Multicore optimization more common; an arguement for the Q9000 and future mobile i7 CPU's

Generally speaking, people assume that a 3ghz Duo is better than a 2ghz Quad for gaming, because (apparently), games still aren't optimized for quad cores. But I'd argue that anything more than 2ghz clockspeed is unecessary for any single laptop GPU up to even the 280M. As well, games are being optimized for quad cores. And the Q9000, becoming a common option and cheaper than the T9600-T9900, deserves better credit.

To test my "theory", I used my overclocked QX9300 at 2.93 ghz and downclocked to 2.13 ghz. To test it as a dual core, I set affinity. The games today are Resident Evil 5 and Clear Skies.

With 4 cores at 2.93ghz, RE5 scores a 56.7 avg.



With 4 cores at 2.13ghz, RE5 scores a virtually identical 57.0 avg. Extra clock speed for RE5, is meaningless with a quad core. This game is optimized well for multicore systems.



With 2 cores at 2.93ghz, RE5 scores a 54.5 avg.



With 2 cores at 2.13ghz, RE5 scores a 50.7 avg. A fullspeed dual core still runs slower than any quad core in RE5. And an entry level dual core really starts to hurt.



Now for STALKER Clear Skies, which is the opposite of RE5. It's poorly optimized for even dual cores, running mostly on just one. You would assume then that clockspeed would make a great deal of difference....

At 2.93ghz:



At 2.13ghz.



But there isn't really, in the GPU intensive spots there's hardly a difference, and at night/rain it's miniscule. Thus 2ghz should be more than enough for the majority of gamers, but having a quad core is becoming invaluable.

 

Really interesting! It would be great to have a similar testing for the 10-20 most popular games.

Rep +!

 

Great read with interesting results starts to make you think that even the slowest QUAD core CPU is the way to go, but I agree if maybe some more games could be tested to see if the trend continues would be great.
Still great work, makes you go HMMMMMMMM!

 

Maybe other users with the QX9300 could contribute with other games

 

Thanks alot mate, really good info, this is much more informative that the '3d' benches, thanks for taking the time.

 

Although I fully agree that Quad Core is the way to go, note that your results are highly arguable since you're only comparing clock speeds. When you set affinity to simulate a dual core, note that you have also halved the amount of otherwise shared resources for the CPU (for example, you would have restricted the available cache to 3MB as opposed to 6MB among other things, which could contribute greatly to the performance since Intel CPUs are cache intensive). A more fair comparison would have been comparing a Q9300 to a P9500 or something.

But certainly, for the gaming enthousiast, going for a powerful quad core over a dual core is a no-brainer (unless if the costs involved are an issue).

 

For the majority of the games I play, from previous testing with my old desktop, they are dual core optimised (especially with more powerful GPU setups) and speed influenced,more than affected by running quad core. In this regards, a cheaper, faster dual core made more sense to me than a much slower quad core.
If anyone tests with something like World of Warcraft they will instantly see similar results (at least in a city for example); the game will mostly maximise the primary core, with a heavy load on the second core. Cores 3,4 will be virtually idle.
By the time the majority of games benefit *well* from a quad core; there will be more options on the market; and we likely will have upgraded anyway.

I dont disagree that a quad is highly useful in some scenarios; however at the moment, the difference is a little superfluous. When you consider a lot of gamers are running dual configs in thier laptops, then the difference is even greater (more with Crossfire than SLI but the point is still relevant), there are very few games that show a real difference from upgrading to a quad core; meanwhile even if its minimal; every game can show slight improvements, especially in the minimum framerates with a faster clocked core.

When you throw on top the fact obtaining a QX9300 is nowadays substantially more than a X9100 (2.53Ghz stock, £300-350 vs 3.06Ghz, £170-220) then you're faced with the simple point of view that whilst games continue the transition into dual core (some games still arent), the movement to quad core will still be a minimalist thing in the market.

The majority of new PCs come with dual cores; and this is where developers will focus thier efforts. As Quads become more and more common; then more programmes beyond rare example and professional applications will take full advantage; but until then you are looking quite frankly at the idea of more cores is better; when with parallel threading and multithreading, the opposite has shown to be true; programmers are still having to learn how to effectively multithread; and expecting a complete transition to quad optimised in the next few years seems a little naive; the dual core system has been on the market now for well over 3-4 years; and we are still not seeing a 100% transition to dual core capable.

