|
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. |
What is your budget?
$1000 CDN
What is your resolution?
1920x1080
What are you using it for?
Playing SC2 and Diablo 3 and general simple computer stuff
What is your upgrade cycle?
3+ years
When do you plan on building it?
ASAP
Do you plan on overclocking?
Yes
Do you need an Operating System?
Yes
Do you plan to add a second GPU for SLI or Crossfire?
No
Where are you buying your parts from?
NCIX
|
On March 22 2013 18:48 Myrmidon wrote:+ Show Spoiler [VRM power/temp, not too important] +On March 22 2013 18:13 Belial88 wrote: I don't really agree with this. I don't know that much about what exactly is going on in there, but your VRM gets hot based on the voltage that flows through it, not the power requirement of the chip (in my experience, but given my understanding of how the VRM works, it makes logical sense that it's not about power requested but voltage supplied as that's what the VRM does). So 1.4v is the same on the VRM regardless if it's AMD or Intel. Just seems to be my experience that doesn't matter how cool or low power the chip is, high voltage gets hot. MOSFET conduction losses while switched on are I^2 * R_DS(on), where I is the current (to the processor, which is going through the MOSFET) and R_DS(on) is the effective resistance between drain and source terminals when switched on. For most of a cycle for a VRM phase, the low-side MOSFET is conducting. More power draw means higher load current I, so more power losses for a given MOSFET. High-side MOSFET also sees conduction loss of I^2 * R_DS(on), but switching loss is I think more relevant for that because it's not on very much when converting 12V to something low like 1.25V. Switching losses are proportional to I, also increase with higher load. If you're willing to accept even more hand-waving, people talk about the efficiency of the conversion process (which in reality doesn't scale linearly with load power consumption). Use more power, and if the efficiency didn't change by a lot, then that means more losses in the power conversion. e.g. 90% efficiency delivering 60W means 6.67W lost as heat in the VRMs. 90% efficiency delivering 100W means 11.11W lost as heat in the VRMs. Higher voltage means more current, as the processor is consuming more power. Anyway, try setting a higher voltage, keep CPU at idle, check VRM temps. Then set lower voltage, load up CPU, check VRM temps.
I see. Well, intel chips will still stress a VRM quite a bit, and you can push over 150w on air with the very high range of voltages. Lower end boards will still definitely get hot enough to limit you, my z77a-g41 was overheating just on 4.4ghz, it even had choke whine.
That's very daft of you. One of the main reasons that people go liquid cooling (custom loops) is noise while overclocked. With a half decent setup you can have a silent computer regardless of the load. Also another big sell is that you'll be stable at lower voltage at any given overclock.
CLC is just terrible though.
Liquid cooling is going to be just as loud as air cooling, because of the fans. If you want to get into the sound difference of air flowing through, say, twin towers vs a radiator, that's very very minimal, less than 5-10% of the total noise. Now if you mean 'liquid cooilng can be just as cool as X air cooler for Y fan speed", yes, but that's less to do with liquid vs air and more to do with simply having a bigger block of metal and better heat transfer.
Closed loops are hardly 'terrible', they are just the same as air heatsinks, with their own benefits and costs. Generally, closed loops are more expensive than air coolers of the same performance, which may mean they are 'terrible' per cost, yes, but at the right price, they are fine. If you got a closed loop at a cheaper price than an air cooler of same performance, it'll be the better buy. It has nothing to do with closed loop vs air vs custom, but how much cooling performance you get per dollar, and how much cooling you ultimately need.
So an H50, which tends to sell much cheaper than Megahalems, is a 'better' buy. If megahalems were cheaper, than they'd be better to get than an h50. Right now on newegg, the Zalman lq320 out performs any air heatsink and is $39, which is just a steal. If it was at it's normal price of ~$90, it'd be a rip-off. it has to do with price.
