|
Krailor posted:Intel's primary customer base (datacenters buying Xeons by the thousands) have indicated that they're much more interested in performance-per-watt rather than just raw performance. So that's where Intel has been focusing their research over the past several years and they've made huge gains in that department. I've physically placed ~3000 e5 2680v3's in boards in the past 5 months. Can confirm.
|
# ¿ Nov 23, 2015 22:54 |
|
|
# ¿ Apr 24, 2024 22:04 |
|
Gravitas, how many K-corners were you running and what hardware setup? I'm personally curious because I work for an integrator and I want to see hard numbers against, say, K80s, given a modicum of optimization. For reference, I have CUDA benchmark numbers for 4 K80s with dual e5-26xx's on our hardware configuration. I know the whole point of K-landing is you shouldn't need to optimize, I just want to know where the sweet spot is for that optimization, especially versus CUDA.
|
# ¿ Mar 30, 2016 03:47 |
|
2695 is the good stuff this cycle. It crushes the 2680 in Multimedia and bests it by about 10% in Arithmetic.
|
# ¿ Mar 31, 2016 21:41 |
|
Kazinsal posted:NATEX usually has a couple in stock. They're used, tested, and in working condition with a 60 day warranty. They're basically a wholesale previous-previous-generation electronics recycling/reseller. I cannot find this place online. Link?
|
# ¿ Apr 18, 2016 21:28 |
|
mmkay posted:Something I've been wondering for a while - how many Intel employees are on SA? I don't work for Intel but I'm in the high end server business and I'm under NDA for everything Xeon. So, close? Related: needed to replace my workstation at work so I retired my 2500k based z97 to that duty and bought a 6700k/z170 with Samsung M.2 in anticipation of GTX1080.
|
# ¿ May 13, 2016 20:25 |
|
We're having some, uh, interesting stability issues with Broadwell-EP on boards that worked just fine on Haswell. AMI and Intel are stumped.
|
# ¿ May 18, 2016 14:40 |
|
Aquila posted:Uh Oh, and Dell just "upgraded" our most recent order to v4 xeons for "free". Essentially a BIOS issue with two faces. If we use the Broadwell BIOS and put a Haswell in, CPU1 doesn't show up if its a "-L." Sorta similarly, non-engineering sample 2680v4's (our primary go-to) don't show up in the BIOS at all, whereas all other v4s work fine. Mind you, the engineering sample 2680v4's work fine in the same serialized board.
|
# ¿ May 18, 2016 21:47 |
|
Anime Schoolgirl posted:seems like they pushed a testing bios to production because there can't possibly be any other reason why Haswell isn't recognized Right but why would there only be issues with the 2680? All non -L Haswells and non-2680 Broadwells work fine.
|
# ¿ May 19, 2016 14:09 |
|
Twerk from Home posted:4+ year old servers are not worth the electricity they take to run, so there's a lot of refurbishers selling machines even with warranties and support for very cheap. ...oh...oh my god. I'm about to waste some perfectly good money.
|
# ¿ Jun 2, 2016 17:48 |
|
Does anyone else have hard data on the performance differences between 970(Superclocked) to 1080GTX on a 6700k. I don't have to upgrade immediately as DOOM is my most intense game right now and I only have a 1080 monitor, but I did build this skylake machine as soon as I got wind of the 1080, in preparation for VR.
|
# ¿ Jun 5, 2016 23:00 |
|
PerrineClostermann posted:Do you think it'd be fine for me to reuse my Corsair H50 as the cooler? Or should I get something new? I've had it since this computer was built in 2011.
|
# ¿ Jun 6, 2016 05:12 |
|
Potato Salad posted:Rare metals often sourced from locations with endless war, monstrously toxic chemicals with expensive storage and disposal, delicate multi-million-dollar fabrication gas chambers or etchers or whatever else in rows upon rows upon rows... I interned at a plant that built steering racks in highschool. The hermetically sealed cell that built stuff for the (then-super secret, hadn't-been-revealed-yet) Camaro was awesome. I can only imagine the PPE requirements to work in a fab.
