StarCraft II Revisited: How Much Gaming PC Do You Need?

[CaptainBib]

Looking at the performance difference between a 460 and a 470 makes me wonder why Nvidia doesn't phase out the traditional 470 and base it off a full gf104 GPU.. sacrifice some tesselation performace, but much better power and temp ratings.

anyway, my 2 cents

 

[the-real-link]

Ah, running an old Quad Q6600 @ 2.4 stock and a 9800GTX+ so after this article, not completely surprised that my 256x256 map w/10 players lags severely. I'm sure it's built to as intensive a level (save starting units) as Tom's map was.

 

[Saljen]

So I run a core i7 920 oc to 3.8ghz, an ATI Radeon HD 5770 1Gb and play SC II at 1680x1050 at Ultra settings. I dont use AA but i still average 45-60 frames consistanly, it runs buttery smooth without any system lag. So does that slightly higher oc on my i7 really make that big of a difference?

 

[kooltime]

Same thing occured when crysis 1st came out was consider a desktop/ graphics killing game. SC2 is just another iteration of that same scenario over time.

 

[Killua]

@Cleeve

If you can't provide the benchmark map, could you do a mini benchmark for me? I know that you said doodad doesn't seem to be the cause of the low performance but I still think its causing some of the problems.

If you could provide a quick test setup for your regular benchmark vs no doodad benchmark. I think Phenom 2 x4 3ghz with 5830 would be a good balance of hardware to test the doodad's effects.

I would be very interested in its effects for a 256x256 map (I'm guessing over 3000 doodads, each doodad requires the same processing power as a marine.)

 

[etrnl_frost]

[citation][nom]greghome[/nom]Those Fermi's were a few months late into the competition, they should perform better[/citation]

By the same token, the Radeon's have had that much more time to mature drivers...

 

[oxxfatelostxxo]

fine on my X58 Extreme3, core i7 920 @ 3.60ghz. 12GB DDR3 1600, 2 evga GTX470SC in SLI, Win 7 64 @ 1920x1080 with aa maxed on ultra.

show off care you maybe share some frame rates with your bragging?

Its ok,, he is still gaming on a 1920x1080 monitor, guess he couldnt afford a nice 30" or even a 26" as those usually are atleast x1200

 

[cleeve]

***I HAVE POSTED THE TOMS HARDWARE GUIDE BENCHMARK TO BATTLE.NET***


TOM'S HARDWARE GUIDE BENCHMARK INSTRUCTIONS
-------------------------------------------------------------------

The first step is to run FRAPS to record the frame rate. Set FRAPS to record the min/max/avg frame rate for 60 seconds. Do not press the key to begin the FRAPS run just yet.

- Start StarCraft 2, and log in to your battle.net account.
- Select 'MULTI PLAYER' button at the top left side of the screen
- In the 'SEARCH BY NAME' search field, type 'toms'.
- Select the 4 player 'Toms Hardware Guide Benchmark' map authored by Cleeve
- Set Game Speed to Normal
- Start Game
- When the game begins, you will immediately see the 'Victory!' message suggesting that the game is over. Click the 'Return to Game' button to begin the benchmark

When I recorded benchmark results, I would center the camera on the protoss mothership at the center of the map in order to be consistent. To do this, select the mothership, press the CTRL and number 1 key to apply the mothership to group 1, and then press the 1 number key to center the camera on the mothership.

As soon as this is done, start the FRAPS utility to begin the 60-second frame rate data collection.

When I take this benchmark, I burn the first results because the initial loading makes for inconsistent benchmark results. In fact, the first time you can just start the game and wait 60 seconds without using FRAPS. After this, exit the map and immediately re-start it. Repeat this three more times and record those three runs to get an average.

If you alt-tab out of StarCraft 2, you will need to burn another run before recording any results as the game will force everything to re-load into memory to cause lags.

Now, I used this benchmark as a single-player map so there might be some variance there, but as far as I can tell you can't search for a single player map. You might have access to the single player map once you've downloaded the multiplayer custom version, I don't know. The single player is preferable because the game seems to pause until you click the 'return to game' button, where the multiplayer version seems to continue whether or not you press the button.

Have fun! Let me know your results.

 

[vvhocare5]

Saying this game is CPU limited appears incorrect. The max framerates would lead one to that conclusion, but looking at the minimum framerates, they are flat between ATI and NV. It appears the game has some sort of clock that is trying to keep the max and min framerates together. It looks like they have a max of 60fps and a min of 30 fps - this has all of the indicators of limiting the max framerate for smoother play.

 

[cleeve]

[citation][nom]vvhocare5[/nom]Saying this game is CPU limited appears incorrect. The max framerates would lead one to that conclusion, but looking at the minimum framerates, they are flat between ATI and NV.[/citation]

That flatness is the reason a strong CPU bottleneck is obvious.

Look at the CPU benchmarks. You will see the min FPS scale very much with different processors. Flat framerates with different graphics cards coupled with heavy dependence on CPU speed is the hallmark of a CPU bottleneck.

 

[IzzyCraft]

game speed normal?

SC2 is most commonely played at Faster, in league matches this does not change and most custom maps also designed under the premise of faster.

 

[cleeve]

game speed normal?

