Gaming Performance: What really matters...

[BenchZowner]

If you're a computer & gaming forums freak I'm fairly sure that you've seen people asking if buying a faster memory ( RAM ) kit will give them more fps in their favorite games, others asking how to overclock their Sandy Bridge 2500K to 5GHz for 24/7 usage in order to "remove" the bottleneck placed by the CPU and to "unleash" their graphics card's performance beast.
Not to forget the infamous "HyperThreading myth", I'm pretty sure that you've seen people claiming that you should disable Intel's HyperThreading if you game because you're losing some performance, etc etc.
I've seen every little thing mentioned as a "game changer" without any facts, no proof, no testing done, nothing, just claims from various users on various forums.

I think it's time to find a crazy person to spend a few hours ( like 250 hours of testing ) to put everything to the test:

CPU Frequency with a single-GPU/VGA
CPU Frequency with a multi-GPU setup/VGA
CPU UnCore/NB Frequency
Memory Frequency
PCI-Express BUS Frequency
HyperThreading: Enabled vs Disabled
BCLK: Stock vs Higher ( same CPU, UnCore, RAM, etc clocks )
Sound Card: On-board vs Add-on
etc etc

That's what I did
I used a couple of VGAs, CPUs, RAMs, Motherboards, Sound Cards, and games & game benchmarks in various frequencies, timings, etc.

What you get is... a 23 page article with some basic theory, ideas, and lots of test results.

If you're a gamer you REALLY NEED to read this, not tomorrow, NOW! ( don't worry, it won't disappear, you can read it tomorrow, and the day after that, and the next day )

So... grab a cup of coffee or your favorite energy drink, and read the article now by clicking... here

p.s. Yes, the link says Part 1, there will be a 2nd part, can't tell you when though ( because I don't know when it will be ready )

Your questions, ideas, input is always welcome and really appreciated

[bbmarley]

tldr sorry just went right to conclusion

[cadaveca]

I've been doing the same sort of testing looking for apps that will help me to show differences in motherboards in my reviews.

The only problem I see with your testing is that you only tested with nV cards, but otherwise , your results greatly mirror my own testing.

Also, no AMD hardware, period = me sad.

I do understand that this is a large undertaking, and I really wish I myself had the time to do things similarily, but look, you already have, so I don't have to.


What is important to me that you missed, is reporting throughout, with stock numbers on both CPUs and GPUs(stock for me includes 1333 MHz JEDEC spec ram).

For example, you test how many cores benefits, but you have singlecore results with less clockspeed than the multicore results. Minor issues, but, just things I'd like to have seen.

[qubit]

Awesome.

[BenchZowner]

Originally Posted by cadaveca

I've been doing the same sort of testing looking for apps that will help me to show differences in motherboards in my reviews.

The only problem I see with your testing is that you only tested with nV cards, but otherwise , your results greatly mirror my own testing.

Also, no AMD hardware, period = me sad.

I do understand that this is a large undertaking, and I really wish I myself had the time to do things similarily, but look, you already have, so I don't have to.


What is important to me that you missed, is reporting throughout, with stock numbers on both CPUs and GPUs(stock for me includes 1333 MHz JEDEC spec ram).

For example, you test how many cores benefits, but you have singlecore results with less clockspeed than the multicore results. Minor issues, but, just things I'd like to have seen.

Long time no "see" man

In the core count tests the motherboard didn't like anything except the 38x Multiplier, don't ask me why, it's Gigabyte
Could be a CPU limitation, I need to check it on another board.

Stock GPU vs OCed GPUs, GPU components OC, performance dependency on settings & features, and more stuff will be included in the 2nd part ( don't know when it'll be ready though, got plenty of stuff to do ), also planned is the addition of AMD CPUs & GPUs

Why is the Sandy chip tested at 4.2 against the 4.5 of the Gulftown? In reality, we all know the Sandy chip has more OC'ing potential, and the Westmere has more capacity per-clock. Hence, Sandy makes up for it from it's high clocks as well as lower TDP.

Also, the sound card part is bull. You don't "need" a sound card no matter how much of a gamer you are. I had a SoundMAX onboard riser off an old Abit board; with Matsushita (Panasonic) fine-Gold caps. It rocked through anything. Directsound HAL is dead as well as EAX to a point.

Even then, a good onboard solution can render through software just as good, if not better. It all comes down on component selection. Hardware or software doesn't matter at this time and date. That "You need a sound card!" or "Sound card makes you HEAR DIFFERENT THINGS!" myth has to die already. We aren't in 2003 with crappy AC97's anymore.

[tigger]

Originally Posted by John Doe

Why is the Sandy chip tested at 4.2 against the 4.5 of the Gulftown? In reality, we all know the Sandy chip has more OC'ing potential, and the Westmere has more capacity per-clock. Hence, Sandy makes up for it from it's high clocks as well as lower TDP.

Also, the sound card part is bull. You don't "need" a sound card no matter how much of a gamer you are. I had a SoundMAX onboard riser off an old Abit board; with Matsushita (Panasonic) fine-Gold caps. It rocked through anything. Directsound HAL is dead as well as EAX to a point.

