Real World Gameplay CPU Scaling @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,602
Real World Gameplay CPU Scaling - Four processors; Intel Core i7 920, Intel Core 2 QX9650, AMD Phenom II X4 810, & AMD Phenom II X3 720 BE. Seven games. REAL WORLD GAMEPLAY at stock clocks and all overclocked to 3.6GHz, head to head, & apples to apples.


CPUs are an important part of gaming, it is impossible currently to play a computer game without one no matter how much you are told these processors are becoming a diminishing portion of your gaming experience. We have recently seen the Infernal Engine scale gameplay physics, not effects physics, across 8 threads of an Intel Core i7 processor. We’ve been seeing CPUs trending towards improving multi-core designs, adding more cores, and deepening pipelines instead of simply being focused on clock speed. Does all this matter to the gamer?
Click to expand...
 
Nice to see that the castrated CPUs from AMD (less core and less cache) are keeping up with the Xtreme edition and the latest tech from Intel.
 
alg7_munif said:
Nice to see that the castrated CPUs from AMD (less core and less cache) are keeping up with the Xtreme edition and the latest tech from Intel.
Click to expand...

You might go back and read the article again. We did not exactly come to the conclusion you did above.
 
Great article, very timely for someone like me who just upgraded to a Q9550 (from a E6400 and yes gaming is better) and a bigger Monitor (1920x1080) , now I feel my 9600GT is getting a little long in the teeth....And only having a MB that will support crossfire, I'm thinking of a AMD video card so for my next upgrade I can CF for cheap.....What single AMD GPU could I get now for about $150 that out performs the 9600?
 
Hey, CPU DOES make a difference! But only with $800+ worth of video cards and 30" Monitors :p

Good article, you guys really concentrate on detail which is always nice to see.
 
Does show i7 as a bit of a useless upgrade for most core 2 quad owners - sure it's *slightly* faster, but not enough to warrant the expense of a new cpu, motherboard and memory. I suspect i5 (which is a memory starved i7) is going to have very little advantage over a core 2 quad.

May take 32nm parts to really give most of us a reason to switch...even then they don't look that great (still same number of cores, apart from the too-expensive-to-buy i7 extreme).
 
sbuckler said:
Does show i7 as a bit of a useless upgrade for most core 2 quad owners - sure it's *slightly* faster, but not enough to warrant the expense of a new cpu, motherboard and memory. I suspect i5 (which is a memory starved i7) is going to have very little advantage over a core 2 quad.
Click to expand...

Well that statement depends on clock speed. We have not seen a i7 920 that would not do 3.6GHz on air. Also, you statement pertains only to gaming.

And yes, I have to agree, i5 could not have much on C2Q in the gaming arena at all.
 
So spend all your money on the video card not the processor. Point, AMD. :)
 
Great article. As much of a pain as they probably are to write, the data tsunami comparison articles like this and the thermal paste shootout are extremely informative.

The GTAIV results on the X3 were especially interesting. The chip has gotten such favorable reviews that I'd begun to wonder if there actually were any real world situations for a gamer where that extra core mattered significantly. At least one, apparently!
 
Citizen86 said:
Hey, CPU DOES make a difference! But only with $800+ worth of video cards and 30" Monitors :p

Good article, you guys really concentrate on detail which is always nice to see.
Click to expand...

Well, lower resolutions will lessen the graphics load and shift it somewhat to the CPU, so differences tend to be more pronounced. I'm sure 2560x1600 on a 30" screen looks amazing, but I'm OK with 1920x1080 on my 46" LCD;)

Great comparison. Only possible gap is that there is no dual core, but with dual being mainstream, you could say enthusiast is now quad:)
 
jfreund said:
Well, lower resolutions will lessen the graphics load and shift it somewhat to the CPU, so differences tend to be more pronounced. I'm sure 2560x1600 on a 30" screen looks amazing, but I'm OK with 1920x1080 on my 46" LCD;)

Great comparison. Only possible gap is that there is no dual core, but with dual being mainstream, you could say enthusiast is now quad:)
Click to expand...

