|
aarstar posted:2 of those in an Abit BP6 was amazing. I had one of these. That was a really cool computer; (entry level) SMP bragging rights for those too poor for a DEC Alpha. Of course, killjoy Intel put an end to all the fun (Alpha too, for, heh, Itanium).
|
#
?
Apr 27, 2013 09:03
|
|
|
|
| # ? May 9, 2024 14:59 |
|
|
Yudo posted:I had one of these. That was a really cool computer; (entry level) SMP bragging rights for those too poor for a DEC Alpha. Of course, killjoy Intel put an end to all the fun (Alpha too, for, heh, Itanium). Wasn't SMP linux kind of crude back in those days. Same goes for WinNT and Win2K.
|
|
#
?
Apr 27, 2013 20:19
|
|
|
Whoever was going high-clocked Pentium 4 and what not (as in, way past the initial models), my trusty dual P3-933 held a candle to anything multitasking with ease. And that on Windows 2000. The thing that impressed others the most is that despite high load, the UI just wouldn't want to fold (slow reactions and redraws), because back then, compared to single CPU systems, there was twice the chance of system and important UI threads getting runtime.
|
|
#
?
Apr 27, 2013 20:34
|
|
|
Shaocaholica posted:Wasn't SMP linux kind of crude back in those days. Same goes for WinNT and Win2K. By today's multicore/hyperthreading aware OSes, yeah, I guess it would be considered rather crude. I remember (keep in mind this was over a decade ago) Linux running like a champ, however. Workstations back then were nearly always dual cpu so there was significant development effort put towards SMP kernels. I eventually sold it to a friend when I bought a q6600 based machine who used it for years more in a home server role, granted with a much more mature Linux platform. I didn't use NT 4 because it was awful, and Win2k was at first limited. I used it only occasionally but it ran okay as well, though processes did not seem as well distributed. Once I got everything compiled, however, Linux ran very well.
|
|
#
?
Apr 27, 2013 21:37
|
|
|
Somewhat interesting Haswell update: Haswell has cut idle power usage so much that a new revision of the ATX power supply specification is required. The previous ATX v2.3 spec required power supplies to power loads as small as 6W on the +12V rail, Haswell will drop this requirement to 0.6W. Because older power supplies may not be able to handle loads this small and will shut down, there will be a BIOS option to disable the new C6/C7 states that provide the idle power savings.
|
|
#
?
Apr 28, 2013 19:52
|
|
|
Does 'older power supply' pretty much mean everything made before the new spec is in production?
|
|
#
?
Apr 28, 2013 20:03
|
|
|
It'll depend on the model, I'm guessing most non-bargain-basement power supplies should work. The old spec said it had to power AT LEAST down to 6W, so I figure many/most power supplies will exceed those requirements.
|
|
#
?
Apr 28, 2013 20:15
|
|
|
Whats the worst that can happen? The PSU shuts down the whole machine if the CPU goes into low power idle? e: I would think all modern PSUs don't shut down unless the are 'told' to regardless of how low the power draw is on any of its lines. Unless there's some conflict with the way '6w' minimum is implemented where the PSU will have to push 6w of power regardless of its its needed or not. Shaocaholica fucked around with this message at 20:30 on Apr 28, 2013 |
|
#
?
Apr 28, 2013 20:21
|
|
|
And/or involves fire.
|
|
#
?
Apr 28, 2013 20:25
|
|
|
Shaocaholica posted:Whats the worst that can happen? The PSU shuts down the whole machine? It causes an arithmetic underflow which leads to the PSU maxxing out and frying your machine 
|
|
#
?
Apr 28, 2013 20:30
|
|
|
Shaocaholica posted:Whats the worst that can happen? The PSU shuts down the whole machine if the CPU goes into low power idle? quote:e: I would think all modern PSUs don't shut down unless the are 'told' to regardless of how low the power draw is on any of its lines. Unless there's some conflict with the way '6w' minimum is implemented where the PSU will have to push 6w of power regardless of its its needed or not.
|
|
#
?
Apr 28, 2013 20:38
|
|
|
Alereon posted:The way power supplies are designed requires a minimum amount of current draw to produce stable power. You can't run a power supply with nothing connected to it, which is why power supply testers have resisters to produce a small load. If current draw is below the minimum for a given power supply, it will shut back off to prevent damage.
|
|
#
?