I already mentioned that custom loops are fine if it's for an 'illogical' reason like aesthetics. I wouldn't quite put low noise as the main reason people buy closed loops, if you need quiet cooling it would be absurd to buy custom loop if running a low overclock, as plenty of air heatsinks can do it fine without a fan at all, whereas custom loops NEED a fan. ie, hr-02 macho fanless vs custom loop on just 4.5ghz intel.
The only 'logical' reason (ie not aesthetics) to get custom water, would be if you simply needed the cooling, which basically means you are pushing extreme voltages like 1.5v+. You could maybe say you go with custom water if you dont want to delid your ivy bridge, but even without delidding you really gotta push fairly high voltages for the custom loop to be worthwhile (and it's kind of silly to spend more than the cost of simply killing an ivy bridge and buying a second one).
|
Custom loops being quieter is about cooling multiple graphics cards. If you're just cooling the CPU there's not that much motivation for that unless you have ridiculous power draws.
There's no way you can cool multiple high-end graphics cards at a mid noise level with the size, positioning, and power draw of those things. Blower designs are too noisy, and larger open-air GPU coolers just end up recirculating a lot of hot air. You can do better by moving the heat to huge radiators elsewhere and put quality 120mm or 140mm fans on them, which actually get to blow the hot air directly out (or in I guess) the case.
Though to be honest, I have no idea if that's what was meant. The discussion seemed to be about CPU cooling.
Is ethylene glycol that much worse than water? What percent do they use? At 50% and say 50C, specific heat capacity is like 83% of water, and it's a little denser than water which means there's more of it per volume. Even if you just focus on the liquid, the bigger deal for CLCs is the lack of reservoir and thin tubing meaning low volume of liquid, right?
On March 23 2013 10:32 Belial88 wrote:Show nested quote +On March 22 2013 18:48 Myrmidon wrote:+ Show Spoiler [VRM power/temp, not too important] +On March 22 2013 18:13 Belial88 wrote: I don't really agree with this. I don't know that much about what exactly is going on in there, but your VRM gets hot based on the voltage that flows through it, not the power requirement of the chip (in my experience, but given my understanding of how the VRM works, it makes logical sense that it's not about power requested but voltage supplied as that's what the VRM does). So 1.4v is the same on the VRM regardless if it's AMD or Intel. Just seems to be my experience that doesn't matter how cool or low power the chip is, high voltage gets hot. MOSFET conduction losses while switched on are I^2 * R_DS(on), where I is the current (to the processor, which is going through the MOSFET) and R_DS(on) is the effective resistance between drain and source terminals when switched on. For most of a cycle for a VRM phase, the low-side MOSFET is conducting. More power draw means higher load current I, so more power losses for a given MOSFET. High-side MOSFET also sees conduction loss of I^2 * R_DS(on), but switching loss is I think more relevant for that because it's not on very much when converting 12V to something low like 1.25V. Switching losses are proportional to I, also increase with higher load. If you're willing to accept even more hand-waving, people talk about the efficiency of the conversion process (which in reality doesn't scale linearly with load power consumption). Use more power, and if the efficiency didn't change by a lot, then that means more losses in the power conversion. e.g. 90% efficiency delivering 60W means 6.67W lost as heat in the VRMs. 90% efficiency delivering 100W means 11.11W lost as heat in the VRMs. Higher voltage means more current, as the processor is consuming more power. Anyway, try setting a higher voltage, keep CPU at idle, check VRM temps. Then set lower voltage, load up CPU, check VRM temps. I see. Well, intel chips will still stress a VRM quite a bit, and you can push over 150w on air with the very high range of voltages. Lower end boards will still definitely get hot enough to limit you, my z77a-g41 was overheating just on 4.4ghz, it even had choke whine. For 150W, are we talking SB-E, SB, IVB, or what? I'd think 150W would be really hard to pull off on air for IVB. This meaning CPU draw and not system draw.
