Look I don't want this to get out of hand I just want to say thanks for the challenge and everyone has contributed some valid points. My explanation there was to simply enlighten you to the fact that NB performance does actually have a big impact on graphics performance. For the sake of accurate information to other members and over clocker enthusiasts this is an important point. Here is your position, quoting from your 1st and 2nd response:
you are completely out to lunch on the level of impact this has. While you can be theoretically on the right track, without actual testing data or any specific real world numbers to properly frame your opinion
What you are missing here, in a CF situation, the bottleneck is the CPU's computational ability - that is improved in a very moderate way by increasing CPU-NB speeds, but it doesn't eliminate the problem.
Of course, the CPU will ultimately be the bottleneck in a CF situation like this, and the computational ability of the cores is not as effected by the NB overclock, which I concede, but that fact is not really applicable since most games utilize only one or two cores (and this is where the NB overclock really shines) Thus, I stand by my original statement that he will have less of a CPU bottleneck (and better FPS) with the high NB speed.
These quotes, direct from the article, should clarify any remaining doubts:
What we did not expect was that a similar pattern would emerge from our CPU-NB testing. Our experiment with Thuban’s CPU-NB was truly strange. In applications that can take advantage of as many cores as 6, the gains achieved by CPU-NB overclocking was rather small. It’s there, but not to the point of writing home about. On the other hand, overclocking CPU-NB greatly benefited less-threaded applications, namely games. This can also be observed from the X264 HD 3.0 test, where the less intense first pass yields much better scaling with CPU-NB overclocking than the second pass.
and:
Both at 3.2 GHz and 4.0 GHz, the CPU-NB overclocking yielded 16% more performance in this game without moving an inch of the CPU frequency.(StarCraft 2, very cpu intensive, with NB @ 3.0ghz)
If you have a minute, can you post up some some results of the testing you did showing that it will improve FPS in games (canned benchmark is best) or even synthetic benchmarks (focusing on the GPU score). I've seen both sides of this discussion before but not really any benchmarks.
Thank your for your message Stormchaser. If you had looked, I plainly stated that I know this is not a phenom or a thuban but I do believe that it plainly shows that although the physics scores were quite varied depending on frequency( more CPU dependant) the actual graphics scores didn't.
The Article you refer to is talking about starcraft which is terribly dependant on the CPU so naturally the score would improve as CPU performance improves. Which is exactly what I.M.O.G. is saying.
Now if you took a new title such as Crysis3, which will utilize a multicore processor, the NB would make very little difference in gameplay. The GPU is fully loaded and the NB won't make it perform any better.
So like you in the spirit of wanting to make sure members are getting accurate information
Look I don't want this to get out of hand I just want to say thanks for the challenge and everyone has contributed some valid points. My explanation there was to simply enlighten you to the fact that NB performance does actually have a big impact on graphics performance. For the sake of accurate information to other members and over clocker enthusiasts this is an important point. Here is your position, quoting from your 1st and 2nd response:
I took an hour last night and ran some synthetic benchmarks which utilize both the CPU and GPU to give the members something to see. That in my opinion plainly shows both sides of the arguments and may make things clearer.
My intention was not to "clutter the thread with irrelevant benchmarks" I felt it was pertinent information or I would have "started my own thread".
If you had looked, I plainly stated that I know this is not a phenom or a thuban but I do believe that it plainly shows that although the physics scores were quite varied depending on frequency( more CPU dependant) the actual graphics scores didn't .... ..... .....
I took an hour last night and ran some synthetic benchmarks which utilize both the CPU and GPU to give the members something to see. That in my opinion plainly shows both sides of the arguments and may make things clearer.
My intention was not to "clutter the thread with irrelevant benchmarks" I felt it was pertinent information or I would have "started my own thread".
Well you filled two posts with just a few words of your own and irrelevant benchmarks. If that's not clutter, I don't know what is. Its just in this case he (the OP) has a Thuban based system, so we should keep all benchmark results associated with that platform. Since NB overclocking performance varies between chipsets and their respective cpus, I'm not sure how you can justify drawing valid conclusions from FX benchmarks for a separate platform, such as the Thuban.
