|
Heh, so, someone on Reddit mentioned memory had gone up and I took a look. I built a 5820K system about 10 months ago and that appears to have been a super good decision. I paid $63 shipped each for two 2x8 GB Geil Evo Potenza DDR4-3000 kits (32 GB total) and right now Newegg has them for $120 a pop  
|
#
?
Feb 24, 2017 06:22
|
|
|
|
| # ? May 15, 2024 20:07 |
|
|
Also - Ryzen will not support ECC. So much for that build.
|
|
#
?
Feb 24, 2017 06:27
|
|
|
Paul MaudDib posted:Also - Ryzen will not support ECC. So much for that build. AMD has never listed ECC support in a desktop product, but should ECC be present and the motherboard have the correct traces, the CPU will use it. The same thing seems to apply to Ryzen, as AMD have been careful not to deny it outright. This is really about market segmentation ( no cheap desktop CPUs being sold in server roles ) than anything, so all AMD needs is the ability to deny warranty support to an OEM if they tried it, not actually lock out the feature. Of course, you'll need to wait until MBs are tested to see if any of them have enabled the ability.
|
|
#
?
Feb 24, 2017 06:36
|
|
|
Paul MaudDib posted:Also - Ryzen will not support ECC. So much for that build. And consumers everywhere gave not one single poo poo.
|
|
#
?
Feb 24, 2017 06:37
|
|
|
Seamonster posted:I just want cpu/gpu combos for sale. AMD should leverage the fact they're the only ones in the world who can do this. How come? Some sort of legal/monopoly reason with Intel?
|
|
#
?
Feb 24, 2017 07:38
|
|
|
I imagine Intel don't have the experience needed to produce that specific kind of chips.
|
|
#
?
Feb 24, 2017 08:27
|
|
|
Yeah, the PS4/xbone SOCs are pretty unique from a piddling engineering perspective, when traditionally the CPU and GPU and I/O were separate for high-performance parts. The Tegra chips are also among the same vein, but they're meant for <8w instead of the 95-140w TDP the PS4/bone chips have. Anime Schoolgirl fucked around with this message at 09:07 on Feb 24, 2017 |
|
#
?
Feb 24, 2017 09:04
|
|
|
Measly Twerp posted:And consumers everywhere gave not one single poo poo. Eh, rowhammer is a thing.
|
|
#
?
Feb 24, 2017 09:30
|
|
|
Kerbtree posted:Eh, rowhammer is a thing. Measly Twerp posted:And consumers everywhere gave not one single poo poo.
|
|
#
?
Feb 24, 2017 09:36
|
|
|
ECC on Ryzen is a cheap and easy way to set up a little NAS for home and I'm looking forward to it.
|
|
#
?
Feb 24, 2017 09:40
|
|
|
Consumers probably should care about Rowhammer given that there's the potential for some javascript to bypass the entire CPU "ring" security model - potentially even to the level of the management engine or secure enclave, from which it is 100% impossible to extract an attacker (by design). http://news.softpedia.com/news/rowhammer-attacks-in-microsoft-edge-can-compromise-entire-pc-due-to-new-research-504574.shtml That actually goes for phones too. ECC really shouldn't be optional anymore at this point - 99.999% security is not good enough against an attacker who gets unlimited tries. PerrineClostermann posted:ECC on Ryzen is a cheap and easy way to set up a little NAS for home and I'm looking forward to it. Yeah I was hoping to build a mITX NAS on it - ideally with a mobo that had an extra SATA controller so I could keep the PCIe open. Intel C2550 it is, I guess (Asrock has one with like 12 SATA ports on it). Paul MaudDib fucked around with this message at 10:43 on Feb 24, 2017 |
|
#
?
