EPTComputer Build, Upgrade & Buying Resource Thread - Page 484
| Forum Index > Tech Support |
When using this resource, please read the 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. | ||
|
EdVonSchleck
11 Posts
| ||
|
Avaroth
Finland44 Posts
What is your budget? Roughly 550e/600usd. I can spend more if it's highly recommended/just makes sense to put in a bit more for a significant improvement. Still, I'm looking to put together something cheap as I pretty much only play SC2/LoL. I considered going all out with a high-end machine but with new gen stuff apparently coming later this year it feels silly to do so. Thus I figured I'd go with a budget build and sell that off later and build next gen if I actually want to play new titles with high settings (which is something I've found myself not to do in the past). What is your monitor's native resolution? 1920x1080. What games do you intend to play on this computer? What settings? SC2 mostly, as well as LoL. SC2 I only play 1v1, rarely 2v2. I want to play GTA V as well, medium/high settings or so are fine. My steam library has a bunch of games I never got around to playing, mostly 2-5 year old titles. I will probably play some of those as well. What do you intend to use the computer for besides gaming? Perhaps some game dev stuff, web surfing obviously and watching movies. Do you intend to overclock? Yes. Do you intend to do SLI / Crossfire? No. Do you need an operating system? No. Do you need a monitor or any other peripherals and is this part of your budget? I'm buying a monitor as well, most likely the ASUS VX229N-W. If you have any requirements or brand preferences, please specify. Not really. What country will you be buying your parts in? China. If you have any retailer preferences, please specify. I'm buying everything off Taobao (Chinese ebay/amazon). Some older parts can be quite cheap while some brands are harder to come by. Alright so here's what I've come up with (I converted the Chinese prices to EUR): CPU: Intel G3258 Dual Core 61,04 € GPU: XFX 7970 3GB Double Dissipation Edition 114,93 € Mobo: MSI H81M-P33 37,76 € SSD: Kingston SV300S37A/240G SSD 78,21 € RAM : 2x4GB Kingston ValueRam 1600MHz 1.5v 24,78 €x2 Case: CoolerMaster K282C 34,18 € PSU: Super Flower 550W 550P14XE 50,60 € Monitor: ASUS VX229N-W 22 inch IPS 128,21 € Total 554,48 € Based on some googling the CPU can be overclocked to around 4,4GHz with the stock cooler using the MSI motherboard. Am I stupid getting this GPU with this CPU? Something like a Sapphire HD7870 2GB would be like 30€ cheaper but I feel like the 7970 is worth it at that price. The older HD 7xxx models seem to be somewhat cheaper on the Chinese ebay. I'd appreciate any feedback/suggestions very much! | ||
|
EdVonSchleck
11 Posts
On April 28 2015 05:25 Myrmidon wrote: If you're willing to spend up to the Core i5-4590k level, get an aftermarket CPU cooler for lower noise, and go for a motherboard that isn't among the cheapest available, you may as well just get a Z97 motherboard model so you have the (official) option to overclock if you want. H97 motherboards artificially don't let you overclock (at least not officially) and many don't really have the hardware or feature support for it anyway. The Enermax cooler is not very good for the price, though. For just a little more you can get a Thermalrigth HR-02 Macho; better options that are cheaper are also available. Double check height and clearance with the case once you decide on a case. The WD Purple hard drives are optimized for a very specific type of workload and probably aren't as good for consumer use. I think the Red (which has some features useful for NAS but is fine for general storage usage) or Green (general storage) models may be quieter too, though maybe not. In any case, the Purple is an unconventional and maybe inappropriate choice. That's not really a good price on a Corsair 200R and you probably want a case with less direct holes and paths for noise to escape. Even if you don't spend up to say the Fractal Design Define level, there are some alternatives. Actually, if you're not interested in an optical drive, the very new Define S is a bit cheaper. From Corsair, the cheapest reasonable option may be the 330R. The power supply is somewhere you can save money because you can find actively cooled options that are sufficiently quiet enough for the power draws you're looking at. Once you hit the CPU and GPU hard enough, their fans (or even a hard drive spinning) may make more noise than the quieter actively cooled power supplies, sometimes by a lot. That said, the 970 Strix is a lot quieter than many other cards. You could use the Cooler Master VS series or if you want to skimp, maybe one of the lower-end Be Quiet models. The CoolerMaster V450SM PSU wasn't in pcpartpicker's list, but I found it for 70€ from a german store, sound good? That would take me to ~1315€ for the build (without optical drive) at current prices. I also found some other retailers that might have better prices on the parts. Here's the updated build: http://de.pcpartpicker.com/p/L8N8wP And thanks again Myrmidon! | ||