Now, if you have any Thuban benchmarks that justify your position regarding NB speed on gaming performance, I would encourage you to post them up. Until then, I will have to agree with the Anandtech research, and I would encourage you to study the gaming results closely in that Article - (in particular the Test Part One). Here is just one (of many) quotes direct from the article that backs up my research and position regarding NB overclocking related to gaming performance:
At the same time, the gains made by overclocking the CPU-NB seemed too good to be true, 3.2 GHz CPU / 2.8 GHz CPU-NB combo surpassing 4.0 GHz CPU / 2.0 GHz CPU-NB combo. We ran the benchmark over 10 times on both ASRock’s 890FX Deluxe 4 and the ASUS M4A89GTD Pro/USB3 and the trend was the same, CPU-NB overclocking benefiting the game as much as, if not more than, the CPU core overclocking. Overclocking CPU by 800 MHz nets an 8% improvement in FPS, yet overclocking CPU-NB by 800 MHz got us 9% FPS increase.
The Article you refer to is talking about starcraft which is terribly dependant on the CPU so naturally the score would improve as CPU performance improves. Which is exactly what I.M.O.G. is saying.
Now if you took a new title such as Crysis3, which will utilize a multicore processor, the NB would make very little difference in gameplay. The GPU is fully loaded and the NB won't make it perform any better.
So like you in the spirit of wanting to make sure members are getting accurate information
What you are missing here, in a CF situation, the bottleneck is the CPU's computational ability - that is improved in a very moderate way by increasing CPU-NB speeds, but it doesn't eliminate the problem.
You say naturally as the CPU clock goes up the game performance goes up. This is of course true, but we need to isolate the NB to see what bearing this has. This is exactly what Anandtech has done. They are not changing the CPU clock, only raising the NB speed from 2.0Ghz to 3.0Ghz. If you were correct in your assumptions, the FPS should remain at a near constant throughout the test, but instead as the NB is raised, the FPS increases accordingly. Sure there are a handful of games that will use 4 cores, but not that many. I've already conceded that NB overclocking does not have a great effect on multi core computations, but since most games use only one or two cores, NB overclocking will really help out. Here is the quote which address the multi core question you and IMOG raised:
What we did not expect was that a similar pattern would emerge from our CPU-NB testing. Our experiment with Thuban’s CPU-NB was truly strange. In applications that can take advantage of as many cores as 6, the gains achieved by CPU-NB overclocking was rather small. It’s there, but not to the point of writing home about. On the other hand, overclocking CPU-NB greatly benefited less-threaded applications, namely games.
Here is a good graph to illustrate NB clocking on gameplay performance:
And in further affirmation, so as to alleviate any remaining doubts, this quote is direct from the Anandtech article, directly explaining the above chart:
Both at 3.2 GHz and 4.0 GHz, the CPU-NB overclocking yielded 16% more performance in this game without moving an inch of the CPU frequency.
And what they are saying is game performance (and FPS) is increasing quite substantially with a simple NB overclock from 2.0Ghz to 3.0Ghz, without even changing the CPU clock speed.
I'd encourage you to run some gaming benchmarks (Thuban related) to find this out for yourself.
Chiefly, I would ask IMOG and Johan45 to explain this Anandtech finding:
Overclocking CPU by 800 MHz nets an 8% improvement in FPS, yet overclocking CPU-NB by 800 MHz got us 9% FPS increase.
Can you test this in CFx which is the context in which the thread is about? What games need and what GPUs need, as IMOG explained, are different it seems.
Both at 3.2 GHz and 4.0 GHz, the CPU-NB overclocking yielded 16% more performance in this game without moving an inch of the CPU frequency. Of course in both the cases the CPU-NB overclocks were massive
And what they are saying is game performance (and FPS) is increasing quite substantially with a simple NB overclock from 2.0Ghz to 3.0Ghz, without even changing the CPU clock speed.
They said that game... not all games or plural even. In x264.. there were no gains so it seems situation dependent to me which warrants some actual testing in the conditions/context of the thread, no?
That said, you may be right.. I dont know. I would just like to see CFx tested to figure it out once and for all. I'd do it, but I'm AMDless.
I'm willing to do some testing as well if ya'll can wait until after my exam this week (and also for the minor issue of actually getting my system back up and running). Any benchmarks that may simulate what is going on with the performance increase in starcraft?
CFx testing that brought this discussion up... synthetics, games... but CFx and the subsequent results of raising the NB has on that setup. Do you have 2 GPUs?
Can you test this in CFx which is the context in which the thread is about? What games need and what GPUs need, as IMOG explained, are different it seems.
They said that game... not all games or plural even. In 7zip and x264.. there were no gains so it seems situation dependent to me which warrants some actual testing in the conditions/context of the thread.