Please dont mistake me, Im not belittling the chance to get a quad in a laptop, BUT I really dont think its as clear cut as some proponents might argue; thier is still a solid position for any high powered dual core processor in the current market. Whilst I agree, if you're looking at comparable cache and core speed this becomes moot, and all being equal, and price being reasonable I'd go quad, but with the current scenario whereby the market is still focusing on dual core; a higher clocked, cheaper dual core can make a lot of sense. I'd certainly much rather have a X9100 at 3.06-3.5Ghz than a Q9000, 2-2.4Ghz, even with the multicore potential; the low natural clock speeds will limit a lot of graphics card setups out there, and judging by ebay prices; they cost virtually the same, with the quad being about £20 (30-35$ cheaper), which is not much in the scheme of things.

 

GTA4 being the biggest exception to this? My friend has 2.53ghz dual core and can only get about 30fps at 1680x1050, whereas i can get 45-50 at same res with my 2.0ghz quad (we both have the hd4850 gpu). Also the advantage of the quad is that when i have enough money i can pop in a qx9300 to give it that last bit of edge for new games before i think about buying a new pc.

Has anyone heard if are there going to be any more dual core releases from intel? or are they going to be focusing on the i7/i5 processors?

Blazin

 

Aye, GTA4 is the largest exception to this; it was designed with the Xbox360 (tri-CPU) architecture in mind, and is generally a pretty sloppy, unoptimised port. Others that come to mind are Supreme Commander (with multi-core optimiser) and a couple of others; but in the vast majority of cases; a faster dual core offers a much better experience than a slow clocked quad, and even Supreme Commander is fine on a high clocked dual core.
There are very few games that bottleneck themselves in the same way as GTA4, whilst a vast number that can benefit from either a single or dual core processor at other clocks; in the same vein as GTA4, Cryostasis is a game that is completely unoptimised for even a dual core, and a high clocked dual core far outperforms a low clocked quad, there are extremes at either end of the tracks so to speak.

 

I keep on hearing this statement that higher clocked Dual Cores will offer better performance in Dual Core optimised games than lower clocked Quads will yet the majority of Dual Core Optimised games are not even CPU intensive.

I am yet to see any evidence or benchmarks with many different games showing any of these statements to be true so it sounds more like a theory to me.

Over 90% of the games I tested out in my Q9000 performance thread were Dual Core Optimised games and I was still getting frame rates from 60-120 fps in 3D intensive games.

I am beginning to think that some people just simply like the look of a 3.0 Ghz clock and feel more secure even if it might only make a 2 fps difference in a few games.

I would like to see benchmarks of games backing up these statements.

 

Actually I've seen and tested it myself, WoW primarily, unfortunately I dont have any pictures atm, and I no longer play. It really depends on what you play.
It's not just a theory; there is a reason so many people have remarked upon it.

 

It is a fact that the CPU will become the bottleneck if nothing else is bottlenecking the game. When that happens (GTA4 is a good example), the performance of the game would depend on whether the game is sufficiently optimised for dual/quad core optimised. In the case of GTA4, a lower clocked quad core will be less of a bottleneck than a higher clocked dual core, since the game is optimised for three CPU cores.

In the screenshots provided by anothergeek, it can be seen that two cores from a downclocked Q9300 is bottlenecking the game. However, the CPU is no longer the bottleneck with two cores from an overclocked Q9300, or all four cores from a downclocked Q9300.

Also note that it's not just the clock speed that matters here. Architecture, cache size, among other things should also be taken into account. For example, a "dual core" Pentium-D will most likely be destroyed in terms of performance by a single core from a Core 2 CPU regardless of clock speed.

 

''AnotherGeek'', you should test run the Tom Clancy's HAWX Demo Benchmark, that game seems to be really intensive and I noticed very high Core usage!

 

You set the affinity to "make" it a dual core?

Why not just restart the machine with 2 cores disabled. That'd be a more accurate representation.

Plus you're forgetting the fact that some GPU's are more CPU intensive as well and while the game may not use the extra clock speed, the GPU does benefit from it.

 

Affinity works 100%. Besides having 12MB cache with the QX9300, which or may not be split when setting Affinity, there's no difference from an actual dual core.