Yeah, but lower the resolution and you're going to get an FPS boost. Thus needing less GPU power to get the same results, and thus needing less CPU to sufficiently power it ;)
 
I really enjoyed this article. I just recently built a Phenon II X3 720 system, and it look like it will be a viable gaming platform for quite a while (especially on a 22" monitor).
 
Very good article. I always enjoy the information given by H. It is always very informative. I would be more interested to see the multi core CPU scaling as it applies to single card rather than multi card setups as I am pretty sure most people prefer and tend to use single card solutions. So a test comparing 1, 2, 3 and 4 core cpu's and at differing frequencies illustrating how much performance can be gained by additional cores and additional frequency versus those additional cores. So how does a E8500 dual core at 5ghz compare to a Q9300 at 3.6ghz versus a 720BE at 4ghz versus a old single core A64 3400+ at 3.2ghz, but as they all behave with a high performance single card and so on.
 
So wait, why weren't any RTS/RTT games used in the testing? That's usually where cpu-limitation comes into play...
 
Nice write up. Would have liked to see some RTS games in there. Wouldn't have minded a Core 2 Duo for reference. Also would love to see the same thing done with Nvidia cards for reference. Of course that would have added another year before I got to read the article so I'll take what I can get. ;)
 
Methinks an E8*00 at 4GHz would have given the tris/quads a real run for their money.
 
Good article as always, but since the whole thing is about CPU scaling I would have thought you'd include some dual core chips for comparison. Plenty of us still running them you know! (I just "upgraded" to one!)
 
well; you can build a whole ddr2 based AMD PII system for less than half the price of a i7 920 system; amd system are the way to go for people with only 1 video card::

thanks for the article; didnt know the games were that cpu dependent
 
Great article. Strange not to see the PII 9xx in the lineup, as it could have provided details on cache vs clockspeed. A Q9x00 and a Q8x00 for the same reason. Maybe for fewer games, but at least itd be a comparison point.

I know SupCom has been out of favor for a while, but still disappointing not to see it. CPU still makes a big difference in sim speed.
 
It would have been good to see a Phenom II 955 in there but it is decent article.
 
It's always great to have more info, but I don't imagine the complexity of such an article scales linearly as you add more configurations to the testing. Each time you add something everything before it must be modified to include it in the graphical as well as verbose comparisons (which become more complex each time you add to them), and eventually the [H] crew dies of old age at the test bench before the article is published.
 
Jospeh said:
It would have been good to see a Phenom II 955 in there but it is decent article.
Click to expand...

agreed

although the article mentioned the price difference; it would still be nice to have some higher priced AMD CPUs in the mix; since the total cost of ownership for the system will still be less than any i7 system
 
Decent article but I mean, no 9XX phenom when your using i7's and extreme Core 2's. I just dont get it...also I really wish when overclocking comes in to play that people would also consider the NB clock on phenoms. It's a very important parameter. The difference in my opinion would explain alot of the times where the 720 just did not keep up with the 810. The 810 had 2.8Ghz NB vs the 720's 2.0Ghz stock.

Really you should throw in a 3.6Ghz CPU/2.6GHz NB Phenom II into a review like this to really see the competition :D
 
I'll add another suggestion.
It could use an e8400/e8500 as those highly recommended for awhile and quite a few people run them. It would be also nice to see how the dirty cheap CPUs (AMDx2 7750) compete.
CPU that really need to be put in:pII 955
CPU that should be put in: e8400
CPU that would be interesting to see in there: AMDx2 7750
 
Nice review.
I find the use of an 810 and QX9650 weird choices.
To say the least - 10 bucks up is a 940 BE with a full 6MB of L3 cache (or the 955 for a head to head comparison if you're set on sticking to DDR3).
For the quad core I think the correct comparison would be the Core 2 Quad 9550 to check the price/performance of a relevant CPU (and in a similar price range to the i7 920).
Anyway great review , no news is good news.
 
alg7_munif said:
I'm sorry, I didn't know that I need to reach the same conclusion as yours. :eek:
Click to expand...