|
Cyro
United Kingdom20297 Posts
| ||
|
EdVonSchleck
11 Posts
On April 28 2015 20:07 Cyro wrote: If you're buying from Germany, the superflower golden green hx450 should be around 70 euro shipped (i think) and is probably better That is good info, thanks. Do you think the power supply will be sufficient then? I'm assuming I won't be overclocking, or not much at least. | ||
|
Cyro
United Kingdom20297 Posts
On April 28 2015 20:43 EdVonSchleck wrote: That is good info, thanks. Do you think the power supply will be sufficient then? I'm assuming I won't be overclocking, or not much at least. Yea other stuff: he meant hr-02 macho (140mm version), the 120 is much worse ga-z87x-gaming 5 is already quite overkill on mobo, no need for the 7 unless there is some feature that you want a lot of better RAM available for the same or even cheaper price - http://de.pcpartpicker.com/parts/memory/#sort=a10&page=1&s=302133,302400&Z=8192002 | ||
|
EdVonSchleck
11 Posts
On April 28 2015 22:29 Cyro wrote: Yea other stuff: he meant hr-02 macho (140mm version), the 120 is much worse ga-z87x-gaming 5 is already quite overkill on mobo, no need for the 7 unless there is some feature that you want a lot of better RAM available for the same or even cheaper price - http://de.pcpartpicker.com/parts/memory/#sort=a10&page=1&s=302133,302400&Z=8192002 Alright, so here's the modfied build with price quotes from mindfactory (incl. taxes): Intel Core i5 4690K 4x 3.50GHz So.1150 BOX €245,53 Asus Z97-A Intel Z97 So.1150 Dual Channel DDR3 ATX Retail €143,68 Thermalright HR-02 Macho Rev.A (BW) Tower €37,89 4096MB Asus GeForce GTX 970 STRIX OC Edition Aktiv PCIe 3.0 x16 (Retail) €359,01 8GB G.Skill Ares DDR3-2133 DIMM CL9 Dual Kit €81,03 512GB Crucial MX100 2.5" (6.4cm) SATA 6Gb/s MLC (CT512MX100SSD1) €192,89 2000GB WD Green WD20EZRX 64MB 3.5" (8.9cm) SATA 6Gb/s €79,79 450 Watt CoolerMaster VS Series Modular 80+ Gold €69,79 Fractal Define R5 Midi Tower €99,69 Sums up to €1309,30, shipping would be extra. What do you guys think? Edit: If you're buying from Germany, the superflower golden green hx450 should be around 70 euro shipped (i think) and is probably better Yeah they didn't have that one at mindfactory, I could get it elsewhere for around €70. How much better are we talking here? Because I'd kind of like to have modular cabling, and the super flower one doesn't have that. If it's way superior to the CoolerMaster one though, I'd get it instead. | ||
|
mantequilla
Turkey779 Posts
Antec VPF450 450 watt Couldn't find a review about this one, but read antec psu are generally ok(?). Xigmatek NRP-PC402 400W This one has a review, mostly good, little bad. I'm building a low power i3+gtx750 (maybe 960?) PC. | ||
|
Incognoto
France10239 Posts
This is a good thing for people who want to use quiet fans which don't have a lot of pressure I guess. Very nice. Seems well priced as well. Very nice : http://www.vortez.net/articles_file/30166_fractal design define s review - install complete.jpg | ||
|
Myrmidon
United States9452 Posts
On April 28 2015 17:00 Avaroth wrote: I can't really tell what pricing should be secondhand on taobao, but that looks okay. I think 4.4 GHz may only be possible if you get lucky enough with the chip you get. It's also a bit dangerous to be abusing a board that cheap not really designed for overclocking. If you decide to get an aftermarket cooler, I would get a downblowing model to ensure more airflow to the VRMs. I hope you don't care about noise with the stock cooler if you run that and want to push the CPU. Some games already have issues loading on dual cores and many do appreciably worse on dual cores, so even a fast dual core is something of a handicap. On April 28 2015 20:43 EdVonSchleck wrote: That is good info, thanks. Do you think the power supply will be sufficient then? I'm assuming I won't be overclocking, or not much at least. I specifically mentioned the Cooler Master VS because I know at least the 550W model had an idle fan speed in the range of 600 rpm and a more relaxed fan curve up to about halfway. Overall the quality is about the same as the Golden Green HX. On April 29 2015 02:46 mantequilla wrote: Hello guys, guess what, I'm asking about psu again. Antec VPF450 450 watt Couldn't find a review about this one, but read antec psu are generally ok(?). Xigmatek NRP-PC402 400W This one has a review, mostly good, little bad. I'm building a low power i3+gtx750 (maybe 960?) PC. The cheapest Antec power supplies are not that good. No reviews yet, but the new VPF I would think should be okay and better than some of the older Basiq and VP models. But even some of those models I would take above that model Xigmatek. | ||