Apr 28, 2013 21:24
|
|
|
Doesn't matter anyway because on my desktop PC no matter what I try to do with the power settings in Windows, the system sleeps for about a split second and then immediately wakes up again. 
|
|
#
?
Apr 29, 2013 09:42
|
|
|
flavor posted:Doesn't matter anyway because on my desktop PC no matter what I try to do with the power settings in Windows, the system sleeps for about a split second and then immediately wakes up again. You might need to go into your device manager, right click on any sort of external peripherals like mouse/keyboards etc and change their power setting to 'cant wake up computer with this' or something. Also do the same check with network adapter and any USB stuff. But yeah the money spent on a new psu is probably not worth having the computer sleeping at 0.6w instead of 6w
|
|
#
?
Apr 29, 2013 10:27
|
|
|
DaNzA posted:You might need to go into your device manager, right click on any sort of external peripherals like mouse/keyboards etc and change their power setting to 'cant wake up computer with this' or something. Also do the same check with network adapter and any USB stuff. Thank you for the inspiration, I'll try this for the 15th time. Why can't they have a setting "ONLY wake up if..." instead? Or some kind of troubleshooting agent like "Hey, I noticed you had sent me to sleep just a minute ago and I woke up based on network traffic [or whatever applies]. Do you want to disable me waking up from this? Yes/No/Ignore in the future". But that would require actual thought about what users might want. Anyway, Haswell contributing to less wasting of power is certainly a good thing.
|
|
#
?
Apr 29, 2013 10:48
|
|
|
powercfg -energy can give you a report on your system, or just run -requests. I had a network share that was keeping my HTPC from sleeping once. Also C6/C7 and S3 are different things
|
|
#
?
Apr 29, 2013 17:49
|
|
|
Why do the Xeon E5450 and X5450 look identical on paper and also have the same release date yet one has a TDP thats far far greater than the other? http://ark.intel.com/compare/34446,33083 Binning?
|
|
#
?
Apr 29, 2013 21:14
|
|
|
Pretty much, yeah. About $75 more per chip for the same performance in 2/3 the TDP.
|
|
#
?
Apr 29, 2013 21:23
|
|
|
Shaocaholica posted:Whats the worst that can happen? The PSU shuts down the whole machine if the CPU goes into low power idle? As the power draw drops below 6W, the 12V rail will start creeping up to a higher voltage. If the PSU is well-behaving, at some point it can creep too far out of spec, and the PSU will trigger an overvoltage protection shutdown.
|
|
#
?
Apr 29, 2013 22:00
|
|
|
In that market (E5450/X5450), Intel is absolutely going to charge extra to lower the TDP because it's a major, major aspect of buying decisions for enterprise server OEMs and resellers. Somehow at the same time most of these customers don't bother thinking about lowering the power supply wattage rating to get increased power efficiency (I know Google and Facebook have customized PSU designs with surprisingly few parts that feed off of DC instead of AC, for example, but I don't consider them traditional enterprises in any manner besides budgets). When you're buying a crapton of blades at $10k+ / pop for several chassis filled, the $75 more is a drop in the bucket. This isn't to say that I believe it's actually cost-effective to buy such things (I'd say to focus on cutting down your likely stupidly bloated software on horribly tuned DBs and JVMs running GBs of dead code), but the point is mostly that Intel's getting margins out of this when consumer CPU margins are getting increasingly worse for the engineering dollars invested. I think of it as enterprise-aimed DLCs (read: "value-adds").
|
|
#
?
Apr 30, 2013 02:33
|
|
|
I like how 'X' means 'Performance' when its really just worse power consumption.
|
|
#
?
Apr 30, 2013 03:29
|
|
|
So I just did my once-every-three-years desktop computer upgrade: From a Phenom II X2 555 (running as an X4 955) to an Intel 3570K paired with a nice Asrock Z75 Pro3. Life is good, computer runs well, etc. Tried my hand at overclocking and found that I can hit 4.3ghz on stock voltage. In fact, I can even hit 4.3ghz with a -.050 voltage offset! However, whether on stock voltage or with the negative offset, Prime 95 was causing 3/4 cores to hit 105C and throttle back! As for cooling, I'm using a Hyper 212+, which I'd heard would be fine for a moderate overclock, but nobody else was reporting the same thermal issues as I was. I tried reapplying thermal paste and reseating the 212+. No progress. In a fit of insanity I used a razor blade and carefully delidded the 3570K. I had never delidded a processor before. I accidentally gouged the edge of the processor exposed a trace, but luckily I didn't cut it. I found it pretty nerve-wracking to delid the first time, but I feel I could do it again easily if I needed to. I peeled away some (but not all) of the black adhesive around the core. Cleaned the older thermal goop off and put a half-pea of AS5 on there. Reseated the integrated heatsink, clamped it down and reseated my 212+. As I was plugging the cables back in I was sweating bullets about the small gouge on the processor, but it booted up just fine! Now under Prime95 the cores all read a much more similar temperature than before (whereas before 1 core would constantly be 10+ C cooler than the rest), and best of all it averages ~75C on all cores under most Prime 95 tests, and ~94C under the most stressful tests. It ran Prime95 for 12 hours straight last night with no errors, so I'm calling it a success! However Intel should be ashamed of itself for using what must be the worst thermal paste known to man. And that's saying something, considering some websites out there have tested toothpaste to be an acceptable thermal paste in a dire situation.
