|
When using this resource, please read FragKrag's opening post. The Tech Support forum regulars have helped create countless of desktop systems without any compensation. The least you can do is provide all of the information required for them to help you properly. |
United Kingdom20275 Posts
I wanted to report some performance information from benching some:
To start out, my current system info: i7 950 @3.8ghz, 216 core gtx260 at stock, 1920x1080
This is a 2v2 replay, 703 supply on map, 237 of that in workers - The majority of units are high tech, for example 9 void rays, 5 immortals, 3 carriers, bunch of archons, probably 15-20 thors, some ultralisks - so it's nowhere near the most intensive bench, i could probably get these framerates in 1v1 if i tried - brood lord/IT in particular has no problems with creating infinite units and wrecking FPS, and every third PvZ in WOL was a max vs max battle with exactly that, god knows how many broodlings and infested terrans, so it's a very real benchmark situation.
I made sure camera position was unchanged and nothing was selected, as well as as some other stuff in the interest of a fair benchmark. I loaded the game on medium - it was not restarted for settings, but i ran through the battle before going back to the same time to benchmark in order to load everything properly with each settings change (it always stutters horribly the first run through)
The data: Low, Medium and Extreme - all with max textures, but Effects, Physics and Reflections off:
Minimum FPS 36. All settings within margin of error, never varied by more than 3fps across a series of tests.
GPU load nowhere near max on any settings, around 50% during the battle on max settings but much lower on lower settings. Average FPS seems slightly higher (5-30%) on lower settings, but minimums seem identical.
Here comes the interesting part: Putting Effects from low to max and physics from off to max cuts FPS min from 36 to 21. I saw massive changes with one or the other off/lowered, but both of them together is where the big numbers come out.
You can word that two ways, it's either a ~42% performance loss from turning them on (FPS cut by 42% from 36 to get 21)
OR
It is a 71% gain from turning them off (FPS of 21 multiplied by 1.71 to get 36)
This shows a lot of important things
Firstly, a 5 year old midrange card being under half load on max settings in a stressful battle (and because of all the 4-6 supply units.. you can get these minimums in 1v1 too or at least close to them) because of CPU limitations - Low and Extreme having the same framerates. Medium and Extreme max the GPU (medium at higher fps than extreme) at the start of the game, but CPU requirements increase far faster than GPU, resulting in FPS dropping as supplies rise and other things happen and GPU requirements relative to CPU dropping until GPU is no longer at 100% load, and then drops lower and lower to the point where a 260 is dipping below 40% load at max settings.
Physics and effects having a devastating effect on performance - even when very few units were dieing - i'd imagine it would be worse in a situation like marines vs tanks, where the physics engine is really active - the Reflections setting apparently affects CPU, but there was nothing to reflect on the part of the map i was on, so it did not seem to change performance or visuals noticably
CPU graphs show as always, primary core maxed (>90% load) and some usage on a second core, later in the game more cores seem to be used some but never much - CPU load averaged across cores during battle even with 1 core maxed does not pass 21% with hyperthreading on (i'd imagine this would be similar for an fx-8k with 8 real cores) but if we say because the hyperthreaded cores are not utilized at all, you can double that and say about 40% CPU usage overall, worst case - 20-25% of that on one core, the other 15-20 spread across the rest. Adding Physics and Effects does not seem to notably change anything here - the only result is FPS being a lot, lot lower.
With minimum FPS increasing with what seems to be linearly (or even slightly faster) with single threaded CPU performance as long as you have 3+ CPU cores, this strengthens the recommendations of going with a "lower end" GPU like the 7770 (about twice as powerful as the 260 used in my examples..) and getting the most powerful CPU possible, preferably unlocked (the i5 3570, i7 3770 perform about the same in sc2) and then overclocking for the performance gain - you can pull out +40% to your minimums here that are otherwise completely unachievable. Those people with extreme edition 6 cores and 3-way SLI 680's? They dont have that performance if they did not overclock - even if they did, the 4-cores will outperform 6-cores per core. It's single threaded performance that's important, dont even get me started on fx CPU's + Show Spoiler +
Any comments/responses appriciated (:
|
I think it's probably something we all knew already, but it's good to have data and a reminder. For those who want to play on higher graphics settings, we should maybe emphasize to turn down physics and effects. They're just eating up CPU cycles and don't really improve visibility (sometimes distract from it).
