Jump to content

batyanko

Registered User
  • Posts

    38
  • Joined

  • Last visited

  • Days Won

    1

batyanko last won the day on December 7 2017

batyanko had the most liked content!

Profile Information

  • Location
    Bulgaria

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

batyanko's Achievements

T|I Semi Advanced

T|I Semi Advanced (3/7)

8

Reputation

  1. Sorry, I have neither bins not even my 2570p any more. I hope some of the other people here do.
  2. Both are fine. In fact, a laptop can handle even higher range of voltages just fine, and especially high-end ones like EliteBook series. Yes, this is a known issue for laptops that were spilled on with Coca-Cola or beer :)) Stock+defaults should be always fine, unless you need something fancy. Maybe read what others achieved modding BIOS in that case Got a ~$5 2xUSB3.0 card on AliExpress got very good speeds, but rather fragile build quality.
  3. Try resetting BIOS defaults, plug one stick at a time, etc. Sometimes BIOS doesn't make good sense of SPD memory info to set correct timings and memory frequency, so you need to give it another chance. A bit surprising for me though, never noticed such problems with IvyBridge. Mostly with an AM3 AsRock mobo (it was so bad it was choosing timings that crashed the system), but there at least I could set timings and frequency manually
  4. That was interesting to test. 3720qm and a 3320m, both at 2.60 GHz nominal freqs. Test bench: Dell Latitude E6430, 2x8GB DDR3-1600 Linux Mint 20 at min. brightness radio off Battery off Idle Watts shown on Kill-a-watt thingie: 3720qm - 7.0 W 3320m - 6.8 - 6.9 W I would call that a 0.15 Watts difference, rather than 1.5-3.0 Watts. Hardly a factor to energy consumption. Though its is another question how often you will be at those Watts in Windows 10. Pretty rarely I guess. So here is a test of a modest load, watching .avi in VLC. Core loads averaging around 3%. 3720qm - ~14.5 W 3320qm - ~13.8 W Here the difference grows to about 0.8 W. Seems that the i7 is getting inefficient due all cores being woken up, while still not much to do. Another interesting note - underclocking CPU and iGPU doesn't seem to matter, those minimum frequencies seem to be the preferred and sufficient mode for it even without forcing these frequencies. Finally decided to pitch 100% load on 2 cores at 2400 MHz against 4 cores at 1200 MHz - to test the popular opinion that a quad-core at low frequencies is more efficient than a dual-core at higher frequencies. So I ran a stress test of 4 threads at 2400 MHz on the 3320m, an of 8 threads at 1200MHz on the 3720qm. That should in a crude way be the same amount of (totally useless) work done per amount of time. (note radios here are on and WiFi is connected, realized that too late... substract, say, 0.8 W to compare with above readings) 3720qm - ~20.3 W 3320m - ~21.2 W Here the 7720qm already seems to be more efficient with around 0.9 W. IMO the third example is the most realistic level of load for everyday work, keeping in mind that modern websites are quite energy-hungry and browsers making good use of multi-threading. So my best bet for energy efficiency would be an underclocked quad-core i7, say, anywhere between 1200 and 2600 MHz. And yes, running on battery is one use case where Linux has clear advantage over Windows 10. Linux tends to do very little background work, compared with Windows 10. So yes, that should do it. Make sure to underclock the iGPU too, that can be quite the power hog. Limitations: these results may be non-representative due to binning. My impression on other i5s is consistent with what I observed above, but I admit this is the only i7 I have observed in detail. Photos of le scientific measurement equipment: https://photos.app.goo.gl/G1ChgoxWdchgiEwX9
  5. Idle consumption of quad i7 should be comparable to dual core, so it seems that certain things create load for the CPU. It seems you will need to do some research on what uses background CPU, which could be something as ridiculous as the Google animation on Chrome start page. Common heavy offenders are website ads or background Windows 10 processes, but the latter is only fixed by switching to Linux... And of course you could underclock, I do this often on battery. 1200Mhz is I think minimum for most Ivy Bridge, 4 cores at that frequency is alright for office work. Nice thing about this, SpeedStep dynamically lowers voltage at lower frequency, so underclocking is extra efficient.
  6. Socket G2 is another name for FCPGA988, it will be fine. They just used random stock image of desktop CPU. $128 is bit of a rip-off, but what do I know, alternative being flaky eBay sellers...
  7. I have been playing recently with battery alternatives for the 8570p (all involving wiring...) 1. One thing I did, I ripped the stock cells from a cheap after-market battery and stuffed it with good 3500MAh cells. It works, but I find calibration a bit tedious, as the OS tends to sleep/power-off at voltages way above cell design termination (something like 3.3V vs 2.5V). Which results in less-than-spectacular Wh ratings 2. Another thing I did - wired another 3 cells in parallel, effectively getting a 9-cell battery. Bit longer-lasting, but re-calibration takes even longer (https://photos.app.goo.gl/qSzCnPaL2KGNvQ7i7) Both of the above rely on the battery (i.e. the board inside) being a non-original one, which generally is known to allow fiddling (original ones for example have pre-programmed max cycle count, and some even reduce capacity to protect ageing cells. Obviously inconvenient for our purposes). However my non-original battery turned out to be too much of a cheapo, it cuts at loads above ~60W which is a problem for some purposes. 3. This is my current project: feeding 19V directly to the power jack, by cable from an external battery. This consists of the following things: 3.1 Battery configuration of your own liking - for example a setup 3x3 of 18650 sticks, resulting in a battery of 11.2V nominal, and capacity of 10500MAh in case of 3500MAh cells. 3.2 DC-DC converter that converts whatever voltage the battery feeds into it, to 18.5 or 19.5 V (or anything in between or around it is usually fine). Probably $10-15 on AliExpress or your local electronics shop. 3.3 Charger jack with the center pin wired with resistors to the V+ wire, so that it gives between 5 and 6 volts (so we fool the laptop thinking it's a real charger). I'm not an expert in electronics so do your own reading if you want to do this one. One mistake I initially made was use low-ohms resistors (like 5 and 14 Ohms), which set them on fire... Now I'm using probably 5K and 14K Ohm resistors. I tested this one and it worked nicely. https://photos.app.goo.gl/7rPJje77v8hg6yZT6 Maybe always good to mention - be careful fiddling with the 18650 cells. They can release ridiculous amounts of energy if shorted (that is, you can set yourself on fire, etc.)