|
|
#
?
Apr 30, 2013 15:52
|
|
|
bobfather posted:Prime 95 was causing 3/4 cores to hit 105C and throttle back! Jesus loving Christ man, throttle back a few hundred MHz. The 2-4% improvement in performance you're getting at 4.3GHz rather than 4.2 or 4.1 isn't worth sticking your processor in a frying pan. e: Recommended max temp on Prime95 is 72. Moment you go over that, your processor's lifespan starts dropping like a stone. coffeetable fucked around with this message at 16:06 on Apr 30, 2013 |
|
#
?
Apr 30, 2013 16:04
|
|
|
coffeetable posted:e: Recommended max temp on Prime95 is 72. Moment you go over that, your processor's lifespan starts dropping like a stone. Given that the guy delidded his CPU, I don't think he particularly cares/already knows.
|
|
#
?
Apr 30, 2013 16:17
|
|
|
.
sincx fucked around with this message at 05:54 on Mar 23, 2021 |
|
#
?
Apr 30, 2013 16:24
|
|
|
Intel is just going to put bath salts in there to piss everyone off.
|
|
#
?
Apr 30, 2013 16:31
|
|
|
coffeetable posted:Jesus loving Christ man, throttle back a few hundred MHz. The 2-4% improvement in performance you're getting at 4.3GHz rather than 4.2 or 4.1 isn't worth sticking your processor in a frying pan. Max recommended core temp is ~70, 70-90 is too hot and 90-105 is getting to be dangerously hot. Note that core temps are different than the overall CPU temperature, which is reported as a different (and usually lower) number. However, it's voltage that kills these chips, not necessarily heat. They're all designed to throttle back if they become unsafely hot. It's probably just the difference between a chip that lasts for 10 years before dying and one that lasts for 7 (under higher heat). Besides, very little that a computer does will load cores like Prime 95 does. I'd be surprised if the computer got higher than 70C under the most demanding of "real" applications. My post was more of a "WTF, Intel" post than anything else, since replacing the stock TIM with halfway decent stuff was able to drop temperatures by 15+C. bobfather fucked around with this message at 16:45 on Apr 30, 2013 |
|
#
?
Apr 30, 2013 16:37
|
|
|
sincx posted:Huh. I really hope Intel is going back to fluxless solder for.Haswell, but given how they haven't said anything to that effect, I'm not optimistic.
|
|
#
?
Apr 30, 2013 16:43
|
|
|
Alereon posted:Intel engineers and people in the know (Anandtech) agree that the issue of thermal contact between the CPU and IHS has been fixed, which should significantly improve overclocking. Intel isn't confirming exactly what was done, but I think our concerns have been addressed. Might still take a generation for people to stop delidding 
|
|
#
?
Apr 30, 2013 19:06
|
|
|
If it's soldered on, delidding it will act like delidding a Sandy Bridge CPU. As in, you can pry the heatspreader off, but it'll take a few million transistors with it. E: Also, TechReport updated on the low-power C-state PSU support for Haswell, and it doesn't look like many PSUs support idle power that low. Intel has a (probably incomplete) database of models that do support +12V draws that low. There are not many. It's likely that most BIOSes will have C6/C7 disabled by default to avoid problems, and Corsair thinks most of its models meet the requirements but hasn't completed testing yet. Factory Factory fucked around with this message at 20:06 on Apr 30, 2013 |
|
#
?
Apr 30, 2013 19:28
|
|
|
If you don't get your computer above 100C you can't be sure your components don't have any built up condensation on them. But that's for one standard atmosphere. This is why you want a negative-pressure cooling system -- to bring down the air pressure in the case far enough such that the liquid-gas phase transition of water happens before you're outside the safe operating temperature of your CPU.
|
|
#
?