Some people really want to crank out that "max" setting though.
|
United Kingdom20275 Posts
I think it's probably something we all knew already
Pretty much, but its more precise data and emphasizes the low GPU requirements, real world performance numbers and complete lack of difference between extreme and low in terms of FPS while also emphasizing the massive difference effects and physics make.
|
Tweaking is fun and occasionally useful. For instance with my integrated graphics playing wings of liberty I learned that I could set textures & a few other settings to the best quality but shaders & associated settings were a complete no-no. The Low or High presets are often remarkably poorly optimized.
|
For GPU load it's almost all just the shaders setting I think.
Anyway, I don't mean to downplay the importance of having some numbers to point to, especially given how much of a difference it makes to keep the settings down that are stealing CPU cycles.
|
Cyro - were there graphs/pictures that accompanied that post? I'm not seeing any (apart from the one in the spoiler).
2nd Q - I thought you had an Ivybridge?
|
I've DECIDED! Thank you everyone for putting up with me! 
I went with the cheaper mobo and PSU. I figure it will be good enough. OC to 4.2GHz should be fine enough.
I'm not too certain on the limitations of the MSI G41, but here it goes!
Microcenter
$190 i5-3570k
$90 MSI Z77A-G41
+$23 tax
-$40 bundle discount
$0 shipping (some gas money)
Newegg
$40 COOLER MASTER Hyper 212 Plus
-$20 rebate
$30 Corsair CX 430
-$10 rebate
+$5 tax
+$0 shipping
-$10 new customer email discount
+3$ adaptor cable for CX430's lack of a second PCIe connector.
= $301 !!!
+ Show Spoiler +with a case in the future
(Newegg or Amazon)
$85 Fractal Define R4 Windowed
+$7 tax
+$0 shipping
= $393
|
United Kingdom20275 Posts
On March 25 2013 11:53 mav451 wrote:
Cyro - were there graphs/pictures that accompanied that post? I'm not seeing any (apart from the one in the spoiler).
No, i didn't make any, sorry.
I was working with mostly approximate data - Fraps only records FPS once per second, which is not nearly enough and i must have ran through that part of the replay like 50 times. I didn't want to count all the frametimes etc and i didn't record cpu/gpu info because i did not have tools on hand to do it, just went by eye, dual screen and fraps benchmarks for verification
![[image loading]](http://i.imgur.com/lrpm23Q.png)
This is the typical CPU usage that i was talking about though, other cores maybe go up a little later in the game, but this is basically what you are looking at for sc2.
+ Show Spoiler +On March 25 2013 11:54 waffling1 wrote:I've DECIDED! Thank you everyone for putting up with me! I went with the cheaper mobo and PSU. I figure it will be good enough. OC to 4.2GHz should be fine enough. I'm not too certain on the limitations of the MSI G41, but here it goes! Microcenter $190 i5-3570k $90 MSI Z77A-G41+$23 tax -$40 bundle discount $0 shipping (some gas money) Newegg $40 COOLER MASTER Hyper 212 Plus-$20 rebate $30 Corsair CX 430-$10 rebate +$5 tax +$0 shipping -$10 new customer email discount +3$ adaptor cable for CX430's lack of a second PCIe connector. = $301 !!!+ Show Spoiler +with a case in the future
(Newegg or Amazon)
$85 Fractal Define R4 Windowed
+$7 tax
+$0 shipping
= $393
2nd Q - I thought you had an Ivybridge?
This is my main system til i (most likely) go Haswell, the ivy is in another poorly thrown together bluray-ripping etc rig.
|
Thanks guys. That was a fun 2 days of learning about comp builds. At first all your posts didn't register fully because it was unfamliar material. I think i repeated some questions, but thank you all for explaining things in depth a couple of times. I reread everything and it made a lot more sense now.
You've successfully converted one noob.
|
Hopefuly I won't get buried again  I appreciate all opinions:
Any other opinions on a good motherboard for OCing a 3570k? I am replacing my Z77 Pro3 and am wondering if I should get another mobo instead (I can buy a new mobo with my replacement order and trhen sell my replacement Pro3 as new to make up for all of/most of the cost of anoher new mobo).
