 |
|||||||||||
|
|
|
||||||||||||||||||||||||||||||
Compared to previous CPU scaling articles a few things have changed this time round. In the past we have set out to find what kind of CPU is required to power the latest and greatest graphics cards. We would test the graphics card or graphics cards in question using the highest possible visual quality settings with maximum AA/AF quality enabled. This places much of the work on the graphics card and portrayed the GPU as being the system bottleneck when using high-performance CPUs.
The idea was to determine if a Core 2 Duo processor could deliver the same gaming performance with the Radeon HD 5970 graphics card as the mighty Core i7, and if so, at what operating frequency would the two meet. Such an article is useful for readers who are looking to upgrade their graphics card but wish to know if their current CPU is powerful enough to take full advantage of the upgrade. However this time around we are testing a pair of Radeon HD 5870 graphics cards without enabling any AA/AF quality settings. Furthermore, we are not just testing at 2560x1600, and will instead include a low resolution test using 1680x1050. This will allow the high-end processors to really stretch their legs, and will uncover which processor would provide the best performance assuming no limits. That said, these results will not necessarily reveal which processor is required to power a pair of Radeon HD 5870 graphics cards running in Crossfire mode. This is because the cards are not being pushed as hard as they could be given that AA/AF is disabled. Therefore, while our previous testing methodology may not have shown much of a difference in performance between a Core i7 processor operating at 3.6GHz and 4.0GHz for example, these non-GPU limited results that you are about to see likely will show quite a significant difference. Something else worth noting is the fact that we will not only be including the results for the average frame rate, but also the minimum recorded frame rate. This is not something we have done before when running so many tests, as it is quite difficult to get accurate minimum frame rates. For example, when testing with Company of Heroes with the Core i7 processor at 4.0GHz, we saw a consistent average frame rate of 155fps at 2560x1600. This frame rate would never deviate more than 1-2fps up or down. However the minimum frame rate would fluctuate between 40–80fps, making it difficult to accurately record this data. Occasionally we would get a result as low as 2fps, which we decided to ignore as this quick drop in frame rate was likely caused by something else happening in the background. In the end we decided that the only way to accurately show the minimum frame rate was to take the average of six runs. Typically we take the average from three runs, which is quite time consuming in itself. Still, this was not enough to accurately record the minimum frame rate and we were forced to double our work load. Finally, please note that the flagship processor from each processor series was used to record all data. For frequencies that required us to tweak any settings past their defaults we did so in a manner that would keep them as close to the default specifications as possible, for the fairest possible comparison. Also please note features such as Turbo Mode for the Intel processors has been disabled to keep the processors locked at the intended frequency.
|
||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
|
JohnnyR |
Posted on: 03/18/2010 03:56 AM
Been waiting for the core 2 duo and quad benches for along time. Are they ever coming? Are you not allowed to show them? A lot of us still have so called old tech. |
|
ProX |
Posted on: 03/08/2010 11:26 AM
Thanks for all the hard work. I was amazed at how well the little Core i3 processors do. I am keen to see how these compare in these tests to the Core i5... |
|
Dave P |
Posted on: 03/08/2010 12:42 PM
well done. it looks like my overclocked phenom ii x2 550 with all four cores enabled is enough to get the most out of my radeon hd 5850. i am keen to see how the slower core 2 duo processors go as well. |
|
Jonny Intel |
Posted on: 03/08/2010 01:27 PM