|
# ¿ Jun 9, 2016 18:21 |
|
PBCrunch posted:
We had this conversation a few pages ago but that is a "ES" i.e., engineering sample, and board compatibility can be wonky.
|
# ¿ Jun 23, 2016 14:09 |
|
wipeout posted:This is how I felt about it; so I just bought a 6600k. Maybe by the time gains (hopefully) start to show, it'll be kaby lake time. LOL. I can't say more than that, sorry.
|
# ¿ Jun 29, 2016 15:04 |
|
PBCrunch posted:The "E5-2650v3" I got from ebay seems to work just fine in my Gigabyte X99-UD3P motherboard. The clock speed is down 100 MHz compared to a real 2650v3, and the max turbo frequency is down 200 MHz, but it is still pretty amazing when allowed to perform multi-threaded work. I saw something like 1200% in top when transcoding videos in Handbrake. Even with only two channels of memory, the chip transcodes in less than half the time compared with ye olde X3450 (i5-750 with Hyperthreading). Absolutely, the binning isn't complete until after ES time. Gigabyte and ASRock boards tend to be the best about handling ES for whatever reason. Intel boards don't like them generally because they change the BIOS to be picky after RTM.
|
# ¿ Jul 4, 2016 00:23 |
|
Aquila posted:Two mysterious kernel panics with our new v4 Xeons (E5-2640 v4) in just a month so far, a little bit worrisome. First one was extra special: That'd be me, but that was an issue with the board not detecting CPU0 in a DP arrangement that worked flawlessly with Haswells, the Broadwells are still giving us headaches.. I would not be surprised if little bugs like this continue going forward, Intel is getting sloppy because they know they need to perpetuate the idea of Moore's law and its very obvious they're running out of ideas, fast.
|
# ¿ Jul 6, 2016 14:05 |
|
BIG HEADLINE posted:I'm not sure what to take from this: Yes, DP Skylake lanes e: Coffee Lake looks to be another 14nm design (tock #4?) also yes, it was also a surprise (lol)
|
# ¿ Jul 21, 2016 22:43 |
|
Yeah they must really be having yield issues with that L4 cache... Of course, the most compelling upgrade reason in 3 years and they kill it off. japtor posted:Skylake-X sounds like the usual Xeon EP derived part, so it could be on the CPU...then again there was that big rear end LGA3647 socket shown a while back. It shows Broadwell-E there as "40/28 lanes", how are those laid out, or does it just vary by model/core count? All E5-26XX v4 have 40 lanes/proc, just like Haswell (v3) if that answers your question. SuperDucky fucked around with this message at 23:04 on Jul 21, 2016 |
# ¿ Jul 21, 2016 23:02 |
|
Sir Unimaginative posted:I'm going to end up riding my 2500K into the ground, aren't I. I mean, I thought that, then I got a big boy job, got bored and went to MicroCenter over lunch on a slow Friday and had one of our techs sleeve my psu cables and watercooling hoses and assembled it. x4 M.2 NVMe and USB 3.1 are nice, I guess?
|
# ¿ Jul 22, 2016 20:44 |
|
Eletriarnation posted:My totally based on nothing guess of what all that means (assuming it's not just made up) is that the -X naming does not designate Extreme Edition, it's just their new term for the HEDT package replacing 2011-3. Kaby Lake-X is going to be a 4-core part because they're not actually making a full server line for Kaby Lake. Although at this rate who knows what will actually drop in the next 6 months let alone 18. Its loving chaos in the midrange server market right now.
|
# ¿ Jul 23, 2016 14:25 |
|
Eletriarnation posted:Yeah, it's a total guess and I have no inside information to that part of the industry so anyone who does could well be shaking their head.
|
# ¿ Jul 24, 2016 18:11 |
|
Potato Salad posted:Particularly with Pascal out, this is solid advice. I've been doing some benchmarking with 4 TitanX's on one of our dual E5-2680v4 rigs for a customer and...
|
# ¿ Aug 1, 2016 23:31 |
|
Combat Pretzel posted:I thought the -E series was now -X? Your wish will be granted. Definitely not. -E will remain big-socket workstation ala Skylake-S for the little sockets.