Even then, a good onboard solution can render through software just as good, if not better. It all comes down on component selection. Hardware or software doesn't matter at this time and date. That "You need a sound card!" or "Sound card makes you HEAR DIFFERENT THINGS!" myth has to die already. We aren't in 2003 with crappy AC97's anymore.

I have to agree with you on the no sound card thing.

[stefanels]

very nice article...

[BenchZowner]

Originally Posted by John Doe

Why is the Sandy chip tested at 4.2 against the 4.5 of the Gulftown?

I see the Sandy Bridge clocked at... 3.4GHz ( Stock ), 4.28GHz ( a easy to hit clock for all 2500K/2600Ks, and also chosen on purpose for a clock per clock ( same MHz testing ) against the 980X... ), and 5GHz.
Think you missed the 5GHz test results ?

Originally Posted by John Doe

In reality, we all know the Sandy chip has more OC'ing potential, and the Westmere has more capacity per-clock.

If by "capacity" you mean IPC, or performance in general ( not everybody's comfortable even with basic technological terms ) you're wrong.
At the same clocks, the Sandy Bridge chips are faster than the "equivalent" in Core Count Gulftowns, Westmeres, Nehalems.

Originally Posted by John Doe

Also, the sound card part is bull. You don't "need" a sound card no matter how much of a gamer you are.

I take it you haven't played any game with a decent or at least semi-decent sound engine ( hardware or pure software doesn't really matter in the end ) ( for example Rainbow Six 3, BattleField: Bad Company 2, F.E.A.R. (1), Unreal Tournament 2004 ) with a decent set of headphones and a X-Fi based card.
Apart from the audio quality, there's a distinguishable difference in terms of audio positioning & how well the sound field is defined.

Especially with headphones, both Creative's CMSS 3D Headphone & Dolby's "Dolby Headphone" make a very good difference in games ( as long as you know how to configure the sound card's options, the windows sound output preferences, and the in-game audio settings ).

If you're gaming with speakers... well, you're missing like most of the sound that you'd hear and "position" if you only switch even to a basic set of Philips/Sony/Bose headphones.

Originally Posted by John Doe

I had a SoundMAX onboard riser off an old Abit board; with Matsushita (Panasonic) fine-Gold caps. It rocked through anything. Directsound HAL is dead as well as EAX to a point.

EAX HD 5.0 is partially dead, but what about the games that use it ? You can't "ditch" something like that ( especially if these games are multiplayer games that you still enjoy playing... )

Lots of games still played by thousands if not millions of people out there, are older games with DirectSound, and yes, using it through Alchemy makes a big difference.

Getting an add-on sound card from Creative, Auzentech, Asus will give you much better audio quality ( just don't expect to hear a difference with a "Noname 2.1/4.1/5.1/6.1/7.1/headphone" set ), and you'll also get CMSS 3D* ( *X-Fi based Creative & Auzentech cards only* ) and Dolby Headphone, both of these do wonders in games if you play with headphones.

Those DSP's are golden, it's an excuse for hackers to stop wallhacking ( yes, I can pinpoint the location of any player and anything happening in the map in Rainbow Six 3, BFBC2, F.E.A.R., UT2004, anytime just by listening... )

By the way the "super duper" WOW, gaming edition, special edition, etc on-board audio solutions are still nothing more than a cheap RealTek or Analog Devices chip, along with 2 capacitors, and located on a crowded area along with NICs, Firewire, USB 2.0/3.0 controllers, PCI-Express logic controllers, and yes... lots of EMI.
Any comparison between that and any semi-decent add-on sound card ( gaming or not ) is really, ridiculous.

And remember, the better your audio equipment is, the easier separating bad audio sources and DSPs becomes, and the sound output quality becomes very noticeable

I'm not a salesman, neither an employee of any related company, nor a marketing douche... sorry censored

Originally Posted by BenchZowner

I see the Sandy Bridge clocked at... 3.4GHz ( Stock ), 4.28GHz ( a easy to hit clock for all 2500K/2600Ks, and also chosen on purpose for a clock per clock ( same MHz testing ) against the 980X... ), and 5GHz.
Think you missed the 5GHz test results ?

No, half the tests were done with the 2600 at 4.2, and the 980x at 4.5.

Originally Posted by BenchZowner

If by "capacity" you mean IPC, or performance in general ( not everybody's comfortable even with basic technological terms ) you're wrong.
At the same clocks, the Sandy Bridge chips are faster than the "equivalent" in Core Count Gulftowns, Westmeres, Nehalems.

Obviously it wasn't enough to make up for the difference. Obviously "250 hours" of testing haven't shown it's results. Have you checked what you posted? Gulftown at 4.5 slightly outdoes Sandy at 4.2 in half those benches. So clocks needed to be reconsidered.

Originally Posted by BenchZowner

I take it you haven't played any game with a decent or at least semi-decent sound engine ( hardware or pure software doesn't really matter in the end ) ( for example Rainbow Six 3, BattleField: Bad Company 2, F.E.A.R. (1), Unreal Tournament 2004 ) with a decent set of headphones and a X-Fi based card.
Apart from the audio quality, there's a distinguishable difference in terms of audio positioning & how well the sound field is defined.