Feb 24, 2017 10:16
|
|
|
Harik posted:Are there any 8C+APU on the roadmap? I've got machines that would be perfect for - no GPU required for driving text logins, and the PCIe slots are used by 10ge and RAID controllers. Would you have any slots free at all? Because you can always just shove some bargain bin low profile GPU into a PCI slot or something if all you need is a text console. Edit: or if you are looking at a board with no PCI slots at all, maybe something that only uses one PCIe lane: http://www.anandtech.com/show/10269/zotac-quietly-releases-pcie-x1-gt-710-graphics-card Rastor fucked around with this message at 10:36 on Feb 24, 2017 |
|
#
?
Feb 24, 2017 10:24
|
|
|
Paul MaudDib posted:Consumers probably should care about Rowhammer given that there's the potential for some javascript to bypass the entire CPU "ring" security model - potentially even to the level of the management engine or secure enclave, from which it is 100% impossible to extract an attacker (by design). These are the same consumers who train themselves to click away red warning screens in chrome.
|
|
#
?
Feb 24, 2017 10:45
|
|
|
Paul MaudDib posted:Consumers probably should care about Rowhammer given that there's the potential for some javascript to bypass the entire CPU "ring" security model - potentially even to the level of the management engine or secure enclave, from which it is 100% impossible to extract an attacker (by design). ECC alone isn't a good enough protection against rowhammer, so no. e: since the memory corruption can possibly flip more than one bit, and ECC can only handle 1 bit errors. If your system isn't detecting the rowhammer attack, multiple attempts eventually work. but current and near future hardware has protections against this type of attack, without the need for ECC. Klyith fucked around with this message at 10:57 on Feb 24, 2017 |
|
#
?
Feb 24, 2017 10:52
|
|
|
https://www.overclock3d.net/reviews/cpu_mainboard/amd_ryzen_rog_am4_crosshair_vi_hero_preview/1 A nice detail that slipped under the radar because ASUS forgot to tell anyone - the ROG Hero is backwards compatible with existing AM3 cooler mounts. Hopefully that extends to their cheaper boards as well.
|
|
#
?
Feb 24, 2017 17:34
|
|
|
https://www.youtube.com/watch?v=mHs0lSfAMVI (put the subs on). Also, some AM4 motherboard links, including a handy table for VRMs, Audio and Network. https://www.reddit.com/r/Amd/comments/5vx5o3/am4_motherboard_links/ Dante80 fucked around with this message at 18:15 on Feb 24, 2017 |
|
#
?
Feb 24, 2017 18:13
|
|
|
PerrineClostermann posted:ECC on Ryzen is a cheap and easy way to set up a little NAS for home and I'm looking forward to it. Is there a good reason for me to give a poo poo about having ECC ram on the box storing my linux ISOs? What is it protecting me against? If the result of a flipped bit in ram getting stored to disk is a flicker of a corrupted frame in a video file, it's not exactly something I think I'd worry about paying extra for, but I don't really know. (I've seen tons of posts re: I WANT ECC FOR MY HOME SERVER and it occurred to me that I didn't know why).
|
|
#
?
Feb 24, 2017 18:29
|
|
|
Gwaihir posted:Is there a good reason for me to give a poo poo about having ECC ram on the box storing my linux ISOs? What is it protecting me against? If the result of a flipped bit in ram getting stored to disk is a flicker of a corrupted frame in a video file, it's not exactly something I think I'd worry about paying extra for, but I don't really know. There's debate on whether ECC is necessary for things like ZFS https://news.ycombinator.com/item?id=13678629
|
|
#
?
Feb 24, 2017 19:07
|
|
|
Gwaihir posted:Is there a good reason for me to give a poo poo about having ECC ram on the box storing my linux ISOs? What is it protecting me against? If the result of a flipped bit in ram getting stored to disk is a flicker of a corrupted frame in a video file, it's not exactly something I think I'd worry about paying extra for, but I don't really know. ECC itself has a pretty small price premium over non-ECC RAM, people are just getting bit in the rear end by Intels market segmentation. Yes, you probably do want ECC on a machine that you're gonna have running 24/7 for years, if only to avoid the random restart when a soft error happens. No, ECC isn't worth a big price premium, but if consumer CPUs/chipsets supported ECC then everybody would be buying ECC for home servers. The issue is that Intel doesn't support ECC on consumer platforms, but AMD might. Edit: Also the more RAM you have the more you want ECC. You're much more likely to run into more soft errors on a Xeon-D NAS box with 128GB of RAM than a bare minimum FreeNAS setup with 8GB.
|
|
#
?
