|
It's almost as if he's producing entertainment and not cold scholarly computer knowledge.
|
# ? Jan 2, 2017 12:25 |
|
|
# ? May 13, 2024 12:56 |
|
I wish someone would sponsor me an 6950X or two, instead of Linus.
|
# ? Jan 2, 2017 14:22 |
|
WAR DOGS OF SOCHI posted:Stop the radiator hate. A fanless pc is a dream of mine and idgaf about trying to cram everything into something the size of a pack of cigs https://linustechtips.com/main/topic/707325-calyos-nsg-passive-cooling-kit/
|
# ? Jan 2, 2017 15:22 |
|
Boiled Water posted:It's almost as if he's producing entertainment and not cold scholarly computer knowledge.
|
# ? Jan 2, 2017 17:43 |
|
Its almost like hes doing it on purpose...
|
# ? Jan 2, 2017 18:00 |
|
He just always comes off like a sophomore competing on his high school forensics team.
|
# ? Jan 2, 2017 18:05 |
|
WAR DOGS OF SOCHI posted:Sorry for the derail, but I love this stuff. Are you saying that they don't bother cooling any more than is absolutely necessary? To the point that a human being could suffer heatstroke or something while in the racks? lmao that's crazy It's not usually hot enough to cause heatstroke for the time someone would actually be in the room, but it'll get up pretty hot at times, easily like 100F. The basic reasoning is that since the server vendors will now agree to warranty their equipment to run constantly at the higher temperatures, why should you try to force temps down to the 60s-70s F range anymore? You spend a ton less on power, which is good for saving money and great for the environment in general. If I remember right, Google was one of the pioneers of the practice and then the influence of their testing convinced the manufacturers to allow it. In many places, this practice results in near 0 use of air conditioning equipment over large stretches of the winter and even spring and fall, since natural cooling to the environment happens quick enough if you set like 100F or 110F as the point where the AC has to come on.
|
# ? Jan 2, 2017 19:01 |
|
Which lead to interesting shenanigans https://www.youtube.com/watch?v=9Sgaq6OYLX8
|
# ? Jan 2, 2017 19:51 |
|
Josh Lyman posted:Pretty sure they don't. I think Linus used to work for Canadian version of Newegg. And he was doing pretty much the same thing, hosting NCIX Tech Tips. I watch quite a bit of their videos, but you shouldn't do it if you want to extract useful information. The reason to watch them is top see them do something silly and ridiculous that you don't usually see. Sometimes they also have interesting ideas, like the HTC Vive extension cord, you just need to wait for someone to execute them better.
|
# ? Jan 2, 2017 21:38 |
|
thebigcow posted:Which lead to interesting shenanigans https://www.youtube.com/watch?v=9Sgaq6OYLX8 That was an extremely enjoyable video.
|
# ? Jan 2, 2017 21:39 |
|
Paul MaudDib posted:
I think I'm pretty safe in saying there's probably very few people here who don't want AMD to make good CPU's and GPU's again. I'm glad we don't have the toxic fanboy shitholers here.
|
# ? Jan 3, 2017 00:05 |
|
thebigcow posted:Which lead to interesting shenanigans https://www.youtube.com/watch?v=9Sgaq6OYLX8 This was worth my time. Thank you for posting this, everyone else should consider watching. That's some hilarious and frightening stuff.
|
# ? Jan 3, 2017 01:02 |
|
Wistful of Dollars posted:I think I'm pretty safe in saying there's probably very few people here who don't want AMD to make good CPU's and GPU's again. I'm glad we don't have the toxic fanboy shitholers here.
|
# ? Jan 3, 2017 01:45 |
|
Awwww yeah!
|
# ? Jan 3, 2017 02:05 |
|
Z270 boards on available on eBay.
|
# ? Jan 3, 2017 02:52 |
|
thebigcow posted:Which lead to interesting shenanigans https://www.youtube.com/watch?v=9Sgaq6OYLX8 I'll nTh watching this talk.
|
# ? Jan 3, 2017 03:15 |
|
It's basically a really good use case for having ECC be standard on everything everywhere. The marginal increase in cost can offset a lot of edge case wonky behavior like this.
|
# ? Jan 3, 2017 04:24 |
|
Methylethylaldehyde posted:It's basically a really good use case for having ECC be standard on everything everywhere. The marginal increase in cost can offset a lot of edge case wonky behavior like this.