|
bluegarfield
Singapore1128 Posts
Just wanna ask if anyone experience have some good guide to recommend? Also software? And anything important that a novice likely to miss? Thanks lots. Oh also, normally how long would one need to properly overclock and do all the stability test? | ||
|
Craton
United States17250 Posts
If your OC isn't stable, generally speaking you'll either freeze, BSOD, or hit a too-high temperature relatively quickly when running the stress test. When that happens, you go back to your UEFI and tweak the settings to give it a little more juice (or lower it, in the case of temperatures being too high). I would say you can do moderate jumps when you start out the overclock (say, 2 steps of the clock multiplier at a time) since these chips usually OC pretty well. As a newbie you can also just choose to take it slow and do one step at a time and do your due diligence at each step. Each time you "up" the OC, record your settings. If you fail to get the OC any better, you'll want to know your last stable state to revert to (that'll be where you stop OCing). With each step up, run e.g. Prime95 for awhile (20-30 minutes is a good quick test). If everything checks out, up the OC another step. Once you think you've reached your best acceptable OC, run e.g. Prime95 for an extended period. A few hours is usually sufficient, but some people run it for 24 hours. I haven't OCed a Haswell; I only know the specifics for Ivy Bridge. Voltage increases generally cause the greatest leaps in temperature (I think it's an exponential increase) and you don't want to raise it too far regardless. The "steps" for voltage increases are around 0.05v at a time. Really, your best bet is just to take things slow. There's not much risk of damaging the chip if you're just doing single step increases of the multiplier and then tweaking the other settings little by little. The CPU should have auto-shutoff capabilities for temperature getting too hot, but it's still prudent to monitor the temperatures as you get into the higher range of OC. | ||
|
bluegarfield
Singapore1128 Posts
But is Prime95 test necessary? I mean it it's way beyond any realistic heavy usage, and I read that the test may push the voltage up as well. Also I am kinda confused because different people recommend different benchmark/testing software, like Prime95, Linpack, cinebench, etc, | ||
|
bluegarfield
Singapore1128 Posts
| ||
|
skyR
Canada13817 Posts
| ||
|
Craton
United States17250 Posts
On April 30 2015 15:03 bluegarfield wrote: Thanks Craton. But is Prime95 test necessary? I mean it it's way beyond any realistic heavy usage, and I read that the test may push the voltage up as well. Also I am kinda confused because different people recommend different benchmark/testing software, like Prime95, Linpack, cinebench, etc, Well that's the thing - they all more or less serve as stress tests. The idea is simply to ensure your system is stable and operating within acceptable ranges (e.g. temperature). Generally speaking they push your system harder than a real-world scenario, but if it's stable at that stressed level then it will also be stable at the real-world heavy load. You can use any of the programs or all of them if you so choose. They're just tools. I doubt there's any real consensus on one being the "best." Regardless of what you use, you do want to use *something*. You need a way to stress the system and make sure it actually runs as desired. Otherwise, down the road you may end up crashing or BSODing out of the blue. Systems can work fine at low load (e.g. browsing) but be unstable under load. That's what you're trying to suss out. | ||
|
bluegarfield
Singapore1128 Posts