That said, you may be right.. I dont know. I would just like to see CFx tested to figure it out once and for all. I'd do it, but I'm AMDless.
Yes, I admit that additional games and CF testing should be done. But we can clearly still demonstrate that NB overclocking plays an important role in gaming performance on a Thuban system. After all, every game that Anandtech tested, resulted in a substantial increase in FPS, not just the one quoted above. Maybe slightly less of an increase for multi-core games, but we must keep in mind that most games only use a couple cores, not all 6.
In the case of H.A.W.X, the FPS increase was dramatic, going from 181FPS @2.0ghz NB to 198FPS @ 3.0ghz NB, with the CPU static at 4.0Ghz. Chart below:
My findings are valid to this thread because the OP has a 4.0Ghz Thuban with Radeon series cards and the NB overclocked to 3.0ghz. AnandTechs findings are valid to this thread because their NB testing is performed on a very similar system with nearly identical specifications, less one radeon card, and shows marked improvements in gaming performance. My research has resulted in the same findings as Anandtech. I must reiterate AnandTech's conclusion and I don't see how a CF system would neutralize the NB's effect on gaming performance.
In conclusion, I will offer up this quotation one more time that speaks to your doubts, earthDog. It is the summation of my findings as well as AnandTechs:
(direct from Anandtech Article) What we did not expect was that a similar pattern would emerge from our CPU-NB testing. Our experiment with Thuban’s CPU-NB was truly strange. In applications that can take advantage of as many cores as 6, the gains achieved by CPU-NB overclocking was rather small. It’s there, but not to the point of writing home about. On the other hand, overclocking CPU-NB greatly benefited less-threaded applications, namely games. This can also be observed from the X264 HD 3.0 test, where the less intense first pass yields much better scaling with CPU-NB overclocking than the second pass.
I think its very clear that IMOG has quite a few more counter points to answer than I do. He' posts runs of Pi and I post real FPS testing of actual games from a reputable source from an actual Thuban based machine. We are not talking about runs of Pi here. We are talking about gaming performance. Minus 1 video card? So what, it still speaks to the core issue at hand.
As of yet I haven't seen any published or reputable Articles speaking to his position. (feel free to add one of your own, if you want to side with him). The majority of testing and benchmarks point to my understanding of NB overclocking and how it in fact, greatly improves gaming performance.
Until you can come up with hard evidence you need all you are saying is just conjecture. And I will again refer you to Anandtech's findings. After all, you really haven't explained the last quote from my prior response as of yet. Which is very much in the context of this thread. As I said before, a CF system would not neutralize the NB's overclocking improvements on graphics performance.
Not siding with anyone... Just want to see some testing of what we are talking about.
I don't think anyone will have trouble swallowing anything when testing is completed and actually proves something within the context of the thread. Both of you bring up valid points as far as I am concerned.
Perhaps a mod can spin this off to its own thread and we can get a member here to help us out.
Don't know about Thuban's, but I checked with a PhII 955BE@4GHz and 5830's xFire.
NB@2000MHz and NB@2800 with Dirt3, Crysis WH and F1 2012.
I had none to very little improvement in FPS@1080 everything maxed out.
By very little, I mean a couple of FPS max (Crysis).
The benefits demonstrated from that very limited example are the product of going from choked ram speeds and stock cpunb to fast ram speeds and well over clocked cpunb.
I think the point storm-chaser makes accurately is that ensuring the memory subsystem on thuban isn't choked yields tangible improvements for gaming. Not huge, but in at least a couple situations which he can show benches for, it has worthwhile gains. I wish he wouldn't leave out the ram frequency also being increased considerably as well to obtain what gains there are.
Its just everything beyond that which is overstated, especially the part about cpunb speed eliminating the CPU bottleneck when running crossfire 7950's on a 4ghz thuban... Which happens to be the topic of this thread.
Thanks for the discussion, interesting stuff. I will stick with my findings. If you do some additional searches on the web for thuban NB overclocking, you will see plenty of positive performance gains that other's have posted about and experienced. Most games will in fact, benefit from quicker L3 cache. The general trend, even with the same ram speed, is higher FPS, and that's what we are looking for. Given his (the thread starters) configuration I was just trying to illustrate how his system is optimized for gaming, with the NB running that high he will see improvements over a stock clock of 2.0ghz.
I will at some point try to get around to running benchmarks on my system which is nearly identical to the OP's so we can isolate the NB completely.
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.