Feb 24, 2017 19:17
|
|
|
I'm running a 32TB ZFS array w/ 32GB of non ECC RAM. 
|
|
#
?
Feb 24, 2017 19:29
|
|
|
If there's not a significant risk of actual data on disk corruption, (And there doesn't seem to be such a risk at all), then it looks like ECC ram isn't really a selling point at all for home NAS use. I'm sure as hell not stuffing 128 gigs of ram in a home server, either, if it's a NAS I'd rather spend that cash on more disks. For a real server, of course, it'd be a given. (My home nas is an older i3, ivb or haswell maybe, running FreeNAS with 8 gigs of ram and 12 4tb reds).
|
|
#
?
Feb 24, 2017 19:31
|
|
|
The ideal ZFS setup is 1gb of ECC RAM per TB, iirc. It's important, because it scrubs through all the data fairly frequently and a bit flip during the scrub will corrupt your data
|
|
#
?
Feb 24, 2017 19:36
|
|
|
PerrineClostermann posted:The ideal ZFS setup is 1gb of ECC RAM per TB, iirc. It's important, because it scrubs through all the data fairly frequently and a bit flip during the scrub will corrupt your data This is straight up wrong and debunked by one of the co-founders of ZFS. 
|
|
#
?
Feb 24, 2017 19:47
|
|
|
ECC protects you from cosmic rays, the deadliest rays of them all
|
|
#
?
Feb 24, 2017 19:51
|
|
|
Dante80 posted:https://www.youtube.com/watch?v=mHs0lSfAMVI I don't get the memory support. Do all the AsRock boards just straight up end at 2667? There's no mention of XMP in their online spec sheet even though its a pretty normal feature by now. And the (OC) support on some boards - is that XMP?????
|
|
#
?
Feb 24, 2017 19:54
|
|
|
PerrineClostermann posted:The ideal ZFS setup is 1gb of ECC RAM per TB, iirc. It's important, because it scrubs through all the data fairly frequently and a bit flip during the scrub will corrupt your data in addition to the ram requirements being wrong, zfs' scrub doesn't work that way at all
|
|
#
?
Feb 24, 2017 19:57
|
|
|
IF they do in fact not support it rest assured you can just manually set the timings and voltage for XMP compatible sticks.
|
|
#
?
Feb 24, 2017 19:57
|
|
|
AMD should have copied the 115x mounting points. Though i guess it still wouldn't make intel coolers compatible due to height.
|
|
#
?
Feb 24, 2017 20:14
|
|
|
hifi posted:in addition to the ram requirements being wrong, zfs' scrub doesn't work that way at all Yea. The ram Requirement was mainly if you did DeDupe last I thought? Outside of that I have had no issues with my little NAS's at home with RED's and 8G ram. Though I think 1 of them is ECC (the little HP Micro Server (7th gen I think. AMD CPU) my last job was just going to throw out because it "didn't work" and kept locking up trying to install ESX/WinServer 2012. Turns out all it needed was a Bios update. I have some FreeNAS -> Nas4Free questions I guess I can ask in the NAS forum instead of here.
|
|
#
?
Feb 24, 2017 20:25
|
|
|
Stanley Pain posted:This is straight up wrong and debunked by one of the co-founders of ZFS. hifi posted:in addition to the ram requirements being wrong, zfs' scrub doesn't work that way at all When did this all happen? 
|
|
#
?