Not so surprised about Z77A-G41, at least by visual inspection and typical price. I didn't check the ICs etc., but it seems like it could be the worst VRM section on any Z77 board.
|
Custom loops being quieter is about cooling multiple graphics cards. If you're just cooling the CPU there's not that much motivation for that unless you have ridiculous power draws.
There's no way you can cool multiple high-end graphics cards at a mid noise level with the size, positioning, and power draw of those things. Blower designs are too noisy, and larger open-air GPU coolers just end up recirculating a lot of hot air. You can do better by moving the heat to huge radiators elsewhere and put quality 120mm or 140mm fans on them, which actually get to blow the hot air directly out (or in I guess) the case.
good point.
Even if you just focus on the liquid, the bigger deal for CLCs is the lack of reservoir and thin tubing meaning low volume of liquid, right?
Right. That's why custom modded closed loops where they add a reservoir perform similarly to custom loops with similar radiators. The tubing on closed loops also tend to be thinner.
Coolermaster is coming out with a new kit where you buy just the cpu block/reservoir/pump all in one, so you just get some tubing and a radiator and you are good to go, which is really cool. You can add more blocks for other stuff, you can add another reservoir or pump, etc. CM were the first to do closed loops though.
For 150W, are we talking SB-E, SB, IVB, or what? I'd think 150W would be really hard to pull off on air for IVB. This meaning CPU draw and not system draw.
my i7-3770k. 150w isn't too hard to pull off for ib on air (granted, it's toward the limit). 170w is, not 150. cpu draw.
Not so surprised about Z77A-G41, at least by visual inspection and typical price. I didn't check the ICs etc., but it seems like it could be the worst VRM section on any Z77 board.
Right now the MSI Z77A-G41 is $34 on microcenter (i see it via email, but it's accessible to anyone, it's just not listed on the website). the g45 and g65 are also on sale but those are terrible boards that are too expensive, given that they don't have offset control. Even the msi mpower board is useless for 24/7 overclocks... (good for ln2 benching, supposedly lol).
But the VRM on the z77-ag41 isn't particularly special in it's low quality. The Extreme4 even uses the same type of mosfets (although extreme4 is special in how terrible it is). Many boards don't have low rds on, a single driver, etc.
|
Hello Team Liquid, I'm looking into building a desktop, but I don't really have the skills to put one together. I live in the wastes of rural Nevada, so there's no one about to whom I can give some money and have one put together. The next best option I seem to have is to resort to a service like cyberpower, although I'm not even really sure that that's the best choice. I've played around with cyberpower's customize tool for a while and come to the conclusion that I may not be computer literate enough to work out if I'm spending money smartly. So, I'd like to as you for help. Budget: $900 with some wiggle room Resolution: 1440x900. I'm upgrading from a laptop and need to buy a monitor (not included in the budget) and this seems to be the best resolution within my price range, and which will fit onto my physical, wooden desk. Use: Mostly gaming. Upgrade Cycle: 3-4 years. Time frame: 4-6 weeks. Overclock: I do not plan to overclock OS: I need an OS Extra GPU: I would like the option to add an extra GPU, but 1) I'm not sure I have the knowledge to make one work 2) it's not a huge priority. Parts: As I mentioned earlier, an online custom assembly shop seems like my best bet, so I'm unsure if this question is relevant. Fry's is not near enough to be an option. Thanks in advance for the help and/or advice.
|
On March 23 2013 11:17 Belial88 wrote:+ Show Spoiler +Custom loops being quieter is about cooling multiple graphics cards. If you're just cooling the CPU there's not that much motivation for that unless you have ridiculous power draws.