Oh you don't! I just don't understand how you could logically make the statement above that you did based on our data. I think maybe your opinion was based on your opinions alone, rather than our data as other forum readers might believe.

enlightenedby42 said:
Great article. As much of a pain as they probably are to write, the data tsunami comparison articles like this and the thermal paste shootout are extremely informative.

The GTAIV results on the X3 were especially interesting. The chip has gotten such favorable reviews that I'd begun to wonder if there actually were any real world situations for a gamer where that extra core mattered significantly. At least one, apparently!
Click to expand...

I would expect to see more in the future as well. But honestly, I think GTA4 is just a POS when it comes to the PC and coding. We have done a lot of testing with it and I think there are scaling issues. But surely multithreaded gaming is coming, but I think you can well get your money's worth out of an X3 720 BE before it is long in the tooth.

jfreund said:
Well, lower resolutions will lessen the graphics load and shift it somewhat to the CPU, so differences tend to be more pronounced. I'm sure 2560x1600 on a 30" screen looks amazing, but I'm OK with 1920x1080 on my 46" LCD;)

Great comparison. Only possible gap is that there is no dual core, but with dual being mainstream, you could say enthusiast is now quad:)
Click to expand...

I think if these CPUs will drive playable gaming at these resolutions, you wil have no problems with lesser video cards and lower resolutions. It just makes sense....

Dual core.....you can still buy those? Seriously, we are looking forward, not back. We are not going to suggest anyone buy a dual core now days unless you are looking for uber-clocks.

jhatfie said:
Very good article. I always enjoy the information given by H. It is always very informative. I would be more interested to see the multi core CPU scaling as it applies to single card rather than multi card setups as I am pretty sure most people prefer and tend to use single card solutions. So a test comparing 1, 2, 3 and 4 core cpu's and at differing frequencies illustrating how much performance can be gained by additional cores and additional frequency versus those additional cores. So how does a E8500 dual core at 5ghz compare to a Q9300 at 3.6ghz versus a 720BE at 4ghz versus a old single core A64 3400+ at 3.2ghz, but as they all behave with a high performance single card and so on.
Click to expand...

We used the video card setup that we did because when you make the games GPU limited, all the graphs are just flat/identical. We did discuss lesser video card setups in the conclusion.

TerranUp16 said:
So wait, why weren't any RTS/RTT games used in the testing? That's usually where cpu-limitation comes into play...
Click to expand...

We don't really give a shit about RTS/RTT gaming around here! ;) Most of the RTS gaming that DOES become impacted greatly by CPU is heavily modded scenarios as we have seen testing SupCom years ago. If you have something to point me towards rather than just an idea that shows it is worth our money to investigate, I will certainly give it a look.

You might check out last weeks RTS DOW article we wrote. It was not challenge for our system.

Until we see true multithreading, RTS is no different than any other game.
1241585304COMHKHr6ym_7_7.gif


vengence said:
Nice write up. Would have liked to see some RTS games in there. Wouldn't have minded a Core 2 Duo for reference. Also would love to see the same thing done with Nvidia cards for reference. Of course that would have added another year before I got to read the article so I'll take what I can get. ;)
Click to expand...

RTS, see above.

Sorry, don't have an extra month to run 42 series of real world gameplay tests to see another GPU that will give us the same results. And yes, you are right, not a year, but more than a month!

Ruiner said:
Methinks an E8*00 at 4GHz would have given the tris/quads a real run for their money.
Click to expand...

I think you are right, but all you guys already own that hardware can tell us all about it. That is what the forums are for. :)

OldPueblo said:
Any way you can just drop in the new 955 Phenom for reference?
Click to expand...

No. I am not sure what you are not able to deduce since PII is already well represented. We show you PII X4 2.6 and 3.6, for 3.2 just guess about in the middle.

Concentric said:
Good article as always, but since the whole thing is about CPU scaling I would have thought you'd include some dual core chips for comparison. Plenty of us still running them you know! (I just "upgraded" to one!)
Click to expand...