Here is the mobo someone recommended. http://www.hardwareversand.de/DDR3/57526/Gigabyte Z77-D3H, Intel Z77, ATX.article
Thanks.
|
The CX430 for $20 deal at Newegg turned into $35 today. sadface.
damnit, the hyper 212 plus lost its $10 rebate on newegg.
and the capstone 450 went up by $5.
what is this, procrastinated inflation day at newegg?
|
|
|
On March 25 2013 05:48 Myrmidon wrote:
I think it's probably something we all knew already, but it's good to have data and a reminder. For those who want to play on higher graphics settings, we should maybe emphasize to turn down physics and effects. They're just eating up CPU cycles and don't really improve visibility (sometimes distract from it).
Some people really want to crank out that "max" setting though.
Is Nvidia's PhysX marketing just marketing? Or do the Physx enabled GPUs actually take control of the physics calculations? Because I know in my settings I can set it to do that, and maybe the 260 just wasn't able to do that, but what does that mean for a supposedly PhysX enabled GPU?
|
PhysX is a middleware physics engine SDK owned by Nvidia. A game's physics calculations need to use this specifically, so calculations can be offloaded to the (Nvidia) GPU, which support these instructions.
SC2 physics calculations aren't PhysX-based, AFAIK.
For the games that actually use PhysX (less common these days), yeah, the hardware acceleration is legit if maybe not that big of a deal.
|
Hey,
I just had a question.
I wanted to build a fairly compact rig (micro atx) with
3750k (but i am not sure the cooling will be efficient enough to OC so i might go for 3750, sparing a few bucks is always nice)
GTX650 TI Amp! From Zotac, which is a small factor graphic card with low consumption (and heat) so it's perfect for mATX case
http://www.techpowerup.com/reviews/Zotac/GeForce_GTX_650_Ti_Amp_Edition/1.html
I wondered it the whole thing was more or less balanced ? If the graphic card would not hinder the cpu perf in gaming ?
Thing is, i wanted to go for a 660 first, but the small factor and lower price convinced me. I could wait for small factor graphic card in the future.
So if you are aware of a small factor GTX660, then i'd be delighted to have a look at it.
What do you think of this ?
|
On March 24 2013 17:22 Myrmidon wrote:They can pop and go out, but usually they just bulge and leak instead, get worse over time. They have those slits / vents in the top so they usually fail in the non-catastrophic way rather than exploding. When a cheap power supply is loaded past what it's capable of and dies with some kind of pop, that's usually some switching transistor or other semiconductor part rupturing. For a better power supply, I'm not sure what the usual modes of failure are. It can be a range of things. But anyway, if a short develops somewhere, or maybe an open circuit in the wrong place, a lot of current could be going through certain parts that aren't supposed to be handling that much, and that could cause a transistor to heat up rapidly and explode. Some bad solder joints work for a long time, but after many many heat / cold cycles can have issues, or something like that. I wouldn't really think there would be an arc unless there's some other problem first. Show nested quote +On March 24 2013 17:04 llIH wrote:On March 22 2013 18:22 llIH wrote:
Hello
I am thinking about getting a laptop that I can carry around when I travel. But I want it to be able to play sc2-HotS on low settings and be able to watch streams at 720p.
How much approximately would I have to pay to get such a laptop? I do not care about looks.
What CPU and GPU would be good enough for low settings for HotS? (Low settings without lag)
I was thinking 15". But any options that are reasonable just throw them at me.
NB: I'm on a budget. Not specifically. But the lower the better. Someone have time to help me? Kind of vastly overkill on the GPU side but a good deal at $550: link, use coupon code 2Q?XNXR2DXQ13G Dell Inspiron 15R Special Edition Core i5-3210M Radeon HD 7730M (way underclocked desktop HD 7750) I'd avoid something with under an Ivy Bridge Core i5. Gotta have that CPU speed. That's not much over the laptops using Core i5 with integrated graphics. Especially if you're not maxing out the HD 7730M, there should be some turbo boost on the i5 there, especially since that laptop can be configured with an i7. Or I guess cheapest possible, Acer, etc., $450: http://www.newegg.com/Product/Product.aspx?Item=N82E16834215992
So for laptops an I5 3210M is the cpu recommended for SC2?
|
@anatase: If you want to overclock, mATX does not have to mean you can't. You have to look at exactly the same things as with ATX. The case has to be wide enough to allow a CPU cooler with 16 cm height. There are mATX cases like that. That also means those cases can take a 120 mm rear fan, so you can still have good ventilation while not having any more noise than without overclocking. It's not really any different than with ATX.
