 5770 ATI video card. Aging I know.
5. > There is one thing that shocked me during the entire testing. Don't ask me why but I wound up with two dissimilar kits of ram in the board. 1x 8 gig in two sticks Kit of Gskil Ares DDR1600. 1x 8 gig in two sticks Kit of Gskil Ares DDR1866. I had made a memory timing change to DRAM READ to WRITE Delay and it really stepped my ram up. Tightened the timings on all sticks to the numbers you will see later and also set 1T and the dissimilar sticks played happily together from about DDR1850 to DDR1900. Yes, the 16gigs has remained in the motherboard through-out the testing and remains in the slots still. Got to love that Gskil stuff and CHV and memory adjustments.
6. > Wish I had thought to do testing on just multiplier clocks since it is tough to land EXACTLY on a specific cpu speed and juggle the other busses for close speeds etc. Just using the multiplier would have been easier but I doubt I would have reached 5.4Ghz with only the upping of the multiplier.
7. > I made a strong effort to keep the cpu speeds falling close to the 4.1Ghz thru 5.4Ghz as I upped the cpu speed incrementally. Also made an effort to keep all the other busses within a reasonable range of each other incrementally as the cpu speed rose. I was able to keep the ram in the DDR1850 to DDR1900 range and as we know ram speed in that range has little to no effect on performance.
8. > When I spoke of keeping all the busses in range of each other in #7 above, I managed to keep CPU_NB and HT Frequency equal and the range was 2550 Mhz to 2610 Mhz through-out testing. I have done testing before and keeping everything to that tight range has never really affected performance good or bad. I did though keep the CPU-NB and HT Frequency up there to gain any performance boost that is within those two parameters.
9. > In my mind a lot of benching is just the ole cpu Mhz moon-shots we used to do. They tell little about the day to day opeation of the system. I tested/benched to 5.4Ghz because (Mr. Nameless), oh well...C_D would not leave me alone. I was going to stop at 5.2Ghz which I had P95'd and make the tests from 4.1Ghz to 5.2Ghz. Start at the same point I started with my FX-8120 and go as far as I had stability. But I kept hearing push it. Push it man push it. So crap I did. However I don't claim full stability for 5.3Ghz and 5.4Ghz. I know that benching is not 25% as tough as P95ing. So that is how that lays out. He got his 2 speeds over 5.2Ghz and glad I did really. When you see what the results are up that fast, then you may be able to leave the big numbers to DICE and LN2.
10. > I have used Cinebench R 11.5 in both the GPU and the CPU test and made performance graphs using the CPU performance only results for all speeds. I will put the graph of my FX-8120 and my newer FX-8350 at the beginning of the bench captures. Then I will post the actual capture of the test of Cinebench R 11.5. Some say that overall Cinebench can be a cross-platform performance test. I am not yet willing to go that far. But it does test the CPU and GPU.
11. > I have used 3D Mark 06 for CPU tests only. No video since it is old but I see some sites use it as a CPU performance predictor. Since I tested with it before, I did so again. 3D Mark 06 results should follow next.
12. > Then I ran 3D Mark Vantage and captured the results of the tests and will post them as well.
13. > Since I do not consider myself a gamer, I did not look for any games that had their own speed tests with them for test. Free or otherwise. I was far more interested in seeing if the CPU continued to produce more performance for each 100Mhz jump in speed from 4.1Ghz to 5.4Ghz. Or would the performance increase per 100Mhz cpu speed increase tend to flat-line after 4.3Ghz as the BullDozer/Zambizies tended to do. Early on the charts should tell that story and the screen captures are there from which I gathered the data.
14. > I will also post FULL F12 captures of my bios at each speed. This of course will be from the CHV mobo. There will be only 3 captures of the bios up until 5.0Ghz. At 5.0Ghz I swapped the CPU_LLC from HIGH to Ultra-High and captured that extra page of the bios to show CPU_LLC on Ultra High. So 5.0, 5.1, 5.2, 5.3 and 5.4Ghz were all run on the Ultra-High CPU_LLC setting. The extra voltage and load at the elevated speed, created the need for Ultra-High LLC. Thank goodness for superb cooling since the HEAT is still there even with PileDriver/Vishera Cpus.
By the way I will just add this note here about CPU_LLC since I have made mention of it. Asus says the setting of High is considered 50% CPU_LLC by them and that Ultra-High is considered to be 75% CPU_LLC by them.
15. > I put borders around places of emphasis on the screen captures. Text is added into the captures hoping for clarity. There will be one set of screen captures that includes HWMonitor at the speed I generally run 24/7. That speed should be in the neighborhood of 4.53Ghz, but I am steadily inching toward the 4.6Ghz mark to stay just on the rise side of the graphed performance leading up to 4.6Ghz.
16. > Okay that is just about all the background information I can think to put here. If I find anything else, then I will come back in and edit this first Post.
RGone...
Attachments
Last edited: