- Joined
- Jan 12, 2012
I wanted to see if I had the same result as you did in post#87, I hadn't seen your new test until after I posted.
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
Testing does FX-83xx seem to flat line after 4.3GHz
- Thread starter RGone
- Start date
OP
- Joined
- Apr 20, 2004
- Location
- JAX, Mississauna
- Thread Starter
- #102
So is my overclock a defeat? Nope. I use my overclock to 4.8Ghz when rendering video with 8 cores since it has bulging muscles that way with my parts and pieces. I do try and stay at the tops of that graph I posted earlier and is why I run very close to 4.6Ghz and again at 4.8Ghz. I want the most bang for the heat and Vcore that is reasonable.
Reasonable? I have not tried it but in the thread earlier, I spoke of the air cooler I bought for my cpu. I will eventually get around to testing it. I fully expect to be able to run 4.4Ghz and 4.6Ghz stable with an air cooler. Nope it is not a $30.00 dollar cooler, but after all I am wanting to overclock and buy parts according to that purpose for best results.
RGone...
- Joined
- Jan 6, 2013
- Location
- Edmonton Alberta Canada
so would the ideal results be that a mild CPUoverclock will give better results then a high over the top large overclock number? Thats kind of what i'm getting from this?
at 4.3ghz I scored better then 4.4ghz and the only change in the whole setup was the CPU multiplier. Voltage manually set in bios remained the same.
now do some of the Auto settings in the Bios have an effect on that? ( I made mention of my NB and VDDC ect being set on Auto throughout my overclock to 4.4ghz and that fluctuations in those settings may be a factor in the overall benchmark score.
Are we trying to determine how much overclock and to what level best affects graphics performance on a computer? It seems that it does affect it but it isn't completely relative to a higher overclock on the CPU. It will generate a higher physics score and combined score in 3dmark11 but is it entirely responsible for generating a higher GPU score as well when a GPU is solely run in a benchmark.?
at 4.3ghz I scored better then 4.4ghz and the only change in the whole setup was the CPU multiplier. Voltage manually set in bios remained the same.
now do some of the Auto settings in the Bios have an effect on that? ( I made mention of my NB and VDDC ect being set on Auto throughout my overclock to 4.4ghz and that fluctuations in those settings may be a factor in the overall benchmark score.
Are we trying to determine how much overclock and to what level best affects graphics performance on a computer? It seems that it does affect it but it isn't completely relative to a higher overclock on the CPU. It will generate a higher physics score and combined score in 3dmark11 but is it entirely responsible for generating a higher GPU score as well when a GPU is solely run in a benchmark.?
- Joined
- Mar 12, 2002
RGone said:
If your getting a drop in points in that range your CPU is generating errors, just not ones serious enough to crash the system, or your motherboard is throttling the chip.
If the issue is not thermal, or APM related you should try what RGone mentioned and bump the vCore up one level it will likely solve this issue you are seeing.
Also be aware that both CPU and GPU scores will vary from run to run even if you dont change any settings. Its quite possible to get a run in the higher range of one speed setting, and then a low score on the next. If you only ran one test at each speed your comparison will have an undefined margin of error.
They are discussing the clockspeed scaling of the Vishera series CPUs in this thread. There is no "Best overclock" every chip is different and every persons needs are different.
Clockspeed scaling is a measurement of how well a processor continues to produce more work when its clockspeed is increased. Bulldozer based CPUs begin to suffer clockspeed scaling degradation beginning around 3.8Ghz and after 4.3ish the curve becomes very steep. This does not mean that the CPU cant do more work by increasing the speed, but it DOES mean that each step forward yeilds a smaller gain than the speed jump before it.
RGone has gone to lengths to show that Vishera maintains a relatively stable clockspeed scaling up through 5.4Ghz. Does that mean people should overclock that high? Not really, but it does mean that IF you wanted to AND you have a chip that can run that speed it will yeild a proportionately faster result from a lower clockspeed.
They are also discussing the importance of CPU-NB overclocking, and HT link speed overclocking, but this is more of a side note. It appears that RGone is now testing whether a lower core count sample will display the same results.
- Joined
- Jan 6, 2013
- Location
- Edmonton Alberta Canada
I'm very curious to see what Rgone will come up with running the 6300 and benchmarking with 3dmark11. Will the outcome be like the original chart in this thread showing the small increases and ups and downs or will it be a constant upward curve for the GPU?
- Joined
- Jan 6, 2013
- Location
- Edmonton Alberta Canada
I'm soaking up as much of this as I can and even at times some of it is over my head!
The Vcore in my benchmark may be of interest for sure and maybe it would be a good idea to run 3 4.3ghz benchmarks and take the average, get to 4.4ghz and run 3 more benchmarks with Vcore adjusted accordingly to get a less skewed result.
Basically in that respect you are saying I could very well break P10 000 if I have appropriate voltage to remove the errors.
I'll back off this thread and see what comes of it , I think i'm bringing a little bit of clutter to some very useful information especially with my lack of experience with much of it.
OP
- Joined
- Apr 20, 2004
- Location
- JAX, Mississauna
- Thread Starter
- #107
Okay in an effort n0t to get my cart up in front of my horse, I have had to now go and look at stability with this FX-6300. MOST but not all the settings are able to remain as my FX-8350 but it certainly seems that Vcore is going to be an item that has to be checked.
I have already noted in a post up above that this FX-6300 needs more Vcore to run 4.4Ghz than my FX-8350 and that IS so for sure. One thing the benches of the FX-6300 are going to show is the 'real' stair-step upwards in Vcore. On my board for sure.
Every 200Mhz on my board, with this FX-6300, is bringing with it the need to up Vcore. A wall of sorts unless Vcore is raised. This particular FX-6300; cpu is exhibiting the stair-step of Vcore to cpu speed in a very defined and much more easily seen manner than my FX-8350 did. Such characteristics have caused me to have to slow down and go and run P95 Blend on 4.4, 4.6, 4.8 and 5.0Ghz to see just what sort of Vcore is needed to run the numbers without the Vcore to cpu speed becoming an issue. This cpu has fooled me and is taking bigger jumps in Vcore than I was used to and subsequently the Bench is hanging and in general acting up.
Tomorrow most likely before I can have the Heaven Benches up using the FX-6300. I only believe it will be Vcore related and not bus or any other voltage. We will see what we see.
RGone...
EDIT:
Thanks to 'ssjwizard' for his post #104. That is a good read in itself.
END EDIT.
I have already noted in a post up above that this FX-6300 needs more Vcore to run 4.4Ghz than my FX-8350 and that IS so for sure. One thing the benches of the FX-6300 are going to show is the 'real' stair-step upwards in Vcore. On my board for sure.
Every 200Mhz on my board, with this FX-6300, is bringing with it the need to up Vcore. A wall of sorts unless Vcore is raised. This particular FX-6300; cpu is exhibiting the stair-step of Vcore to cpu speed in a very defined and much more easily seen manner than my FX-8350 did. Such characteristics have caused me to have to slow down and go and run P95 Blend on 4.4, 4.6, 4.8 and 5.0Ghz to see just what sort of Vcore is needed to run the numbers without the Vcore to cpu speed becoming an issue. This cpu has fooled me and is taking bigger jumps in Vcore than I was used to and subsequently the Bench is hanging and in general acting up.
Tomorrow most likely before I can have the Heaven Benches up using the FX-6300. I only believe it will be Vcore related and not bus or any other voltage. We will see what we see.
RGone...
EDIT:
Thanks to 'ssjwizard' for his post #104. That is a good read in itself.
END EDIT.
- Joined
- Jan 12, 2012
That is because the chip has become accustomed to MORE V CORE, this is me making the assumption of who the lender was
- Joined
- Oct 21, 2012
- Location
- Phoenix, Az
what i like most about this thread is the level of detail and attention to solid repeatable testing.
groundbreaking stuff. nice to know performance of the Vishera scales in accordance with it's clock, almost perfectly. As far as I can tell, cpus haven't done that since the k7... that's really encouraging results actually.
which means clock speed really is king for this chip. Gets my blood flowing just thinking about it. What someone should do is clock one up a bit higher and see if there is a point this stops being true. Cause clearly 5.4 isn't the ceiling on this chips performance.
groundbreaking stuff. nice to know performance of the Vishera scales in accordance with it's clock, almost perfectly. As far as I can tell, cpus haven't done that since the k7... that's really encouraging results actually.
which means clock speed really is king for this chip. Gets my blood flowing just thinking about it. What someone should do is clock one up a bit higher and see if there is a point this stops being true. Cause clearly 5.4 isn't the ceiling on this chips performance.
OP
- Joined
- Apr 20, 2004
- Location
- JAX, Mississauna
- Thread Starter
- #112
If you remember from the first post of this thread there is some wonkiness out between 5.1Ghz and 5.4Ghz. From what I could tell the performance went dopey in that range and then recovered somewhat. The recovery is speculation since that FX-8350 was at the very edge of being able to run on reasonable voltages without having to go sub-zero as 'ssjwizard' has said. Phase change being the most likely cooling that could be used for extended periods of running.
And it would take a hefty phase change unit to deal with the heat of the FX processors. Not just an old SS unit that dealt with the heat of 2.5 year old Intel heat. More speed is not the real answer but 'more' work done for each cycle of speed already in use.
More completed IPC is the real answer. And we are not going to see that anytime soon. I doubt we see anything new or better for the desktop until toward the very end of 2014. That means all the rest of this year and most of 2014, we are about at the peak of anything AMD. For the pure desktop anyway. Winders 8 did not help the performance of the AMD cpus as most have found out. That was just hype from a year or so ago. I started to buy Winders 8 when it was so cheap at release but decided that was just a waste of even that small monies for me. My thinking and others can do with their money as they please to do.
A floating point unit for each core of the AMD cpu would have been great, but AMD is designing for servers where the money is and not for the desktop user or clocker. There are just too many 'other' places where the coming money is thought to be for a 'strapped' or appears 'strapped' company to focus too much energy on.
Oh there are more ideas we can try with the Vishera processor. Disabling cores maybe and looking into any other thing that might be done for more performance where we can run the processor. Such is the actual reason for this thread. We can sort of play at things that are not the normal run of the mill, gimme some numbers to key in for my overclock and lemme out of here.
RGone...
- Joined
- Oct 21, 2012
- Location
- Phoenix, Az
if there is a gain to be had in disabling anything it would be in disabling the 2nd core in each module... scaling the chip back to a 4 core. That apparently (if it can be done) should increase the IPS.
Just not sure how much it will scale if you can do it.
Just not sure how much it will scale if you can do it.
OP
- Joined
- Apr 20, 2004
- Location
- JAX, Mississauna
- Thread Starter
- #114
Hurried to get done...
...As said yesterday, I had to go back and do some down and dirty stability runs to determine just what Vcore this FX-6300 needed to run stable. Takes a little more than my FX-8350 does but got it sorted anyway.
I also said that this FX-6300 showed easily that inceasing the Vcore is needed at each 200Mhz jump in cpu speed. Pretty easily seen stair-step up in Vcore required for stable increase of 200Mhz.
I am in a rush to RMA this 6850 that kills my sleep mode. Did not know there was a history of that with this "my price point" video card but there is. One of the later drivers even mentions same as it being a "fix" for such crap. NOT. Not with this one anyway. So I don't have time to swap benching programs in mid-stream. Fortunate to get all these done with Heaven DX-11 and posted up.
Of course I had all the captures right on my desktop where they were handy for viewing and I don't see a real difference in the scores between this FX-6300 and my FX-8350 when benched at the "identical" speeds and settings .So if gaming were the only criteria for use in an AMD setup, the FX-6300 might well be the better buy since it seems to run cooler overall. That is a consideration if one is using air-cooling for sure.
Two visible screen captures and the rest hidden in spoiler. Click the spoiler for the other 3 captures with each video card used.
RGone...
FX-6300 4.4Ghz 5770 Video Card.
Click Spoiler for more captures.
Okay FX-6300 with Stock 6850 Video Card.
Click Spoiler for 3 more captures.
...As said yesterday, I had to go back and do some down and dirty stability runs to determine just what Vcore this FX-6300 needed to run stable. Takes a little more than my FX-8350 does but got it sorted anyway.
I also said that this FX-6300 showed easily that inceasing the Vcore is needed at each 200Mhz jump in cpu speed. Pretty easily seen stair-step up in Vcore required for stable increase of 200Mhz.
I am in a rush to RMA this 6850 that kills my sleep mode. Did not know there was a history of that with this "my price point" video card but there is. One of the later drivers even mentions same as it being a "fix" for such crap. NOT. Not with this one anyway. So I don't have time to swap benching programs in mid-stream. Fortunate to get all these done with Heaven DX-11 and posted up.
Of course I had all the captures right on my desktop where they were handy for viewing and I don't see a real difference in the scores between this FX-6300 and my FX-8350 when benched at the "identical" speeds and settings .So if gaming were the only criteria for use in an AMD setup, the FX-6300 might well be the better buy since it seems to run cooler overall. That is a consideration if one is using air-cooling for sure.
Two visible screen captures and the rest hidden in spoiler. Click the spoiler for the other 3 captures with each video card used.
RGone...
FX-6300 4.4Ghz 5770 Video Card.
Click Spoiler for more captures.
Okay FX-6300 with Stock 6850 Video Card.
Click Spoiler for 3 more captures.
OP
- Joined
- Apr 20, 2004
- Location
- JAX, Mississauna
- Thread Starter
- #115
Thanks Fugu man...
Fugu hooked me to some 3Dmark11 so I hurried to pump out some benches on this less that muscular 6850. The overall 3Dmark11 score seems to only give any points bennies for the power of the video card and not for the cpu flexing its' muscle as I saw when I had all the scores in and tallied up.
I don't pay any attention to the bench results until I am finished and all the scores are in the spread sheet. Too much effort to make sure that a level playing field is being maintained and that the bench itself is running the same each time. Rebooting for each test and well you get the drift. Benching is not my most relaxing pursuit.
If I get time before Monday, I hope to try and see how Cinebench tests out since I have quite a bit more time in the saddle with it. Got a lot to do while the FX-6300 is kindly on loan and I still have another mobo to see what gives.
RGone...
Fugu hooked me to some 3Dmark11 so I hurried to pump out some benches on this less that muscular 6850. The overall 3Dmark11 score seems to only give any points bennies for the power of the video card and not for the cpu flexing its' muscle as I saw when I had all the scores in and tallied up.
I don't pay any attention to the bench results until I am finished and all the scores are in the spread sheet. Too much effort to make sure that a level playing field is being maintained and that the bench itself is running the same each time. Rebooting for each test and well you get the drift. Benching is not my most relaxing pursuit.
If I get time before Monday, I hope to try and see how Cinebench tests out since I have quite a bit more time in the saddle with it. Got a lot to do while the FX-6300 is kindly on loan and I still have another mobo to see what gives.
RGone...
- Joined
- Jan 6, 2013
- Location
- Edmonton Alberta Canada
I'm reviewing my results as well to see how similar the results are between each step to figure out the percentage of change vs points gained. Good information for someone trying to reach a specific number on 3dmark11 (i'd like to break P10 000 on a single GPU as an example)
7% change in my scores (P7500 to P8000) earned an increase in points of 500 points on the physics score alone This is the jump from 4.0ghz to 4.4ghz the line is a constant increase and my highest scoring run was P9807 by using GPU tweak and a CPU clock of 4.4ghz. Meaning i'd need an increase of only half as much as that.
0.05% increase should yeild apx P250 points meaning i'd need a CPU overclock of 4.6ghz in my current system to acheive the goal of P10 000. My hardware is not capable of controlling the temps to achieve the score i'm looking for and would have to look into watercooling to make it happen!
Seems like a 7% score increase in Performance mode on 3dmark11 requires a 10% CPU overclock increase.
This is how i'm looking at it. another 5% CPU clock increase to 4.5ghz should yeild another 3.5% increase on my score for a total Physics score of: 8313 and judging by the way the total score is factored the combined score will only increase by 75 to 100 points and if this is scalable to the GPU Tweak runs that I did later on with 3dmark11 and applied to my score of 9807 a 4.5ghz overclock will most definitely not be enough to reach P10 000 but a 4.6ghz overclock will do it.!
Rgone, you can see it in my graph as well. The scores from 4.2 to 4.4 increase slightly slower then going from 4.0 to 4.2 and the large jump between 4.0 to 4.2 is noticeable on your graph as well.
7% change in my scores (P7500 to P8000) earned an increase in points of 500 points on the physics score alone This is the jump from 4.0ghz to 4.4ghz the line is a constant increase and my highest scoring run was P9807 by using GPU tweak and a CPU clock of 4.4ghz. Meaning i'd need an increase of only half as much as that.
0.05% increase should yeild apx P250 points meaning i'd need a CPU overclock of 4.6ghz in my current system to acheive the goal of P10 000. My hardware is not capable of controlling the temps to achieve the score i'm looking for and would have to look into watercooling to make it happen!
Seems like a 7% score increase in Performance mode on 3dmark11 requires a 10% CPU overclock increase.
This is how i'm looking at it. another 5% CPU clock increase to 4.5ghz should yeild another 3.5% increase on my score for a total Physics score of: 8313 and judging by the way the total score is factored the combined score will only increase by 75 to 100 points and if this is scalable to the GPU Tweak runs that I did later on with 3dmark11 and applied to my score of 9807 a 4.5ghz overclock will most definitely not be enough to reach P10 000 but a 4.6ghz overclock will do it.!
Rgone, you can see it in my graph as well. The scores from 4.2 to 4.4 increase slightly slower then going from 4.0 to 4.2 and the large jump between 4.0 to 4.2 is noticeable on your graph as well.
Last edited:
OP
- Joined
- Apr 20, 2004
- Location
- JAX, Mississauna
- Thread Starter
- #117
Thanks again man...
...Fugu it is hard to tell much from a graph of only three points. That understood your Graphics score is falling from 9,054 backdown to 9,056 after a good rise in between and the green line which says score, I assume is the 3Dmark11 score is falling off as well. BUT the Physics score has been nearly a straight-line rise over the same period and yet the score falls. Not that it falls back to the beginning point but it is certainly on the decline. That movement of the 3Dmark11 score is not at all tied to the rise of the Physics score.
Said another way; step back and look at the chart. See the yellow line go up then down? See the green line go up and tilt down? See the purple line rise continuously? The yellow and green lines are following each other and seemingly with next to NO regard for the increase in cpu score. That is why I said this bench only gives a hoot about the graphics card and not about the cpu performance.
No more than I admittedly know about games, this bench tells me only one thing that might relate directly to gaming and that is the video card is king. Otherwise the reality of the situation is that the final score seems to pay NO attention to the increase or likely the decrease of the cpu performance. What other way could the graph be looked at for pure results? None that I see.
All of this stuff is little more than hype when you look at it. What you think MSI and Antec paid to have their little toys within the bench? No telling what was arranged.
Back when I was a part of a team that benched at HWBot (screw it) we never looked at the scores to see what to buy. We knew what to buy to get the bench scores. Hehehe. The biggest baddest setup you could get. Talking big marks and not the niche benching. Which goes on as well.