|
# ? Jan 3, 2017 05:00 |
|
Professor Science posted:everybody's using ECC now because of https://en.wikipedia.org/wiki/Row_hammer if not any of the other million reasons I'm not sure who your definition of "everybody" includes, but it certainly isn't desktops, laptops, or mobile basically anything (phones, tablets, etc).
|
# ? Jan 3, 2017 05:09 |
|
DrDork posted:I'm not sure who your definition of "everybody" includes, but it certainly isn't desktops, laptops, or mobile basically anything (phones, tablets, etc).
|
# ? Jan 3, 2017 05:12 |
|
Let's also pretend software error detection/correction never existed and those extra ECC bits in the RAM will also be given away for free.
|
# ? Jan 3, 2017 06:43 |
|
DrDork posted:I'm not sure who your definition of "everybody" includes, but it certainly isn't desktops, laptops, or mobile basically anything (phones, tablets, etc). the same results he mentions at the end are why you don't see ECC in desktop/mobile yet--you just don't see that many failures right now due to the kinds of things that ECC would prevent.
|
# ? Jan 3, 2017 08:45 |
|
Why do I read comments? quote:3. Once a system architecture is completely fault-tolerant there is NO LIMIT to system size, so you can step-and-repeat many large processors on a wafer and reasonably expect them ALL to work - regardless of the thousands of bad transistors sprinkled throughout them. Hence, by simply connecting the adjacent pads of adjacent processors on a wafer, you can cheaply manufacture entire modern supercomputers-on-wafers - for a few thousand dollars each. Hmmm sounds pretty easy
|
# ? Jan 3, 2017 09:17 |
|
Professor Science posted:the same results he mentions at the end are why you don't see ECC in desktop/mobile yet--you just don't see that many failures right now due to the kinds of things that ECC would prevent. I guess what's implied is that any specific user doesn't see many failures, but if radiation-related memory bit flips are really that common you're making an argument for using ECC in more places for sure. If it was supported on Intel's consumer platforms, would we be recommending people spend the 20% more on RAM to have ECC?
|
# ? Jan 3, 2017 15:41 |
|
Depends on if the platform were still flexible enough to take non-ECC RAM (if not, it's all just RAM now), and if the economies of scale wouldn't pull the price closer in line to what non-ECC RAM costs.
|
# ? Jan 3, 2017 15:42 |
|
dont be mean to me posted:Depends on if the platform were still flexible enough to take non-ECC RAM (if not, it's all just RAM now), and if the economies of scale wouldn't pull the price closer in line to what non-ECC RAM costs. I don't have a good estimate of how large the laptop/desktop vs server markets are, but I'd expect that the trend is certainly towards more server RAM being consumed and less laptop/desktop, given that the future seems to hold ARM-based mobile clients and IOT devices talking to ever-growing datacenters. I wouldn't be surprised if right now similar volumes of ECC and non-ECC memory are sold.
|
# ? Jan 3, 2017 15:51 |
|
DrDork posted:I'm not sure who your definition of "everybody" includes, but it certainly isn't desktops, laptops, or mobile basically anything (phones, tablets, etc).
|
# ? Jan 3, 2017 16:04 |
|
I want ECC too right now, though more for an upgrade to my NAS than anything else
|
# ? Jan 3, 2017 17:17 |
|
I have that in my NAS from the get-go. Kind of pointless to have a fully checksummed on-disk format, when poo poo can get hosed up in memory waiting to be written to disk. That Youtube video earlier kind of wants me to harden up my desktop now too.
|
# ? Jan 3, 2017 17:24 |
|
One sore place for ECC RAM is for laptops. ECC SO-DIMMs are super obscure enough that I spent a week just trying to find a place that had one listed besides "call us for volume pricing" as its details. There's a couple mini ITX boards that would be great for home media / file server uses whose use of ECC SO-DIMMs is both great and terrible. Add in the market for folks that'd like 32 GB of RAM in their laptops (read: doofuses like me that run a bunch of Docker containers and VMs to test and develop modern horsecrap software) and it's really strange why the industry's been silent on this area.Twerk from Home posted:I wouldn't be surprised if right now similar volumes of ECC and non-ECC memory are sold. Also, another interesting part of the equation is that Intel CEO Brian Krzanich said a couple years ago that for each mobile device sold Intel is selling several more CPUs to support the services that phone consumes. It's kind of sobering how inefficient the software produced by the software industry is when each user's impact upon a backend is that massive.