|
# ¿ Aug 24, 2016 01:19 |
|
Paul MaudDib posted:He's right, they renamed Skylake-E to Skylake-X. I assumed the -X was going to be the new "extreme edition" enthusiast chip and that they'd keep a -E "workstation" chip that's pretty much the same thing except with ECC and brand it a Xeon. Guess not, but from reviewing my non-confidential sources, it looks like Kaby-E is still on the table? gently caress's sake, Intel. These days I keep my head buried in -EP stuff as not to get confused. I LIKE TO SMOKE WEE posted:4.5GHz boost clock is impressive, but aren't all Kaby Lake processors going to be limited to 24 PCI-E lanes, or is that just on the lower power packages? (or outdated reporting) No? But you'll have to step up to socket3647(?) for more than 24PCIe. Also, blame your fiscally prudent self, I built my 2500k machine in mid '11 and finally got bored this summer and went skylake. SuperDucky fucked around with this message at 14:31 on Aug 25, 2016 |
# ¿ Aug 25, 2016 14:28 |
|
I LIKE TO SMOKE WEE posted:Isn't socket3647 intended for their server garbage?
|
# ¿ Aug 26, 2016 20:52 |
|
Combat Pretzel posted:Who cares about the socket? That changes way too often. Haha. Like half of this thread self-reports still being on 1150 i5-2500/2600k's. But yeah 5+ years isn't long enough for a socket. e: for content, finally grabbed a 1070 and bothered to overclock my 6700k for the first time last night. With literally no tweaking other than hitting the option that says, "hey, motherboard, don't throttle the bastard when it's idle, and give it whatever juice you think necessary for it to hit 4.6GHz, thanks" I hit 77th percentile on Fire Strike 3d mark. SuperDucky fucked around with this message at 01:28 on Aug 28, 2016 |
# ¿ Aug 28, 2016 01:24 |
|
Otakufag posted:1- For those of you who have upgraded from a 2500k to a new skylake: can you feel noticeable differences when playing recent games or doing other windows stuff? 1. I upgraded when I got back into gaming and about the most strenuous thing I do is battlefront at 1080, natch, so not a lot to comment on, there. 2. i767k. 5GHz is effortless even on decent quality air. 3. I know the argument is always that something better is right around the corner so if you're on the fence, go for it, but Kaby should be hitting general availability around christmas time. Personally, I'd wait now that you're this deep into SkyL's cycle. Obviously, your 25k isn't really harming you at this rate, just hang onto it and see what happens next quarter, you might even be able to grab a bangin' deal on a SkyL when Kaby comes out.
|
# ¿ Aug 31, 2016 18:24 |
|
HalloKitty posted:I'm pretty sure that a consequence of them being more cautious with Xeons, so they do the desktop runs of the new chips first, to make sure all the bugs are ironed out, before going big. JFC. That's what I get for believing my Intel rep on availability. We're expecting engineering samples in Q4 though. HalloKitty posted:I'm pretty sure that a consequence of them being more cautious with Xeons, so they do the desktop runs of the new chips first, to make sure all the bugs are ironed out, before going big. This is absolutely the case. Remember, Intel may be making some coin and clout with enthusiasts these days by virtue of not being utter poo poo compared to AMD, but they know where their bread and butter is--the datacenter. Though, if they don't get off their asses and do something new, soon, there are a couple of players who would love to dethrone them, there (ARM)
|
# ¿ Aug 31, 2016 19:04 |
|
I've got a couple of DP socket 2011 supermicros lying around if someone wants to throw me a few bones for them.
|
# ¿ Sep 2, 2016 14:53 |
|
priznat posted:Got an Intel SDK in with LGA 3647 processors Yeah its' something else, ain't it? I can't imagine standards like PICMG 1.3 are going to be able to embark them without some creative daugherboards or something. Paul MaudDib posted:What were you looking for for these?