There's no way you can cool multiple high-end graphics cards at a mid noise level with the size, positioning, and power draw of those things. Blower designs are too noisy, and larger open-air GPU coolers just end up recirculating a lot of hot air. You can do better by moving the heat to huge radiators elsewhere and put quality 120mm or 140mm fans on them, which actually get to blow the hot air directly out (or in I guess) the case. good point. Even if you just focus on the liquid, the bigger deal for CLCs is the lack of reservoir and thin tubing meaning low volume of liquid, right? Right. That's why custom modded closed loops where they add a reservoir perform similarly to custom loops with similar radiators. The tubing on closed loops also tend to be thinner. Coolermaster is coming out with a new kit where you buy just the cpu block/reservoir/pump all in one, so you just get some tubing and a radiator and you are good to go, which is really cool. You can add more blocks for other stuff, you can add another reservoir or pump, etc. CM were the first to do closed loops though. For 150W, are we talking SB-E, SB, IVB, or what? I'd think 150W would be really hard to pull off on air for IVB. This meaning CPU draw and not system draw. my i7-3770k. 150w isn't too hard to pull off for ib on air (granted, it's toward the limit). 170w is, not 150. cpu draw. Not so surprised about Z77A-G41, at least by visual inspection and typical price. I didn't check the ICs etc., but it seems like it could be the worst VRM section on any Z77 board. Right now the MSI Z77A-G41 is $34 on microcenter (i see it via email, but it's accessible to anyone, it's just not listed on the website). the g45 and g65 are also on sale but those are terrible boards that are too expensive, given that they don't have offset control. Even the msi mpower board is useless for 24/7 overclocks... (good for ln2 benching, supposedly lol). But the VRM on the z77-ag41 isn't particularly special in it's low quality. The Extreme4 even uses the same type of mosfets (although extreme4 is special in how terrible it is). Many boards don't have low rds on, a single driver, etc. But Extreme4 has twice the number of MOSFETs and an actual heatsink**. There's also definitely differences between different models of MOSFETs that happen to be packaged in DPAK (TO-252), though like I said I didn't check the ICs so I would just be blind guessing.
** hey wait a sec, in your old model you explained (not ragging on you, but a shift in thinking is required), having more parts wouldn't reduce the heat produced by each one. In reality, having more means less current through each one, so less heat for each. Also, Extreme4 would have more capacitance and other better parts. And again, G41 doesn't even have heatsinks. That said, I wouldn't be surprised if you tried putting heatsinks on the G41 and didn't mention it.
How'd you measure 150W or 170W? If it really happened, I'm guessing it had to be over 1.4V, way past what's probably sustainable long term? Maybe I misworded that: by "really hard" I mean requiring higher-end cooling and non-wussy VRMs.
|
On March 23 2013 10:11 Zanzibar wrote:+ Show Spoiler +What is your budget?
$1000 CDN
What is your resolution?
1920x1080
What are you using it for?
Playing SC2 and Diablo 3 and general simple computer stuff
What is your upgrade cycle?
3+ years
When do you plan on building it?
ASAP
Do you plan on overclocking?
Yes
Do you need an Operating System?
Yes
Do you plan to add a second GPU for SLI or Crossfire?
No
Where are you buying your parts from?
NCIX Core i5-3570k - $240 http://ncix.ca/products/?sku=70541&promoid=1293 + Show Spoiler [price match $227] +http://www.newegg.ca/Product/Product.aspx?Item=N82E16819116504
Gigabyte Z77-HD3 - $110 http://ncix.ca/products/?sku=78647&promoid=1360
Zalman CNPS10X Optima - $30, -$10 MIR http://ncix.ca/products/?sku=67639
G.Skill 2 x 4GB DDR3 1600MHz - $57 http://ncix.ca/products/?sku=57953&promoid=1293
Powercolor HD 7770 - $120, -$30 MIR http://ncix.ca/products/?sku=74323&promoid=1293
Samsung 840 120GB SSD - $100 http://ncix.ca/products/?sku=77210&promoid=1293
Western Digital Caviar Blue 1TB (or whatever size is needed) - $79 http://ncix.ca/products/?sku=74462
XFX Core 450W - $52, -$10 MIR http://ncix.ca/products/?sku=63238
Antec GX700 - $50 http://ncix.ca/products/?sku=79112&promoid=1293 + Show Spoiler [alternative] +
Windows 8 (or 7, whatever) 64-bit OEM - $108 http://ncix.ca/products/?sku=77180 + Show Spoiler [price match $95] +http://www.tigerdirect.ca/applications/searchtools/item-details.asp?EdpNo=4864601
The HD 7770 is a little bit on the low side if you want to max out the games you mentioned, but they really don't need much. You could bump up to say HD 7850. Past GTX 660 is probably not worth it. If you wanted to overclock the CPU a lot, you'd want a better motherboard there too, probably. You could just skip the SSD to have more room in the budget, but I wouldn't recommend that personally.