Apr 30, 2013 19:44
|
|
|
Shaocaholica posted:Might still take a generation for people to stop delidding Haha, let them. I wonder if anyone tried to delid their Sandy Bridge? That would be hilarious
|
|
#
?
Apr 30, 2013 19:49
|
|
|
Factory Factory posted:If it's soldered on, delidding it will act like delidding a Sandy Bridge CPU. As in, you can pry the heatspreader off, but it'll take a few million transistors with it. Haha, right. I guess its only going to take a few people to figure that out before people get the point.
|
|
#
?
Apr 30, 2013 21:13
|
|
|
HalloKitty posted:Haha, let them. I wonder if anyone tried to delid their Sandy Bridge? That would be hilarious  quote:the pan method: cut the silicon, flip the cpu with thermal paste on a pan and wedge the blades in all four corners of the IHS. Heat the pan slowly on your kitchen gas heater and wait for a pop. Remove the cpu, let it cool down (VERY SLOWLY) and then remove the solder with either sand paper (2000 grit) or with a thin razor blade. Easy as cooking an egg! 
|
|
#
?
Apr 30, 2013 21:30
|
|
|
shrughes posted:If you don't get your computer above 100C you can't be sure your components don't have any built up condensation on them. But that's for one standard atmosphere. This is why you want a negative-pressure cooling system -- to bring down the air pressure in the case far enough such that the liquid-gas phase transition of water happens before you're outside the safe operating temperature of your CPU. I dunno if this is a joke or not (sorry) but this is the vapor pressure curve for water http://en.wikipedia.org/wiki/Water_(data_page)#Water.2Fsteam_equilibrium_properties and fans will pull like, a tenth of a psi, so negative pressure in the case makes very little difference. There also the water pressure drop from the core to the case atmosphere (probably worth a few #s of pressure), and the fact that the temperature's high enough normally to transport water from the core to case by diffusion until you equilibrate at whatever the Rh is in the room. Fart booger.
|
|
#
?
Apr 30, 2013 21:49
|
|
|
bobfather posted:Max recommended core temp is ~70, 70-90 is too hot and 90-105 is getting to be dangerously hot. Dude, no. The processor should be low 70s under Prime95, not 90s. You have done something wrong. Heat will most definitely gently caress up processors. Please get thee to the overclocking thread.
|
|
#
?
Apr 30, 2013 22:18
|
|
|
bobfather posted:So I just did my once-every-three-years desktop computer upgrade: You might want to check to make sure your motherboard isn't automatically raising the voltage with the overclock. Stock voltage should never get an IB that hot.
|
|
#
?
Apr 30, 2013 22:26
|
|
|
bobfather posted:However, it's voltage that kills these chips, not necessarily heat. They're all designed to throttle back if they become unsafely hot. It's probably just the difference between a chip that lasts for 10 years before dying and one that lasts for 7 (under higher heat). Yeah, no. What kills chips is electromigration, and the mean time to failure of a conductor due to electromigration has an (approximately) exponential dependence on temperature. If you're a frequent gamer and your games bring your chip up to 90-100C, I wouldn't be surprised if it burns out before the year's done. Also, the processor will throttle down to prevent imminent damage. It won't stop long-term damage from accumulating. coffeetable fucked around with this message at 00:34 on May 1, 2013 |
|
#
?
Apr 30, 2013 22:30
|
|
|
Mofabio posted:I dunno if this is a joke or not (sorry) but this is the vapor pressure curve for water Yeah, it's a joke. Not to mention that lowering the air pressure proportionately lowers heatsink performance.
|
|
#
?
Apr 30, 2013 22:38
|
|
|
|
| # ? May 9, 2024 14:59 |
|
|
KillHour posted:You might want to check to make sure your motherboard isn't automatically raising the voltage with the overclock. Stock voltage should never get an IB that hot. I did some fiddling around with the only setting I thought might cause an issue, which was the CPU load-line calibration. I was using a fairly aggressive value to compensate for vdroop, but it turns out that setting might have been giving a boost in voltage, perhaps +.05 or so. Changing the setting to be more conservative has resulted in a more reasonable core voltage of 1.280 under full load, for temps of ~70 for all cores under Prime 95. Thanks for the tip!
|
|
#
?
May 1, 2013 00:24
|
|