On May 01 2015 02:30 skyR wrote: 5th generation is Broadwell so you don't need to flash it. Thanks, I went ahead and tried. On May 01 2015 06:03 Craton wrote: Well that's the thing - they all more or less serve as stress tests. The idea is simply to ensure your system is stable and operating within acceptable ranges (e.g. temperature). Generally speaking they push your system harder than a real-world scenario, but if it's stable at that stressed level then it will also be stable at the real-world heavy load. You can use any of the programs or all of them if you so choose. They're just tools. I doubt there's any real consensus on one being the "best." Regardless of what you use, you do want to use *something*. You need a way to stress the system and make sure it actually runs as desired. Otherwise, down the road you may end up crashing or BSODing out of the blue. Systems can work fine at low load (e.g. browsing) but be unstable under load. That's what you're trying to suss out. Thank you much. I settled with Prime95 27.9, x264 and Cinebench for stress testing. Will try to get some result tomorrow. Anyway, howlysheet I hit a brick wall so damn hard. For some reason Gigabyte (Gaming 7) BIOS UI is kinda different from most of the overclocking guides, cant find Adaptive/Offset voltage thingy to turn off so I can run those synthetics tests (edit: well I did see that Adaptive is not present on Gigabyte board, but then what to do with those C6/C7 states). I only found C3, C6/C7 states, I guess I will disable those then OC. And then I got locked out of the system because someone changed the windows password huehuahue T_T And Gigabyte smart tweak UI sucks, it keeps jumping around every time I click something, making me wonder if I accidentally change some wrong values | ||
|
Cyro
United Kingdom20297 Posts
Prime95 is what most people use to check stability Not for any CPU released after 2012 TBH The synthetic tests that stress FPU on haswell (small fft, linpack etc) get the CPU ~30-40c hotter than the tests that don't heavily stress the FPU. That means you can use one OC profile only during testing, and get vastly different temperature results. You could have a video encoder, for example - and run it 100% load on all cores and it'd peak 60c. You could run fft 1344-1792 on latest prime and it would peak ~65-70c. you could run an FPU stress test and it would hit 95-100c and throttle. ^Out of all of those, the only test that could crash would be the large FFT - because voltage requirements are not necessarily the highest on the load that draws the most power (creates more heat). It's 20-30c cooler than the hotter tests, but needs the most Vcore and crashes unstable OC's easier - therefore the only real purpose for using it is to simulate worst case scenario temps when using crazy FPU stress, and the reality is there's no reason to do that when no real world program that i've found in two years comes within ~30c of it (and those 30c matter a lot for tweaking OC) Prime95 and some other programs were popular in the past when they heavily stressed the CPU and temperatures were hot - but they were not THAT much hotter than regular loads. The introduction of AVX instructions changed that a bit - the temperature difference between worst case FPU synthetics and almost all other loads increased. Haswell added avx2 and fma3 - which made that gap huge compared to every other CPU. It's very very fast when dealing with the type of load that prime95 small fft or linpack presents to it - it's capable of >200gflops at 4ghz, while avx1 CPU's can only do 110 or so - but it consumes a lot of power and gets hot when it uses those new instruction sets. The only guide that you should use is this: http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics/ You should enable all c-states @ OC. I use only c7 personally, no other c-states or EIST because it's very latency-friendly to only use c7 and still has idle low voltage, but you probably can't see any practical difference aside from niche cases to prevent you from using all c-states + EIST. You also shouldn't use motherboard software to change voltages in OS generally. Just change stuff in bios. You should definitely use the x264 test IMO (mega link in that guide). Realbench is a bit easier to set up (doesn't need a short list of instructions) and also has an x264 test - though very slightly worse for stability testing probably it's functionally pretty much the same - and some other tests in there. You can ballpark some values with 5-60 minute runs of x264 test, then you can run an overnight test with some kind of logging (and bluescreenview to check the timestamp of bluescreen if it crashed that way). It will either pass or you'll see how long it took to fail. ----- Basic settings that you should set before starting OC: Uncore to 33x Uncore voltage to 1.15v VRIN LLC to turbo Variables during OC: Core clock speed core voltage (vcore) VRIN (input voltage) - NOT ring/uncore. "VRIN" looks close to "VRING" which is potentially fatal if you mix them up, don't  gigabyte calls it VRIN/input and uncore so you probably won't mix it up, but some boards call the uncore "ring" so that's dangerous to refer to on some guides/comments.