On March 23 2013 11:35 aquelmerde wrote:+ Show Spoiler +Hello Team Liquid, I'm looking into building a desktop, but I don't really have the skills to put one together. I live in the wastes of rural Nevada, so there's no one about to whom I can give some money and have one put together. The next best option I seem to have is to resort to a service like cyberpower, although I'm not even really sure that that's the best choice. I've played around with cyberpower's customize tool for a while and come to the conclusion that I may not be computer literate enough to work out if I'm spending money smartly. So, I'd like to as you for help. Budget: $900 with some wiggle room Resolution: 1440x900. I'm upgrading from a laptop and need to buy a monitor (not included in the budget) and this seems to be the best resolution within my price range, and which will fit onto my physical, wooden desk. Use: Mostly gaming. Upgrade Cycle: 3-4 years. Time frame: 4-6 weeks. Overclock: I do not plan to overclock OS: I need an OS Extra GPU: I would like the option to add an extra GPU, but 1) I'm not sure I have the knowledge to make one work 2) it's not a huge priority. Parts: As I mentioned earlier, an online custom assembly shop seems like my best bet, so I'm unsure if this question is relevant. Fry's is not near enough to be an option. Thanks in advance for the help and/or advice. Come back when you buy, even if it's just to get a prebuilt, in case prices shift. But seriously, assembling is not much harder than Legos + referencing a manual once or twice. Almost everything is keyed to only fit the correct way.
Also, how much were you thinking for the monitor? I'd suggest getting a 1920x1080, at least spending $150. You could take a little bit out of the $900 budget for the box.
|
United Kingdom20275 Posts
How'd you measure 150W or 170W? If it really happened, I'm guessing it had to be over 1.4V, way past what's probably sustainable long term?
Belial's running 1.5v 5ghz 24/7. AFAIK it's sustainable on ivy bridge reasonably long term - it would be pretty much instant death for sandy, nehalem etc though. Ivy is extremely tolerant of voltage IF you can cool it - hence the delid and nh-d14 with i cant even remember how many fans.
|
You don't necessary need a custom loop to cool multiple GPUs at good temp/noise. There do exist CLCs that you can install aftermarket that do a pretty good job, but they are permanently affixed, each have a separate radiator needing to be mounted, and are difficult to install. Hose length to mounting point can definitely be an issue beyond the first.
http://www.newegg.com/Product/Product.aspx?Item=N82E16835186067
Obviously a custom loop would be better :p.
|
hmmm i can buy the rosewill capstone 450w for $44...
|
Almost nobody's owned Ivy Bridge longer than a year though, not long enough to really know. It's a lot of guesswork, but maybe my assumed sustainability requirements are more stringent than others.
Maybe the tri-gates or 22nm are different because of _____? Anybody know or had a plausible explanation? I'm loathe to just trust less than a year's worth of enthusiast self-reporting, anecdotal evidence, for the obvious reasons.
|
United Kingdom20275 Posts
edit: Oops. 1 year. Not 2.