Great article about CPU scaling instead of the usual super jizzed up 1920x1200 8xMSAA/AF 5970 CFX, which obviously is limited by the GPUs. Though you could have thrown in better multi-cores games such as Dragon Age, RE5, Dirt2, AvP, BF:BC2, GTA4 etc. |
|
chemiczny_ali |
Posted on: 03/08/2010 02:55 PM
Who've done those graphs? All graphs are wrong. They've added min fps and max fps. LOL. Someone better fix this cos it's looking a bit embarrassing. |
|
Steve Posts: 75 Joined: 2010-02-08 |
Posted on: 03/08/2010 08:26 PM
Posted by chemiczny_ali on 03/08/2010 03:55 PM
Embarrassing? Hardly they have been arranged by their combined min/average result on purpose. |
|
chemiczny_ali |
Posted on: 03/08/2010 09:22 PM
On purpose? Maybe to mislead the reader? What combined graph shows? Absolutely nothing! If you want to show performance of several processors/gpu's do it right way... http://www.xbitlabs.com/articles/video/display/radeon-hd5830_6.html#sect1 |
|
Dave P |
Posted on: 03/08/2010 10:53 PM
It showed nothing? More often than not the results aligned as they should anyway so I don't know what you are crying about. |
|
Steve Posts: 75 Joined: 2010-02-08 |
Posted on: 03/08/2010 11:04 PM
Posted by chemiczny_ali on 03/08/2010 10:22 PM
I was almost not going to bother but I will. If it were you making the graphs how would you arrange the data, by the average or minimum result? |
|
chemiczny_ali |
Posted on: 03/09/2010 05:30 AM
I would arrange the graphs with the average BUT with adding minimum which is far more important for me. Have you looked at my link? But I'm starting realize that your graphs are not that useless I was thinking at first point. |
|
Telimektar |
Posted on: 03/09/2010 08:22 AM
Those graphs are perfectly fine, thanks for a great article. |
|
Steve Posts: 75 Joined: 2010-02-08 |
Posted on: 03/09/2010 09:42 AM
Posted by Telimektar on 03/09/2010 09:22 AM
You are more than welcome and thanks for the feedback, makes all the hard work worth it  |
|
Dan |
Posted on: 03/09/2010 05:32 PM
Good article. But I do think that the stacked charts are biased and favor the intel processors (in this case). AMD Processors which are usually slower both in mininum fps and average fps are punished "twice" in such a stacked chart and the overall bar length gives impression of bigger performance difference then the actual difference. Also, although the observation that i7/i3 processors consistently provide better minimum fps is very valuable for potential buyer, minimum fps are based on a single data point and I do not think have the same weight as average fps. |
|
graf |
Posted on: 03/09/2010 09:49 PM
The graphs were confusing to me as well, until i figured out what you were doing. If you explained how you were graphing in the text, i missed it. I would suggest putting in a subtitle with something like 'summation of average and minimum values' and/or move the text for the average to the left side of its colored area. Other than that, the article was interesting and informative. |
|
bob smith |
Posted on: 03/10/2010 07:37 AM
sweet article, it helped me with a number of purchase decisions, i was worried about amd x4 965 vs intel i7 920 and now i know the amd will be just as good as the intel for gaming, i also know that i don\'t really need to over-clock to get great performance, i saw that the amd system only had 4 gigs of ram and the intel had 6 gigs, did the intel system really need 6 gigs? did the amd system get close to using all 4 gigs of ram? i am still trying to decide how much ram to get, the graphs are great, at this level of detail for all these tests i almost want a spread-sheet to do sorted comparisons where i can type in the prices to see added value, thank you for the value judgements on overall performance, the conclusions about the intel i3 are as funny as the conclusions about the amd x4 955 and the radeon 5970, i almost fell out my chair laughing when i first thought about buying a $200 cpu for a $600 gpu (i have no idea why my co-workers call me a nerd), newer dx11 games and multi-core game tests will be an amazing next article (supreme commander 2 on 3 monitors), is there any chance you could post some of the huge resolution screen shots of the min fps game-play? i would love to see how many units moving and shooting or huge explosions it took to bring down the intel i7 @ 4GHz to around 30 fps, i am