  8. Giving it fresh air would help, but blowing more air into the fan doesn't seem much use. I think it takes quite tight engineering to achieve a turbine-like system of fans, so that the air flow speed at the fins increases noticeably. In this case MioIV's considerations of the heatpipe being a bottleneck already makes sense. If you want to improve from here, you rather need to widen the heat transfer bandwidth, that is, connect more pipes or a bigger pipe to the cpu.
  9. Yes, drilling holes above the fan will drop a few degrees. I consider this sufficient for mostly throttle-less experience. Okay, if you want it to last decades you may want to keep it yet cooler, but otherwise I find these temps safe enough. Btw. I got bottomless and never found a (reasonably priced) bottom cover, thanks to all you guys buying spare ones to drill on xD The pipe is surely not quite up to par, but if you decrease the temps at one end of the heat transfer system (i.e. at fan radiator), that will still lead to similar decrease of temps on the other end (i.e. at the chip). The bigger the temperature difference, the more powerful the heat transfer. Air at fan will be colder with drilled holes, because it comes directly from outside, rather than from other parts of the laptop as per standard design.
  10. Yep it's a nice idea, that room can be utilized. That copper plate can interface with the original heat pipe, is that what you intended? You can use thermal glue for OK results or solder it for maximum performance. Soldering however takes skill and consideration (a lot of things that can mess up). Another thing, how about using heat pipes instead of that copper plate? Pipes are more effective than plain copper, and may even be easier to get. However I am not knowledgeable if the cheap pipes on Aliexpress are good enough.
  11. Having a cousin with the same laptop sounds like a fortunate thing :)) You can just try combinations between CPU, battery and laptop (i.e. motherboard) until you narrow it down to one component. Unfortunately I still imagine this to be faulty board or CPU, as you recently mentioned that you have experienced problems on charger, too. A 65W charger is completely alright even with a 45W quad, as the rest of the laptop is remarkably power-efficient. Also I have played with my 3720QM on one board that totally wouldn't deliver more than 35W - in that case the CPU and iGPU still remained stable, but simply throttled frequency to minimum.
  12. If that only happens on battery, it sounds that there is weak/faulty power supply from the battery. That might originate in the battery but also somewhere on the power circuitry on the board (VRMs, caps, etc): Reasons are that when you pull power from the battery, you eventually convert 12V to Vcore, and from PSU you pull 19V, which possibly uses other power circuitry. As you try to get something rendered, the iGPU gets involved which I have noticed produces quite some load spices, especially on the quad i7s. Edit: The quad i7 basically uses x2 power than any of the dual-cores Now that's theory and I don't know how this particular motherboard is organized, and also not quite sure if the whole thing depends on chipset/graphics drivers. So you can still try a different Windows or Linux version. I don't like to repeat myself too much :)) but I would again suggest the easiest way to be a live Linux USB stick with something like Linux Mint. Just boot from it and you get a ready OS with Intel's drivers to mess with. Open a browser, run some 3D shooter from the store or download FurMark (look for v. 0.7)
  13. I had a very similar problem with a 2540p, but only with AC power connected. So as @Phobosse said, probably a power circuit issue. That 2540p turned on about 1 in 5 times, with ~5 sec wait necessary between retries, and I still wouldn't complain. So if your laptop always turns on the second time, that still sounds like good enough :)) Anyway it `might` be worth to try another AC adapter, if you happen to have one around. I find the 2570p quite pretentious about AC adapters, refusing other brands even with the same voltage. For example it won't take a Dell charger, while a Dell Latitude E5430 will do alright with the HP charger.
  14. Looks quite interesting, can you post us with some details on the soldering process when you are done? I am going to play with some video card cooling these days, so I'm looking for advice on suitable process and equipment: type of blowtorch, type of solder (chemical composition details would be great), etc.
  15. Are you guys sure it is not cached RAM that just stays there, waiting to be overwritten when needed? 16GB being not enough for normal work sounds kinda extreme, I actually exchanged my 2x8GB for 2x4GB + some $$, I just couldn't fill it up over 8GB. My typical usage is Android Studio + 1-2 Android Emulators + Browser with ~10 websites. Many would call that a demanding combination, And it still doesn't add up to 8 GB. All this on Linux Mint / Kubuntu, but it isn't supposed to differ that much from Windows regarding RAM usage (though other stuff works quite a bit faster than on Windows...).
×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use. We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.