Still, Sandy degrades really fast at 1.4v+, even heard dead cpu's slightly below that, but there's few/no reports of it happening with ivy bridge at those voltages aparantly. From my own reading it seems pretty solid to use even 1.5v 24/7.
|
On March 23 2013 02:13 MisterFred wrote: We did seem to get distracted by your request to help you decide how to choose for yourself. Here's what I'd recommend in terms of hardware:
No worries. I think i may have overemphasized the learning part. I think the best way to discuss is to have recommendations reasons, and alternatives with supporting reasons, as you foresaw.
CPU: i5-3570k ($190) in-store @ microcenter
I will go with this one - most likely from microcenter for their $40-off deal. Even with 8% tax on $300 (includes mobo price), it still nets positive savings.
Mobo: Asrock z77 Extreme4 ($95) in-store @ microcenter. Price above includes a $40 bundle deal for buying it with the CPU. If they don't give you the bundle deal right off, have a fit until they do. Mobo: Gigabyte ga-z77x-ud3h ($115) same comment as above. Change made as Cyro below has quality concerns about the Extreme4.
I think mobo is the only thing I need to pick now. The sales rep mentioned how GB motherboards have software locks that make it not friendly for overclocking. does the one you recommend not have that problem, or isn't a big deal? They had a lot of ASUS boards. are those any good? reminder: i'm only using 1 graphics card. I'm not sure what features I will be using or not using besides that. There's overclocking of course. Can you recommend any other options along with your excellent reasons/comparisons like before?
Total: ~$400-$420. If you want to spend another $100 or so you can get an i7-3770k, which will probably cut down on rendering time a little bit, maybe 10-20% compared with the i5-3570k? Your call if that's worth it.
that's exactly the kind of info and easy-to-adjust recommendations I was hoping for. Thanks
I did the final price for both, and the final comes to an 8 dollar difference of $76 to $68 (even less difference if i buy more stuff in one shipping from newegg). I'll go with the capstone.
Great. $20 is pretty cheap. Approximately how much difference in will there typically be compared to having the stock fan when under large usage load? if degrees are used, please be sure to specify C or F.
I'll stick with my RAM for now. If my RAM says 1600, is that the max, or can it go beyond that when overclocked? http://www.newegg.com/Product/Product.aspx?Item=N82E16820145260
Video Card: Probably want to stick with AMD Radeon cards here, as they seem to to a tad better in Solidworks for the price than their competition. Again, not a priority unless you run games OTHER than Dota 2 or SC2, which are not very demanding on GPU. This Sapphire 7870 ($220) would put you a bit over budget, but is the card I'd recommend: http://www.newegg.com/Product/Product.aspx?Item=N82E16814202025 I will be sticking with my 5830 GPU. it's serving me fine. But would my graphics be the bottleneck with this build for future upgrades? I had no idea about Radeon cards being a better value for Solidworks. You must read a ton.
Don't forget to thank your e-sports sponsors if you buy from them:
how exactly do i do this? lol.
P.S. For Windows 7: if you still have the installation files for the "upgrade copy" you downloaded, you might just be able to re-use them if you put them on a flash drive. Many "upgrade" versions of Windows are just the full version labeled differently. If what you downloaded is not in fact a full version, then you could just download an image of a Windows 7 installation CD & use windows own tool for putting it on a flash drive (google/windows website search). Note the important point here is whether or not the upgrade is a full license or an OEM license. Ordinary Windows licenses allow you to use the program on multiple generations of computers (not unlimted), so you're in the clear. However, OEM licenses (including some upgrades) are valid ONLY for the current computer (defined by hardware, generally motherboard). If that's the case, then the above downloading recommendation would be pirating & would be illegal, which makes for big frowny faces here at TL tech support.
ooo i'm not sure if i have the files. I checked the download link (which is years old), and it's not available. I got it from Digital Rivers, an authorized reseller. I recall that they allowing a few hardware switches, but I could be wrong. Is there any way to recover this without forking over another $100 just to transfer it over to my SSD? If not, don't rack your brain over it too much. I've contacted them, and maybe I can just live with having it on my hard drive. Ohhhh wait. The new mobo might trip the machine licence cock block anyway...