----- you should adjust core clock speed and tweak vcore and input voltage to be stable. You need enough vcore to be stable (if you don't have enough, no amount of tweaking input voltage will help you) but you can also have plenty of vcore and be unstable if the input voltage is out of range. For that you use that turbo LLC for input voltage, and you adjust it some with vcore. A setting 0.6 above vcore is probably good. Vcore is automatically 0.02 above bios value for some loads, so a profile that's great for starting out would be 1.2vcore and 1.85 input voltage. You can set that and then the only thing that you'll have to change is Core clock speed - it'll be stable or it won't. For example 4.0ghz is pretty much guaranteed to work. 4.3ghz will likely work, but might not. You can find what works there, and then work from there adjusting those three settings and testing. Generally ~0.55 - 0.65 gap between vcore and input voltage is good. Some people with haswell cpu's before the 4690k/4790k needed to use more in some cases (like 2.1 for 1.4vcore) but 4690k/4790k seem to be fine with lower numbers (lower isn't necessarily better though, there's no reason to try to lower it too far) After you're 100% sure that your settings are stable and you know how to play around without risking crashing due to core clock, vcore and input voltage (it's good to play your games, do your cpu loads etc for a few days to a week beforehand, or add a bit of vcore/input for this) you can raise the uncore clock and voltage. It's negligable for performance in almost every task, but it can be nice to run it at whatever runs fine with uncore voltage 0.1 below vcore - for example, 4.6ghz core @1.3vcore, 4.2ghz uncore @1.2 uncore V. Sorry for late reply, i was up for a long time and not feeling good then slept for ages so i didn't spend much time on forums for a while | ||
|
bluegarfield
Singapore1128 Posts
Anyway, I got some decent start until the first crash, but I only did 15-20min x264 stress test so far. 4.6Ghz@1.2V bsod immediately when x264 starts, and 4.5Ghz@1.2V was fine with x264 (only run 2 loops 20mins though), fine also on Prime95 small FFT but froze after 5mins with Prime95 blend test. I am trying to adjust the voltage now and hopefully can get a stable 4.6Ghz. Dont have the time for 24hrs stress test now so probably will run a 4hrs x264 and may be 6hrs Prime95 blend test (or may be not, it looks pretty destructive XD). Temperature-wise, ~60*C when running x264 and ~73*C running Prime95 | ||
|
Cyro
United Kingdom20297 Posts
On May 01 2015 14:04 bluegarfield wrote: Thank you so much Cyro, that explains a lot of things for me. Take care also, sleep is good Anyway, I got some decent start until the first crash, but I only did 15-20min x264 stress test so far. 4.6Ghz@1.2V bsod immediately when x264 starts, and 4.5Ghz@1.2V was fine with x264 (only run 2 loops 20mins though), fine also on Prime95 small FFT but froze after 5mins with Prime95 blend test. I am trying to adjust the voltage now and hopefully can get a stable 4.6Ghz. Dont have the time for 24hrs stress test now so probably will run a 4hrs x264 and may be 6hrs Prime95 blend test (or may be not, it looks pretty destructive XD). Temperature-wise, ~60*C when running x264 and ~73*C running Prime95 You should just leave prime entirely IMO, or maybe use it for nothing but version 28.5 large FFT (manual 1344-1344 or 1792-1792 size is good for a consistent test) no need for 24 hour test either. You can just start running a test when you go to bed and stop it when you get up - you could run 45x like i said and in the morning it will have either crashed (then you can see how long it took to hit that random fatal error - if it took longer, it's more likely that you were closer to complete stability) or not crashed (very close to solid, you can just add something like 0.02vcore for safety at this stage on a typical OC) In your case i would just try to make ~46x@ ~1.25vcore with 1.95 VRIN pass ~30 min, then 4.7ghz around 1.32v if that went ok. When you get to temps breaking 85 on x264 in peaks, that's towards the hotter side for an OC. | ||
| ||