The more powerful your GPU, the more powerful CPU you need. If you have sli GTX 280M or CF 4850, your experience will differ from mine. By having a single GTX 280M and producing these results, we can conclude that the Q9000 is an optimal price/performance pairing for even the highest-end of current single GPU setups (with strictly gaming in mind).

A CPU can run at full speed, but it does not mean it is being fully utilized. I'm familiar enough with my computer that by temps alone I can recognize if the game is CPU intensive. Prince of Persia for example, has gotten my temps to mid-high 60C with the full overclock in place. Since it is a console port, and consoles are multicore optimized, this makes sense.

Since console ports are more common these days, generally speaking quad cores are taken advantage of. And in more rare cases, the speed of the CPU has a more profound effect. #1 example being GTA4, but a bad example would be RE5, where we see whether at 2.13 or 2.93ghz, the speed is the same. But in any case the quad core is optimized for, even at the slowest speed of 2ghz, it will outperform a dual core that is 50% faster. And the numbers of games that fit this case is steadily rising.

 

Great work, anothergeek, but would you mind also running the Crysis benchmark and maybe the DMC4 one as well? Those two seem to be pretty popular and may give us a more relative idea.

+rep

 

Thanks for running those AG.

Now we need similar tests for L2 cache. Benchmark a p8700 v/s a p9500 and an OCed Q9000 against a Q9100

 

I'm far from an expert,but I dont trust the validity of this test. Hes not testing 2 different processors,which is the point of the test,hes just making his quad "emulate" a dual core.

You also need to remember that,even with affinity set,the computer will still run background apps with the other 2 cores most likely, which would still throw the dual core test off.

 

Well I've signed up to be the counter argument to the Q9000 when my NP8662 arrives with the P9700 2.8 GHz Core 2 Duo. LaptopNut and I have identical machines except for CPU. I will test what he's already tested plus whatever else makes sense and we will have a good comparison then.

 

My previous gaming benchmarkss were in XP SP3 with 3GB of RAM so I will do a fresh test since I am on Windows 7 X64 RC7100 with 4Gb of RAM now.

I think other good titles to test will be Crysis, Prototype, Oblivion, Street Fighter 4, Batman Arkham Asylum Demo, Brothers In Arms Hells Highway. If you have any more suggestions or ones you think will be a good test we will see what we have.

All of those titles are Dual Core Optimised so it will be very interesting to see the fps.

We should both make sure we have the same GPU drivers, Physx drivers, game settings and Nvidia Control Panel settings too. This will be interesting.

I use Rivatuner in Windows 7 and have setup the OSD so I can always see CPU core usage and fps, hopefully you can do the same. We can note down the range of high, medium and lowest fps we see and also look at the Rivatuner log files.

 

Ah, I see htwingnut has embraced a Clevo.

 

I would test something older,thats NOT optimized to see if the high clock of the dual core effects it.

 

[email protected]


[email protected]


[email protected]

 

I dont get it sgilmore62, what are you trying to show us? Although the change in the minimum framerate is more than I would expect from just a CPU frequency change, thats just the dual core you're showing us...how does that help us compare to quads? Am I missing something?

 

Sorry, getting lazy. Wingnut was saying something

about higher clocked duals not performing any better . You can't really make any inferences about the min framerates b/c it stutters at the beginning of the first loop every time. Had a driver that didn't, forgot which one it was. From my experience a higher clockspeed is the difference between playable @ high settings and not. I ran those benchmarks in dx9, the results in x64 dx10 are more pronounced.

 

laptopnut, htwingnut your competition would be really interesting! and it would provide useful info for future buyers. so let the game begin : )

 

Very good post AG, I fully agree with you here, realistically, looking at it in a different point of view, speed really is capacity these days, i may run some runs of a game that demands both of the GPU and CPU and see how changing affinities, or running dual and quad set up effect the game on the q9000.

GTA would be a good test, even as demanding as it is, we will see a bottleneck with say a duo core at 3 GHz+ against a stock q9000, but I'm not to sure.

 

Were those benchmarks run with SLI enabled?

Also I notice in the Global settings it has benchmark_GPU instead of benchmark_CPU.

I am assuming that you ran the benchmarks with the specs in your sig?