P.P.S. Technically you're a tax evader if you don't pay CA state taxes on your online purchases. In practice this means precisely jack. But hey, don't knock newegg too much, they charge you CA sales tax but don't make you a criminal!
lol, ok. Nice Newegg, good Newegg.
|
United Kingdom20275 Posts
Great. $20 is pretty cheap. How much difference in will there be with the stock fan when under large usage load? please specify C or F for degrees.
Like 30c, with less noise or something, stock fan is cheap and small - a combination that is not very good. Its basically night and day difference with any tower heatsink like the gaia (which is low end - but unthinkable performance gap)
Maybe not 100% accurate on that - but its not a small gap at all.
|
|
belial, did you mean F for every time you used C?
Cyro, you're saying there is a 30 C DIFFERENCE between the stock fan and an aftermarket fan, or that the machine runs AT 30 C with the aftermaket fan? the prior seems quite unlikely to me. I'm interested in the difference in temperatures.
|
United Kingdom20275 Posts
belial, did you mean F for every time you used C?
He did not
Cyro, you're saying there is a 30 C DIFFERENCE between the stock fan and an aftermarket fan, or that the machine runs AT 30 C with the aftermaket fan? the prior seems quite unlikely to me. I'm interested in the difference in temperatures.
Stock heatsink at non overclocked might max out at 85c under stress testing load - when i say for example 20c difference, the same load would put the other heatsink at 65c. Maybe 30 is a little bit over the top for a gaia - but it is NOT for other, better coolers - like the hr-02 macho for example, which im pretty sure can take like 40c margins over the stock cooler. Gaia is just cheap option to allow for some decent overclocking and/or keeping temperatures down and comfortable - because it's very easy to get a lot better than the stock cooler.
|
On March 23 2013 11:55 Cyro wrote: edit: Oops. 1 year. Not 2.
Still, Sandy degrades really fast at 1.4v+, even heard dead cpu's slightly below that, but there's few/no reports of it happening with ivy bridge at those voltages aparantly. From my own reading it seems pretty solid to use even 1.5v 24/7. Damage accumulates over time. I guess we don't really mean "dead" as much as "can't sustain same speeds at same settings anymore". Or at least, that's what I'm talking about. This stuff is hard enough to predict if you do reliability testing at Intel, much less for anybody on the outside. Seems plausible that a new process etc. could sustain higher voltages (though smaller feature sizes means for example less hot-carrier injection required for problems to occur), but I also wouldn't be surprised at all by some sadface in a year or two.
If the reports are all that favorable then maybe there really is something to the trigates or something like some secret technique invented and I could be way wrong.
On March 23 2013 11:51 Craton wrote:You don't necessary need a custom loop to cool multiple GPUs at good temp/noise. There do exist CLCs that you can install aftermarket that do a pretty good job, but they are permanently affixed, each have a separate radiator needing to be mounted, and are difficult to install. Hose length to mounting point can definitely be an issue beyond the first. http://www.newegg.com/Product/Product.aspx?Item=N82E16835186067Obviously a custom loop would be better :p. Quick, somebody find a build pic where somebody's using an H80 / H100 / whatever for the CPU up top, one of those for the top graphics card place on the rear exhaust, and one of those for the bottom graphics card placed on the bottom (floor) fan mount. That would be most pro.
edit: hardwaresecrets heatsink testing leaves a loooooooot to be desired, but: http://www.hardwaresecrets.com/article/Xigmatek-Gaia-CPU-Cooler-Review/1410/6
It really really depends on the test setup. Actually, they remove the case fans off and remove the side panel, which helps down-blowers, including the stock heatsink, by a lot relative to the side-blowing tower coolers. But yeah, a big difference even then.
|
let me preface by saying that practically, 1.3-1.4v is the max for ivy bridge for most people.
However, 'delidding' changes the dynamics of how the chip works given that it drops temps 20-30*C on average.