still wondering about the bottlenecks around a radeon 5850, which cpu do you think would be the right match for gaming? not too slow to cause a bottleneck but not too fast to be a waste of money for gaming, i hate bottlenecks, these tests kick ass, i think this shows another example where gaming hardware is way more advanced then gaming software, crysis had the best graphics and stupid ai, i hope game coders realize that they can use the cpu more to make smarter ai while still having great graphics or use the cpu to help with more graphics, like the old unreal engine with 100% cpu graphics, with all this hardware pc games should be way better than ps3 and xbox36, right? |
|
ProX |
Posted on: 03/10/2010 08:41 AM
The Core i7 system is triple-channel and the Phenom II is dual-channel. So one has 3 sticks of 2GB memory and the other has 2. At the end of the day I am yet to see a game that performs better with 6GB of memory when compared to 4GB so this is a none issue, maybe Steve has more info on this. Also for most people 2560x1600 is a huge resolution, add another 30" screen and you can pretty much just halve the performance. Not sure what Xbox 360 games you have been playing but the PC equivalent always looks worlds better in my opinion. Half the reason why Crysis never went to console is because it was way to demanding. |
|
don´t like it |
Posted on: 03/10/2010 09:16 PM
There´s no such information about phenom II NB Overclock speed. In fact the Crossfire Scaling changes a lot with NB overclocked to 2.6ghz or 2.8ghz. Phenom II is bandwith Starved. It´s not true for the I7. Core i7 has turbo mode. Phenom II does not. So when overclock a Phenom II you have to overclock the NB clock ALSO. cheers Rodrigo |
|
Please read this!!! |
Posted on: 03/10/2010 09:32 PM
http://www.anandtech.com/mb/showdoc.aspx?i=3533&p=5 the guys at anand got 12% more speed just with NB Overclock. Well, I got huge 3000 points in 3dmark 2006 JUST WITH NB OVERCLOCK with phenom II. Also I got some 7000 ponts in 3dmark 2001 just with NB OVERLOCK The rule is... if you overlock your processor to 4ghz , you will need at least 2.8 to 3.0ghz NB Speed. |
|
ProX |
Posted on: 03/10/2010 11:04 PM
Clearly neither of you guys can read. First of all they were not using turbo mode. Second of all they said they overclocked the processors as close to their standard settings as possible for fair results. Finally if you look at the 2.0GHz and 2.4GHz data the margins are the same at 4.0GHz so put that in ya pipe and smoke it. |
|
Steve Posts: 75 Joined: 2010-02-08 |
Posted on: 03/27/2010 10:24 PM
Posted by White on 03/27/2010 10:18 PM
Thanks for the feedback. The Core 2 Duo E8xxx series and Core i5 results along with half a dozen other processors will be online in the first week of April. |
|
White |
Posted on: 03/27/2010 09:18 PM
Great job Steve. I've been waiting for this article and you put a lot of arguments to rest. This is the first properly done article I have seen today that compares i7 and Phenom II in gaming performance. The dual core results are interesting as well. Great job! 2 thumbs!! PS: On part 2 if you are doing C2Q please do 45nm. Also, would have been nice to see how i5 fairs with these result, although I expect them to be very similar to i7. |
|
Julian |
Posted on: 04/07/2010 12:37 AM
Great article keep the benches coming! |
|
Pdp |
Posted on: 09/05/2010 11:17 PM
Yes we all know Intel have a better chips at the top end, benchmarking is fair enough but thats just bragging rights. But, in the real world of gaming the i7 965extreme costs £650 the x4 965 £120. But ya sure don't get 5.5 x the performance. On some benchmarks even the little i3 has better performance than its bigger brother. My point being its all about bang per buck. Even the X2 555 gets to within 75% minimum frame rate (crysis) of the i7 965 and that costs a meager £70, thats a wopping 9x price difference. Then you have the possability of unlocking 3 or all of the cores. With a descent motherboard. So its £800 i7/+mobo vs £170 X2 555/+ mobo. I know what I'd choose everytime. The £600 saving I go for X2 HD 5850 and water cooling. For even better performance. Then again any improvement over 60fps you won't notice any way as it's above the refresh rate of 90% of tft monitors!!!!!!!!!!! |