 

Here's my exhilarating Crysis benches... as I had expected, unexciting results.

Quad:



Dual:

 

Hes running it in dx9 mode, i think that could cause a good frame increase, but am not too sure.

 

I don't think DX9 would give much of an increase but SLI would. I noticed his specs say SLI in his sig.

 

For having sli 9800M GTS, his performance is not much better than mine

DX9 or DX10 won't make a significant impact in the benchmarks. It's driver dependent as well. Generally speaking, DX9 for me is smoother, with a slightly higher average and less peaky lows and highs. DX10 spikes higher, but minimum FPS is also lower, and overall just a bit lower average. When it comes to playing the game, DX9 performs better, but I think there's a bit more lag with depth of field than with DX10.

 

A single gtx 280m should outperform 9800m gts sli

 

It's funny you say that, but you are in fact wrong

Now perhaps at lower resolutions the GTX 280M would take the lead. But because of the dual 256 bit memory bus, the sli 9800M handles higher resolutions more efficiently.

 

I guess I should have put the qualifier, "according to this link".

 

Notebookcheck is worthless!

I need to stop using that smilie. Terrible.

 

Notebookcheck is not worthless. Their ranking of cards, however, is.

 

Nope; it depends on the game and drivers. If a game is very SLI friendly, then there is no reason that SLId 9800mGTS could beat a GTX 280m. If the game is not very SLI friendly, then the GTX 280m > SLId 9800mGTS. But "GTX 280m > SLId 9800mGTS" is not a static statement, as it will not always be true, and as the future comes around and games are more multi GPU friendly, the GTX 280m will fall to more and more SLI configurations.

 

Ahh, didn't realise that he was sli'd but yeah, his fps are barely higher then yours, and thats a 2 v 1 scenario.

 

Went ahead and ran 2.26 and 2.545 x64 dx10.

 

some fps drops as expected, but nothing major, may try that tool and try run again on my system, but its quite buggy.

The good thing is crysis is fluently playable

 

Are you stock or overclocked? I get the same FPS overclocked, but I don't play at that speed. So you've still got me beat. I'd suspect an X9100 would put you up a few more frames, Sli tends to be more CPU bound.

 

stock, don't get much benefit from oc'ing the gpu's, definitely get a boost from oc'ing processor

 

I was just wondering whether you have a game called Arma 2? If not they do have a playable demo.

I have been testing this out at native resolutions, normal graphics settings at high quality and patched to 1.03. The game runs at something like 39 fps but easily crawls to 25 fps with AA, shadows and increased view distances.

Sometimes I wonder if it is more intensive than GTA IV lol. Apparently view distance is this game is more true to real life or something.

 

I do have ArmA 2. I loved the original Operation Flashpoint and wouldn't pass this one up. I ll benchmark it.

 

I can't tell if that game is poorly optimised or just very CPU intensive. I just disabled 2 Cores and the game became near unplayable but maybe the higher clock rate of the 2.85 of each Core will prevent this from being an issue because with 3 Cores, this game runs well.

I saw some interesting benchmarks for many CPU's here http://tinyurl.com/mdm6ee

 

I'll give ArmA 2 a try, however, it runs rather well on my desktop E8400 @ 3.6GHz. But I do think it is a poorly optimized game. It has plenty of bugs. I am still miffed that I can't complete the training missions because my soldier keeps getting caught when trying to pick up the injured dude to administer first aid! LOL.

 

This is where the "native quad-core" might matter. I don't know how you disabled your 2 cores or how it works on the Core 2 Quad, but if you somehow disable 1 core on each of the two dies rather than just disabling one die, it might do bad for performance since the two dies communicate through the FSB. According to that benchmark linked by you, it should only lower performance by 20%, which isn't enough in most cases to make it going from playable to unplayable.

 

I disabled the cores by changing the affinity in task manager. When the 2 Cores were disabled this way, there was a lot of stuttering and slow downs, particularly if you changed direction quickly. I class it as unplayable but I guess that is relative to what your expectations are.

That is a very interesting factor that I didn't consider before regarding the Core2Quad that share the same die Vs the ''native quad cores''. In some games, disabling the same 2 Cores doesn't make any difference but in others, there are big slow downs though.

In Arma 2, the frame rates are slower at night for some reason too.