On ivy bridge you can put a ton of volts into the chip no problem, it's really heat that's the problem with ivy bridge. If you can keep your chip under 90-100*C (i personally go with 90*C as my max but 95,100 as an absolute max is 'okay' as long as you aren't running compute workloads 24/7 and consistently above 90*C). 1.4v is very conservative for ivy bridge, most people say 1.45v, 1.5v as the absolute lowest max voltages I've ever heard. People are running 1.6v for 24/7 overclocks with no problem.
There has really only been a single report of ivy bridge degrading, and that was a guy running 24/7 1.98v on air (if you can even call it 'air' when you stick your computer outside in a blizzard). Obviously, ivy bridge is new, it's possible that all of us running 1.5v+ are going to have our chips suddenly explode like a self-burning spy letter on the 9000th hour, but many people have been running 1.55v+ and run degradation checks monthly, and so far no one has seen any signs of any degradation at all. Until I see anyone, anywhere, say their ivy bridge degraded, I will not even think twice about running above 1.5v as long as temps are good.
Ivy bridge isn't anything like sandy, it's quite a hardy chip, you can pump insanely high voltages into it no problem.
Practically though, temperatures are a huge barrier on ivy bridge, and it's difficult to get past even 1.3v. You have to 'delid' the chip:
![[image loading]](http://i.imgur.com/OVaBbnPl.jpg)
with a razor blade in order to get far past 1.35v, no matter what kind of cooling you have. Personally, my temps are maxing at 85*C, but that's after delidding (which resulted in a 23*C temp drop), using coollaboratory liquid ultra on-die and IHS, using 120/140/140mm fans, and having 5 case fans (2 of which are 140s). The computer room isn't exactly 'toasty' either.
So most people will say yea, ivy bridge max voltages are around 1.35v, 1.4v, but that's because you have to delid the chip in order to push past those voltages without getting insanely hot. Even after delidding, it's hard to keep under 100*C with even high end cooling on above 1.5v, personally I start to see temps get way too hot around 1.55v, I would need a better cooler if I wanted to do 5.1ghz on my chip (which would require ~1.55-1.58v).
Ivy bridge actually runs much cooler than Sandy Bridge, but the problem is a sort of 'manufacturing defect' with the chip. Both sandy bridge and Ivy bridge have a fairly large gap between the actual chip, the die, and the IHS, the metal heatspeader on top. On sandy bridge this isn't a problem, because they use fluxless solder and it transfers heat near-perfectly. On ivy bridge, they use a slightly high end thermal paste, which is a great paste but it's not nearly as good as fluxless solder.
So what you have to do (short of replacing the TIM with fluxless solder, if you are comfortable with that...), is remove the glue on the chip that creates the gap (people who've delidded, rubbed the black glue off, and kept the original paste, reported similar temp drops).
![[image loading]](http://cdn.overclock.net/f/f2/120x120px-LS-f2601117_SpinningIHSshortfinal.gif)
On average, delidding results in a 20-30*C temp drop, which is much larger than going from air to custom loop even. In a sense, ivy bridge is great for budget overclockers willing to 'risk it all', people like myself who overclock for value and don't have much money.
Now delidding has it's risks, you are sticking a razor blade into the chip, but as far as I know, as a member of the largest official delidding club on the internet, not a single person has screwed up delidding when they used the right razor blade:
![[image loading]](http://www.killyourclutter.com/wp-content/uploads/2012/06/razor-blade.jpg)
A few people have screwed up delidding, but that is because they used a utility blade or knife and were morons. And most of those people, they didn't really have a dead chip, they just could no longer run RAM on dual channel, which imo is worth having a 20-30*C temp drop.
|
ic. that's insane. i didn't realize stock fans were that bad.
Belial mentioned ivy bridges (3rd gens) don't overclock well. And then he seemed to say they overclocked well. I'm sure he was referring to different aspects. Could you clarify?
|
|
|
|