SC2 is most commonely played at Faster, in league matches this does not change and most custom maps also designed under the premise of faster.

You can bench it at faster if you like, but the battle will be over a lot faster than 60 seconds.

 

[FloKid]

I can play SC2, on G-Force 250GTS at little more than 30 fps on 1920x1200 without AA. It works smooth, but, when I play on 2v2 or more it starts slowing down. Surprised that, the new cards can't hold up.

 

[cleeve]

I can play SC2, on G-Force 250GTS at little more than 30 fps on 1920x1200 without AA. It works smooth, but, when I play on 2v2 or more it starts slowing down. Surprised that, the new cards can't hold up.

It's not the card that can't keep up, it's your CPU...

 

[geok1ng]

"As things stand, a high-end Core i7, something in the realm of a Core i5-750, or a fast quad-core Phenom II X4 CPU is the best way to experience StarCraft II."

If the game only uses two cores, it is better to opt for dual cores with plenty of cache and highly overclockable, something like the E8600 and the 32nm dual core i3/I5s. These CPU have at least 3MB of cache per core AND OC on the 4-4.5Ghz range.

 

[numaru7]

the info in this article seems a bit useless to me. Even if you have the best card and proc in the world, chances are, if you're a pro you'll never ever set it on ultra, maybe... just maybe..high. Also, this game will run on almost anything, so the suggestions they have for me equals zero.

As an example, i've tried it on my old PC(Athlon Xp 1.7 OC-ed at 2.6+ and an Ati X1600, 1.5Gb Ram...so it's more than 7 years old), albeit on Low, and it's more than decent in multiplayer - why would u even want to play with AI?!

 

[CptTripps]

[citation][nom]kooltime[/nom]Same thing occured when crysis 1st came out was consider a desktop/ graphics killing game. SC2 is just another iteration of that same scenario over time.[/citation]

Except Crisis had an obvious reason for killing systems, amazing graphics/physics far ahead of their time. Don't get me wrong, SC2 looks nice but I see no reason why it benches so low on high end rigs.

 

[woshitudou]

Here are some screen shots on my 5970 comparing AA on various settings
http://www.xyvo.com/sc2/

 

[nerdbox]

Huh... I find these results interesting, I am running the game on ultra settings full AA with a core I7 OC to 3.5 and a 5870 and have seen not even a stutter. Why in the world is your setup showing such poor performance?

 

[top_playa]

Run Fantastic on my Gigabyte UD7, i7 975X running at 4.2ghz with 6GB Corsair Dominator GT 2000Mhz 7-8-7-20 and BFG GTX 285 OCX,
Crucial 128gig SSD Sata-3 (6gig sata)
All Ultra settings enabled 1900x1200

 

[the-real-link]

Thanks for putting the map online Cleeve!

Results:
11 / 40 / 24
12 / 42 / 24
12 / 42 / 24
(min / max / avg)

I'm running @ 1600x1200 fullscreen. All effects on high and post processing on medium. 512 MB 9800GTX+ / Q6600 Quad @ 2.4 / 4 GB RAM / Win 7 Pro x64.

 

[pjladyfox]

If you want to keep things playable during epic battles, we recommend at least a 3 GHz dual-core or a 2.8 GHz triple-core processor
. We also suggest that you opt for a 3.4 GHz Phenom II or a Core i5-750 at the very least to ensure smooth gameplay at all times.

I'm curious if this applies to those CPU's with a L2 cache below 2 x 1MB per core or those CPU's with no L3 cache available at all? The reason I ask is that I've seen many titles that are more greatly affected by a low amount of L2 or complete lack of L3 more than actual GPU speed of course followed by the lower core count.

This would at least help users try to decide to either get an Athlon II X2 260 or a Phenom II X2 555, or for that matter a Athlon II X4 640 vs a Phenom II X4 545, for their AMD or Intel builds for that matter since I'm sure some are still running older 775 CPU's in their current rigs.

 

[karma831]

Great article. It would be great if you guys could another followup to see how high a person would have to overclock their CPU in order to get rid of the bottleneck.

 

[Shirosaki]

[citation][nom]letsgetsteve[/nom]i wish the test was re-run with a bigger overclock so we could see how cpu limited the game really is and what card will really let it stretch its legs.[/citation]
It's not CPU intensive at all.

Had a 2.4 ghz Intel core 2 duo, 2 gigs of ddr2 ram, and a 460 1 gb gtx, WinXP 32 bit and I was getting 40-60 FPS in Ultra. I upgraded to I5 750, 4 gigs of ram, Win7 64 bit, Same GPU, and I get the same FPS as I did with the old set up. 40-60. There may be a few FPS difference I didn't write the exact numbers down. But it was still in the 40-60 range in the places I did pay attention to. Thought my old setup would be bottlenecking my GPU. Guess not.

 

[cleeve]

It's not CPU intensive at all.

Had a 2.4 ghz Intel core 2 duo, 2 gigs of ddr2 ram, and a 460 1 gb gtx, WinXP 32 bit and I was getting 40-60 FPS in Ultra.

Under what kind of load, though?

Saying it's not CPU intensive because you're running a low load is like saying a jet isn't fast because it sits in a parking lot the same speed as a volkswagen.