I will repeat again, that my only reason to come back to benching for myself is to see if the speed all the newbs come seeking, really gives a result for effort. Some I would say but not worth it overall. The key really is to buy a GiNormous video card and just make sure that the rest of the parts and pieces are not in the way. Now that is what one would get if they actually sat and studied the scores. IS that how things are in the 'real' world? I would guess my supposition to be at least 75% to 80% accurate.
That is why I did my best to say to you just get what you can get out of your FX-8350 and be happy since I just did not see the cpu speed to gaming performance based on video benches to mean much.
Now what I have said above is just based on the numbers. I don't care how many benches people run. In truth most charts on websites do not move me one way or the other. I know charts still have to be relevant. They need to be charts made from data that is related to real world use. This one seems certainly not that for pointing to any need of cpu peformance to be able to game. Then of course I could just be blind.
RGone...
...Fugu it is hard to tell much from a graph of only three points. That understood your Graphics score is falling from 9,054 backdown to 9,056 after a good rise in between and the green line which says score, I assume is the 3Dmark11 score is falling off as well. BUT the Physics score has been nearly a straight-line rise over the same period and yet the score falls. Not that it falls back to the beginning point but it is certainly on the decline. That movement of the 3Dmark11 score is not at all tied to the rise of the Physics score.
Said another way; step back and look at the chart. See the yellow line go up then down? See the green line go up and tilt down? See the purple line rise continuously? The yellow and green lines are following each other and seemingly with next to NO regard for the increase in cpu score. That is why I said this bench only gives a hoot about the graphics card and not about the cpu performance.
No more than I admittedly know about games, this bench tells me only one thing that might relate directly to gaming and that is the video card is king. Otherwise the reality of the situation is that the final score seems to pay NO attention to the increase or likely the decrease of the cpu performance. What other way could the graph be looked at for pure results? None that I see.
All of this stuff is little more than hype when you look at it. What you think MSI and Antec paid to have their little toys within the bench? No telling what was arranged.
Back when I was a part of a team that benched at HWBot (screw it) we never looked at the scores to see what to buy. We knew what to buy to get the bench scores. Hehehe. The biggest baddest setup you could get. Talking big marks and not the niche benching. Which goes on as well.
I will repeat again, that my only reason to come back to benching for myself is to see if the speed all the newbs come seeking, really gives a result for effort. Some I would say but not worth it overall. The key really is to buy a GiNormous video card and just make sure that the rest of the parts and pieces are not in the way. Now that is what one would get if they actually sat and studied the scores. IS that how things are in the 'real' world? I would guess my supposition to be at least 75% to 80% accurate.
That is why I did my best to say to you just get what you can get out of your FX-8350 and be happy since I just did not see the cpu speed to gaming performance based on video benches to mean much.
Now what I have said above is just based on the numbers. I don't care how many benches people run. In truth most charts on websites do not move me one way or the other. I know charts still have to be relevant. They need to be charts made from data that is related to real world use. This one seems certainly not that for pointing to any need of cpu peformance to be able to game. Then of course I could just be blind.
RGone...
- Joined
- Jan 6, 2013
- Location
- Edmonton Alberta Canada
I agree that with the 3dmark11 benchmark CPU score doesn't matter and the emphasis is on GPU power!
I was using a situation I was familiar with (mine) and trying to crunch numbers to see what it would take to acheive a slightly higher score which of course means nothing more then bragging rights to my friends lol.
It seems though that you do get a steady increase in Physics score as you climb the graph, it's slow and miniscule but it does go up. In score comparisons even the 1000.00 intel chips don't get a very high physics score!
- Joined
- Jan 10, 2012
I really don't think Rgone is using enough vcore.
some idiot around here really pushes for high clocks and he REALLY, REALLY uses high vcore with great results.
cpu's train to vcore. look at Rgones 8120. they train up very quickly and train down very, very slowly.
some idiot around here really pushes for high clocks and he REALLY, REALLY uses high vcore with great results.
cpu's train to vcore. look at Rgones 8120. they train up very quickly and train down very, very slowly.
- Joined
- Jan 12, 2012
I wonder who you're talking about Caddie
Similar threads