You guys own dual cores and don't need us telling you what they will do.

GLSauron said:
Great article. Strange not to see the PII 9xx in the lineup, as it could have provided details on cache vs clockspeed. A Q9x00 and a Q8x00 for the same reason. Maybe for fewer games, but at least itd be a comparison point.

I know SupCom has been out of favor for a while, but still disappointing not to see it. CPU still makes a big difference in sim speed.
Click to expand...

See above, already answered.

mjz_5 said:
agreed

although the article mentioned the price difference; it would still be nice to have some higher priced AMD CPUs in the mix; since the total cost of ownership for the system will still be less than any i7 system
Click to expand...

We are fairly sure you are smart enough to make those deductions for yourself. But if you want to spend a hundred extra bucks to get the name per that you get out of the 810, I say go for it, it is your money. :) And we simply have limitations on time and money spent doing an article of this magnitude. Other sites do not give you this information because it is simply too big of an investment in resources.

Tom128 said:
Decent article but I mean, no 9XX phenom when your using i7's and extreme Core 2's. I just dont get it...also I really wish when overclocking comes in to play that people would also consider the NB clock on phenoms. It's a very important parameter. The difference in my opinion would explain alot of the times where the 720 just did not keep up with the 810. The 810 had 2.8Ghz NB vs the 720's 2.0Ghz stock.

Really you should throw in a 3.6Ghz CPU/2.6GHz NB Phenom II into a review like this to really see the competition :D
Click to expand...

PII is well represented and we give our readers the benefit of being smart enough to figure it out for themselves without having to shown them every baby step of the MHz scale.

Jospeh said:
I'll add another suggestion.
It could use an e8400/e8500 as those highly recommended for awhile and quite a few people run them. It would be also nice to see how the dirty cheap CPUs (AMDx2 7750) compete.
CPU that really need to be put in:pII 955
CPU that should be put in: e8400
CPU that would be interesting to see in there: AMDx2 7750
Click to expand...

When we have unlimited time and money we will get that covered. :eek:
 
altcon said:
Nice review.
I find the use of an 810 and QX9650 weird choices.
To say the least - 10 bucks up is a 940 BE with a full 6MB of L3 cache (or the 955 for a head to head comparison if you're set on sticking to DDR3).
For the quad core I think the correct comparison would be the Core 2 Quad 9550 to check the price/performance of a relevant CPU (and in a similar price range to the i7 920).
Anyway great review , no news is good news.
Click to expand...

Again, the model numbers of what is used is really insignificant to architecture and MHz. We give our readers credit for being able to make some of their own deductions from the vast supply of data given. It might make it easier to understand if you just think of it as PII X3 vs. PII X4 vs. C2Q vs. Ci7 at the given MHz.

I am still unsure why so many of you are not able to see around a model number.
 
An interesting read on how current CPUs hold up considering the big focus on video cards for the last several years. I know time constraints are a factor, but I was surprised to see a Core2Extreme and a X4 810 in the mix; for some reason I expecting the Q6600 (considering how long ago it was released and how its still holding a respectable place) and the X4 920 (seems to be a fairly popular choice).

I think a dual-core comparison to the triple/quadss would of been nice to see the limits of dual core, considering the popularity of the E8400.

Edit: I just saw Kyle's new post; you do make a good point after all.
 
Last edited:
It's all about clock frequency. So does that mean a 7 ghz P4 would spank them all?
 
Kaiyoko-Desu said:
An interesting read on how current CPUs hold up considering the big focus on video cards for the last several years. I know time constraints are a factor, but I was surprised to see a Core2Extreme and a X4 810 in the mix; for some reason I expecting the Q6600 (considering how long ago it was released and how its still holding a respectable place) and the X4 920 (seems to be a fairly popular choice).

I think a dual-core comparison to the triple/quadss would of been nice to see the limits of dual core, considering the popularity of the E8400.
Click to expand...