Feb 24, 2017 20:43
|
|
|
PerrineClostermann posted:When did this all happen? The co-founder never said that ECC isn't important. It's just that ECC is as important for ZFS as it is for any other file system. https://arstechnica.com/civis/viewtopic.php?f=2&t=1235679&p=26303271#p26303271 I set my little Dell T20 homeserver to log corrected errors and it found two in the last ~6 months — both could have been potential crashes or data corruption on non-ECC systems. I'm not sure I'd go out of my way to buy a Xeon + C-series chipset MB + ECC ram over the consumer versions to replace this machine but it'd be nice to have consumer Ryzen support it.
|
|
#
?
Feb 24, 2017 22:12
|
|
|
Yea, I think two just potential crashes in 6 months means I can quite safely not give a poo poo for home server use.
|
|
#
?
Feb 24, 2017 22:33
|
|
|
I don't care about crashes but I do care about data corruption. Here's an Ars article that shows how a single bit flip can corrupt a file: https://arstechnica.com/information-technology/2014/01/bitrot-and-atomic-cows-inside-next-gen-filesystems/
|
|
#
?
Feb 24, 2017 22:37
|
|
|
Seamonster posted:And the (OC) support on some boards - is that XMP????? I know there were memory modules that supported AMP for AM3 systems and that wasn't a very popular overclocker's platform so if they'd put them out for that they'll put them out for AM4 eventually too.
|
|
#
?
Feb 24, 2017 22:43
|
|
|
Goons: consumers should care about ECC memory on their home fileservers to protect against single-bit errors in their next-gen journaling filesystems corrupting backups! Consumers:  Prosumers: Yes, backups! You should make backups! I got this nifty hard drive box from seagate that does backups when I push the button.
|
|
#
?
Feb 24, 2017 23:02
|
|
|
No true consumer would have a home server  Also NAS/ZFS/RAID/etc. is not backup.
|
|
#
?
Feb 24, 2017 23:08
|
|
|
PerrineClostermann posted:No true consumer would have a home server Also they should reallllllly care about quad channel memory because reasons.
|
|
#
?
Feb 24, 2017 23:26
|
|
|
I dunno man that sounds a bit excessive
|
|
#
?
Feb 24, 2017 23:35
|
|
|
https://forums.anandtech.com/threads/official-amd-ryzen-benchmarks-reviews-prices-and-discussion.2499879/page-32 Rumor going around (Well, it is the largest tech publication in Turkey) that the R7 1700 is only capable of 4.0Ghz across all cores and will literally kill the VRMs if you attempt to push anymore voltage with low end boards. However, the R7 1700X and 1800X, from the same source as above, will beat the 7700Ks stock performance (so 4.4Ghz 1800X ~ 4.5Ghz 7700K?). Temps, power draw, voltage, etc were within safe limits unlike R7 1700. Seems R7 1700s (and likely all non X CPUs) are the real garbage parts which can't clock worth a drat. This is a similar story to Polaris 11, @ 850Mhz it pulls 30-35W, @ 1200Mhz it'll pull near 65W+. I guess Kabylake still has a niche in absolute niche for ST performance but honestly if the 1800X is hitting 4.4Ghz all cores without throwing up any alarms, then later 1600X and 1400X should be monsters, and a 1400X vs 7700K comparison would heavily favor the 1400X from a price/perf point.
|
|
#
?
Feb 24, 2017 23:52
|
|
|
|
| # ? May 15, 2024 20:07 |
|
|
I'm more interested in when goons / sites that don't partly(?) break NDA start reviewing them. But this kinda stands out:quote:"Single thread score will be so great. According to this performance we can say that 7700K will be history, even for gaming, from now on" he said. There is so much hype about the platfrom right now, the aftermath will be fantastic if it sucks balls. GRINDCORE MEGGIDO fucked around with this message at 00:47 on Feb 25, 2017 |
|
#
?
Feb 25, 2017 00:09
|
|