Especially with headphones, both Creative's CMSS 3D Headphone & Dolby's "Dolby Headphone" make a very good difference in games ( as long as you know how to configure the sound card's options, the windows sound output preferences, and the in-game audio settings ).

If you're gaming with speakers... well, you're missing like most of the sound that you'd hear and "position" if you only switch even to a basic set of Philips/Sony/Bose headphones.

Have been playing the best implemation of EAX, F.E.A.R (and Perseus) for years. Played it on Elite Pro, SoundMAX and Bravura. SoundMAX IS up there period. It can emulate EAX, A3D whatever you name it. It has Matsushita's for deep sound and bass, which trump the Chinese, crap G-LuXon/Jamicon's of Creative boards.

Originally Posted by BenchZowner

EAX HD 5.0 is partially dead, but what about the games that use it ? You can't "ditch" something like that ( especially if these games are multiplayer games that you still enjoy playing... )

Lots of games still played by thousands if not millions of people out there, are older games with DirectSound, and yes, using it through Alchemy makes a big difference.

By the way the "super duper" WOW, gaming edition, special edition, etc on-board audio solutions are still nothing more than a cheap RealTek or Analog Devices chip, along with 2 capacitors, and located on a crowded area along with NICs, Firewire, USB 2.0/3.0 controllers, PCI-Express logic controllers, and yes... lots of EMI.
Any comparison between that and any semi-decent add-on sound card ( gaming or not ) is really, ridiculous.

I'm not a salesman, neither an employee of any related company, nor a marketing douche... sorry censored

DirectSound is dead, Alchemy reapplies EAX. With that aside, there's nothing SoundMAX can't do over X-Fi. It even has a good ADI panel to tweak surround (instead of CMSS).

The only games that still use EAX on top of my head are UT3 engine and DIRT etc., both of which do it via OpenAL. Add that any sound card can render OpenAL just as good, if not better depending on component selection. The SoundMAX games with those explosion effects on your back YES.

"If you're a FPS gamer you NEED to have a sound card ( CMSS 3D"

No, you don't. This is %99 nonsense. As for EMI, it's nothing. They have EMI shields and even without, amount of EMI simply isn't enough to crackle sound.

You don't know what you're talking about. If you haven't made that statement, you wouldn't have been trying to save yourself.

[BenchZowner]

Sorry I don't feel like talking to a brick wall.

Originally Posted by BenchZowner

Sorry I don't feel like talking to a brick wall.

If that's the best you do, then I'm done here.

[ViperXTR]

hmm, no test on i3 500/i3 2000 series or i5 600 series HT on and off. I think these series are more affected if HT helps or not.
I made a test long ago (GTA4, multi core hungry game) on my i3 2100 with enabled and disabled HT. Framerate is slightly higher when HT is enabled but usually can't really see much difference by playing, but disabling HT actually introduced several stutter/pauses/ lag ingame. I think i still have the comparison screenshot, imma look at it later.

[BenchZowner]

Originally Posted by ViperXTR

hmm, no test on i3 500/i3 2000 series or i5 600 series HT on and off. I think these series are more affected if HT helps or not.
I made a test long ago (GTA4, multi core hungry game) on my i3 2100 with enabled and disabled HT. Framerate is slightly higher when HT is enabled but usually can't really see much difference by playing, but disabling HT actually introduced several stutter/pauses/ lag ingame. I think i still have the comparison screenshot, imma look at it later.

You do realize that just what I tested took already well over 200 hours of testing, and another hundred of chart creation, text writing, screenshot taking/resizing/uploading
Also that I don't have all the time in the world to dedicate it in a single article/review.
I also have no i3/i3 2nd gen and 1st gen i5 at my disposal nowadays.

In part II I will add an AMD FX-8150 and maybe a Phenom II x4 or x6.
It depends on what's going to find its way to my place, and how much time I'll have in my disposal ( unfortunately product reviews have a higher priority than articles ).

[ViperXTR]

Yes i do and i appreciate it, but some games only use around 4 threads as well (in which already covered by the quad core 2600K when HT is disabled, more so for the 6 core 980X), and HT can be quite a deciding factor if it would make things better or worse for these lower end dual core processors since they also use HT.

[BenchZowner]

I can't promise anything, but I might add an i3 2xxx for example.
Don't know when and how far I'll take it.
Just HT on & HT off at stock and "overclocked" probably.

It depends on my schedule and how much free time I'll have to sacrifice.

[Completely Bonkers]

I'm not going to nitpick over a few things I would have done differently, but overall, that is an interesting set of results you have posted, and there are no surprises. And it does, objectively, put to rest a lot of old wives tales about OC'ing or pushing certain components.

You spend a lot of time benchmarking, and putting up those nice flash charts. I'll send you a collective thanks from everyone for that.

How did you do those charts BTW? Have you embedded flash files, or does your server create those dynamically from a set of xml data you have uploaded?

[BenchZowner]

I use Swiff Chart 3 Pro to create the charts ( it accepts manual & automated input from excel sheets ).
It's what PowerPoint is for presentations, only much better at what it does