|
# ¿ Sep 12, 2016 18:12 |
|
LiquidRain posted:On that subject would it be worth it to run an NVMe SSD on a Z97 board on the south bridge PCIe lanes? (PCIe 2.0) gently caress's sake. No. Not unless you're planning on buying a drive tomorrow and upgrading to a sky/kaby in under 6 months but even then, info on the Samsung 960 is already trickling out.
|
# ¿ Sep 14, 2016 16:15 |
|
My customers in the embedded segment are super stoked about the triple 4k outputs provided by the proc on Sky/Kaby without having to add a dedicated GPU FWIW.
|
# ¿ Oct 3, 2016 14:15 |
|
Potato Salad posted:But will any mobo manufacturer put three hdmi/dp outputs on their poo poo? We do 1 DP external and 2 DVI-D on the board itself. My Asrock z77 extreme4 had HDMI, VGA and DVI.
|
# ¿ Oct 3, 2016 17:53 |
|
Paul MaudDib posted:
I don't disagree, but we had to do it that way due to compatibility restrictions. We have a chassis with a built-in monitor on the faceplate that has to have DVI, also, I'd be scared of surface mounting a DP connector on the board, its too tall, not enough support. e: you couldn't believe the hell we've gotten for not having VGA on this Skylake-S board. The very reason we never went to HDMI was that it didn't have a collar/lock on the connector. DP did, so we bit the bullet. PLUS, iirc, there's no way to pull an analog/vga signal out of the HD/P530 graphics on Skylake-S without an offboard BMC or graphics controller, which there was absolutely no room for on the already slam-full PICMG 1.3 board. SuperDucky fucked around with this message at 01:24 on Oct 4, 2016 |
# ¿ Oct 4, 2016 01:20 |
|
blowfish posted:
|
# ¿ Oct 5, 2016 17:15 |
|
canyoneer posted:
We'll still build Williamette P4 boards for someone that wants to order a min. quantity. Everything else we had to EOL because we just can't get some of the IC's necessary to build 'em.
|
# ¿ Oct 5, 2016 19:07 |
|
Picked this bad boy out of the trash at work today. AFIK, world's only dual socket 1366, x16, x8, x8, triple channel ECC DDR3 microATX mobo, complete with EC5549s and 24GB of RAM. Riser card is not our design, but was also in a "to recycle" chassis. This son' bitch gets HOT in a 2u even with the 3, 102CFM chassis fans screaming.
|
# ¿ Oct 7, 2016 02:09 |
|
SlayVus posted:Your best bet would probably be to pick up some aftermarket coolers. I think Noctua makes server sinks that would fit that socket design. Those dinky OEM sinks are just really not enough. Agreed, the stock sinks are loud as all gently caress. Our new OEM supplier on our fully in-house form factor is dynatron, they have a socket2011 low-profile blower that works really well, I need to see if our rep can ship me some 1366 "samples" mwuhahaha. edit of the edit: eh, this info is too fresh to be that specific about, Intel prob wouldn't like it SuperDucky fucked around with this message at 04:45 on Oct 13, 2016 |
# ¿ Oct 13, 2016 04:40 |
|
Rexxed posted:I recently put a couple of Hyper 212 EVO coolers on a dual socket 1366 board. The support on the back of the motherboard was permanently affixed (at least the LGA retention bracket was also secured to it so it'd have been really hard to remove and replace both) so I used some M3 10mm standoffs instead of the hyper 212's stock standoffs to get them mounted. The 212 is big but I moved the motherboard from a 1U rackmount case to an ATX case so it worked for my situation. Good to know, that's the same way the retainers are on mine as well. I'll have to stick with a blower style cooler, though, this board is staying in that 2U because of the 8, 3.5" hot-swaps.
|
# ¿ Oct 13, 2016 13:25 |
|
|
# ¿ Apr 24, 2024 22:04 |
|
priznat posted:We have one of the Skylake/Purley Intel SDKs at work and got sent the upgrade to Beta CPUs and when I swapped them out I was like what the fuuuuuck because the chips were so huge under the heatsinks. Also the heatsinks snap right on to the CPU and you have to carefully pry them off which is ridiculous. You're breaking nda, friend, just fyi.
|
# ¿ Oct 16, 2016 22:34 |