|
# ? Jan 3, 2017 17:56 |
|
Combat Pretzel posted:I have that in my NAS from the get-go. Kind of pointless to have a fully checksummed on-disk format, when poo poo can get hosed up in memory waiting to be written to disk. That Youtube video earlier kind of wants me to harden up my desktop now too. My NAS is my old e6750 system I used before this 2600k machine.
|
# ? Jan 3, 2017 18:14 |
|
First proper review I've seen (German) https://www.computerbase.de/2017-01/intel-kaby-lake-test-core-i7-7700k-i5-7600k/ 7700k is pushed to the limit, no undervolting possible and you need a beefy air cooler to keep it from temp throttling at stock. You'll need to delid to get any significant OC results. Intels thermal paste still sucks. 7600k has a little more room to uv or oc unsurprisingly. AVX microcode bug that requires a BIOS update to fix, spikes power draw. Also Z270 introducing AVX offset BIOS option (apparently that's already a thing serverland) that limits clocks when AVX is used.
|
# ? Jan 3, 2017 18:29 |
|
So Kaby Lake has zero IPC gains. Not surprised at all. AMD, please be good this time! E: reviews were posted some time ago already https://www.techpowerup.com/reviews/Intel/Core_i7-7700K_vs_6700K_Game_Performance/ http://www.anandtech.com/show/10968/the-intel-core-i7-7700k-91w-review-the-new-stock-performance-champion
|
# ? Jan 3, 2017 18:31 |
|
http://www.anandtech.com/show/10959/intel-launches-7th-generation-kaby-lake-i7-7700k-i5-7600k-i3-7350k Anandtech's article is up, although their quality has gone way down since Anand left. They also found the 7600K to run cooler and have more OC headroom than the 7700K. i3-K looks like a badass part. Overall there's not much here for people over Skylake, but a smidge more clocks.
|
# ? Jan 3, 2017 18:32 |
|
I3 K is also dual core and has terrible pricing. For same price you can buy a faster cpu used, with more cores, or if you want new, a quadcore for not much more..
|
# ? Jan 3, 2017 18:34 |
|
necrobobsledder posted:One sore place for ECC RAM is for laptops. ECC SO-DIMMs are super obscure enough that I spent a week just trying to find a place that had one listed besides "call us for volume pricing" as its details. There's a couple mini ITX boards that would be great for home media / file server uses whose use of ECC SO-DIMMs is both great and terrible. Add in the market for folks that'd like 32 GB of RAM in their laptops (read: doofuses like me that run a bunch of Docker containers and VMs to test and develop modern horsecrap software) and it's really strange why the industry's been silent on this area. ECC has been an option in laptops since what, late 2015?
|
# ? Jan 3, 2017 19:00 |
|
New NUCs up on Intel's product page: http://www.intel.com/content/www/us/en/nuc/products-overview.html If pricing stays as rock bottom on the Celeron based NUCs as previous generations, the J3455-based one should be a great value HTPC with 4K-capable decoding and output hardware.
|
# ? Jan 3, 2017 19:08 |
|
Ihmemies posted:I3 K is also dual core and has terrible pricing. For same price you can buy a faster cpu used, with more cores, or if you want new, a quadcore for not much more.. Besides like the 7700K, how much OC can you realistically get for a K-chip that is already at base 4.2GHz....maybe 10%? I would rather see a AMD branded i5-K Haswell at <$150 but I'm not holding my breath for that either. Palladium fucked around with this message at 19:27 on Jan 3, 2017 |
# ? Jan 3, 2017 19:25 |
|
Anand's sample was reportedly middling and their non-AVX OC topped out at 4.9ghz before voltage and heat got to be too much for their cooling. So like 14% ish.
|
# ? Jan 3, 2017 19:28 |
|
|
# ? May 13, 2024 12:56 |
JayzTwoCents must have gotten a golden sample, he's got his 7700k running stable at 5GHz using an old H100i for cooling.
|
|
# ? Jan 3, 2017 19:32 |