Again, the model numbers of what is used is really insignificant to architecture and MHz. We give our readers credit for being able to make some of their own deductions from the vast supply of data given. It might make it easier to understand if you just think of it as PII X3 vs. PII X4 vs. C2Q vs. Ci7 at the given MHz.

I am still unsure why so many of you are not able to see around a model number.

Oh and I will say this. We use the CPUs we have on hand. Realize that I had to spec and build systems here that worked perfectly fine at stock and overclocked speeds then ship them from Texas to NC for Brent to pull the gameplay data on. It just gets to a point where we have to work with what we have. Currently we just cannot afford to spend thousands more for a specific model number of CPU, when in the great scheme of things, the model number makes NO DIFFERENCE to our comparison. Insignificant L2 cache changes, make NO DIFFERENCE in real world gaming. This is a look at the core architecture and what you can get as you scale MHz. It is up to you to pick the right model number that will fit your needs and budget.

krotch said:
It's all about clock frequency. So does that mean a 7 ghz P4 would spank them all?
Click to expand...

In gaming, absofuckinglutely. Of course that is just my professional opinion. :p
 
You guys own dual cores and don't need us telling you what they will do.
Click to expand...

I don't exactly own a 4870x2x2 setup to test against your data :p. That being said, I think the biggest thing is people who do own a Core 2 Duo want to know if the higher clock speeds they are getting (4.5 from some people) are still going to show gains. But I think the overclocking the quad core answers a lot of these questions.
 
vengence said:
I don't exactly own a 4870x2x2 setup to test against your data :p. That being said, I think the biggest thing is people who do own a Core 2 Duo want to know if the higher clock speeds they are getting (4.5 from some people) are still going to show gains. But I think the overclocking the quad core answers a lot of these questions.
Click to expand...


Well, let's take a look at what we provided you with today and make some educated guesses. Only in one game did we see threading become and issue, and quite frankly, I think it is because the game is "coded like shit." Any proof of that? No, but that is what I think.

No games out to my knowledge will take advantage of anything beyond a dual core currently.

We have shown here today that MHz is still king when it comes to gaming and CPUs.

Now take all sorts of multitasking and background programs utilizing CPU cycles fully out of the picture and I would suggest that a C2D is going to hold its ground in a pure gaming arena. But I think this is where the dual core actually is going to get thrown under the bus in the real world. Unless you are savvy enough to keep your operating environment mean and lean enough that you are not wasting any resources on any other application or background process, you are going to see dual core gaming suffer compared to triple and quad core.

If you make SURE you have all the other shit turned off and trimmed down to nil, the C2D are going to shine. But in the real world I think dual core guys are going to be giving up some perf somewhere. IMO of course...
 
Concentric said:
Sorry, I don't see how that can be an arguement - you could say the same for Quad core and i7 owners.
The point is to let people who might be thinking of upgrading know what the performance will be like, and i'm arguing that a comparison of popular dual core offerings would be useful, if not fundamental to a "CPU scaling" article.
Click to expand...

Thank you sir. Sorry to have wasted your time with our article.
 
I was just thinking of the cache difference between the 810 and a 9xx series, but I don't know if that would've mattered anyway. Cache has mattered in the past with some games, but not all.
 
Tom128 said:
People who spend their free time reading hardware articles can make the leap I am sure. We see that it's being done and can adjust, but that's also why you are seeing requests for a 9-series Phenom 2 in an article like this. You're taking a i7 and a Core 2 with EXTRA cache, and comparing it to a cache starved or core-lacking competing part. Just looking for an AMD vs Intel showdown with so few what-if's that it truly is a simple matter of Mhz scaling. A Q9550 at 3.6Ghz != QX9650 at 3.6Ghz and PhII 810 != PhII 955. Close sure, but perhaps enough that those 4x AA results would have looked a bit different if the core 2's had been represented by a Q8400 and the PhII's had been represented by a 955. In fact they would have looked alot different.
Click to expand...


Point noted. Lot easier to sit around and "what if" an article to death than actually spend the resources to put one together. Hopefully some other website will fit your needs.
 
You must log in or register to reply here.
Back
Top