The Zotac GTX 660 is short compared to the other, normal GTX 660s. I think it's only a little bit longer than that 650 Ti you linked.
EDIT: Palit and Gainward also have short GTX 660s, and I think I've seen Gainward recommended for building silent PCs.
EDIT2: You may have to be careful about what kind of CPU cooler is the largest possible for mATX. My CPU cooler is about the size of the usually recommended Coolermaster Hyper 212 Evo, is also using a 120 mm fan, etc., and it barely fits. With mATX motherboards, the graphics card is in the very first slot, while ATX usually has a pci-e 1x slot as the first, and graphics card as the second slot. Something with a 140 mm fan would not fit in mATX, I think.
|
On March 26 2013 23:05 Ropid wrote:
@anatase: If you want to overclock, mATX does not have to mean you can't. You have to look at exactly the same things as with ATX. The case has to be wide enough to allow a CPU cooler with 16 cm height. There are mATX cases like that. That also means those cases can take a 120 mm rear fan, so you can still have good ventilation while not having any more noise than without overclocking. It's not really any different than with ATX.
The Zotac GTX 660 is short compared to the other, normal GTX 660s. I think it's only a little bit longer than that 650 Ti you linked.
EDIT: Palit and Gainward also have short GTX 660s, and I think I've seen Gainward recommended for building silent PCs.
From what i saw the 660 is 4cm longer than the 650, it can mean quite a lot in a compact rig but it turns out the case i picked should be large enough to go up to a 21cm long card so i will probably go for the 660.
Indeed Gainward looked like it had several -kind of- small card factor, will still dig but that's should be it.
THX
ps: btw since you're german, what online shop do you go to in germany ? I'd like to compare prices, the only deutsch shop I know is computeruniverse
|
For the German market, go to geizhals.at / geizhals.de.
|
First time computer build. Read a lot through this forum as well as several others and think I have a general idea of the components I want. I was looking to make a budget machine for around $600. It's really difficult to adhere to that limit being there is always a better choice for just a little bit more money. I am currently $150 over budget. I'm looking for feedback on the parts I've picked out as well as any ideas for cutting cost.
Processor:
Intel Core i5-3570K Ivy Bridge 3.4GHz (3.8GHz Turbo) LGA 1155 77W Quad-Core Desktop Processor Intel HD Graphics 4000 BX80637I53570K
$219.99
Motherboard:
ASRock Z77 Pro4 LGA 1155 Intel Z77 HDMI SATA 6Gb/s USB 3.0 ATX Intel Motherboard
$104.99
GPU:
ASUS GeForce GTX 660 GTX660-DC2O-2GD5 Video Card
$219.99
Hard drive:
http://us.ncix.com/products/?sku=58746&vpn=ST31000524AS&manufacture=Seagate&promoid=1293
Seagate Barracuda ST31000524AS 7200.12 1TB SATA 32MB Cache 3.5in Internal Hard Drive OEM
$70.00
RAM:
http://us.ncix.com/products/?sku=57953&vpn=F3-12800CL9D-8GBXL&manufacture=G.Skill&promoid=1293
G.SKILL Ripjaws X F3-12800CL9D-8GBXL 8GB 2X4GB DDR3-1600 CL9-9-9-24 Memory
$55.99
Power Supply:
CORSAIR Builder Series CX430 430W ATX12V v2.3 80 PLUS BRONZE Certified Active PFC Power Supply
$44.99
Case:
http://www.newegg.com/Product/Product.aspx?Item=N82E16811119227
COOLER MASTER Elite 430 RC-430-KWN1 Black Steel / Plastic Computer Case
39.99
Subtotal: $755.97
Thanks for any help.
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/KBAuRCF.png)
