|
"Self," says I, "Maybe you should pop your head in the Intel thread to see if they have a better idea than the rest of the internet as to when general Ice Lake or next-gen graphics availability is before you pull the trigger on that 2019 LG Gram. Sure wouldn't do to stick your brother on old graphics on his birthday present." edit: page snipe yooooo. SwissArmyDruid fucked around with this message at 13:40 on Mar 30, 2019 |
#
?
Mar 30, 2019 13:29
|
|
|
|
| # ? Jun 8, 2024 19:40 |
|
|
Intel has slated products with Ice Lake/Sunny Cove/Gen11/whatever that SoC is called at this point for the holiday season this year so you have a while to wait yet before anything with Kaby Lake-R/Whiskey Lake is superseded.
|
|
#
?
Mar 30, 2019 14:51
|
|
|
a OEM manufacturer published their roadmap for Intel CPUs — it suggests that Elkhart Lake, Comet Lake, Ice Lake U are scheduled for Q2/2020. https://www.cnx-software.com/2019/03/31/intel-elkhart-lake-processors-2020/ has a link to the PDF
|
|
#
?
Apr 1, 2019 16:01
|
|
|
My head hurts when I try to decipher anything sensible from post-CFL Intel roadmaps.
|
|
#
?
Apr 1, 2019 16:21
|
|
|
Intel just launched a huge amount of enterprise stuff. https://www.anandtech.com/show/14155/intels-enterprise-extravaganza-2019-roundup In addition to new NICs, FPGAs, and Xeon D, there is also the full Cascade Lake replacement for Skylake-SP. Heres the new stack:  And the new 9200 series stack, which are two dies on one BGA package:  112 threads, 400 watts. Lordy. https://www.servethehome.com/second-generation-intel-xeon-scalable-sku-list-and-value-analysis/ https://www.servethehome.com/intel-xeon-platinum-9200-formerly-cascade-lake-ap-launched/
|
|
#
?
Apr 2, 2019 19:27
|
|
|
Be aware that the turbo clocks are for a single core. All-core turbos are considerably lower and the binning for which model # can hit what clocks on how many cores is an absolute nightmare where not even the product pricing matches aggregate. Intel releases spec sheets for SOME of their models showing you exactly what clocks its binned for with X # of cores active but they are not good or consistent about disclosing that info.
|
|
#
?
Apr 2, 2019 20:23
|
|
|
God drat I will never get over how harrowing it is to lock down an LGA chip on an Intel board, especially when the board costs as much as the damned chip.
|
|
#
?
Apr 4, 2019 06:07
|
|
|
BIG HEADLINE posted:God drat I will never get over how harrowing it is to lock down an LGA chip on an Intel board, especially when the board costs as much as the damned chip. No risk no reward, friend Now please don’t remind me again, I have to fight the impulse to curl into a ball and weep
|
|
#
?
Apr 4, 2019 06:50
|
|
|
I like the LGA parts, it seems like it would be easier to drop a cpu and bend a pin on a PGA than one of the contacts in an LGA socket. LGA parts seem pretty hard to gently caress up! The sockets do worry me as well though.. especially the giant xeon-sp 3647 ones.
|
|
#
?
Apr 4, 2019 08:56
|
|
|
priznat posted:I like the LGA parts, it seems like it would be easier to drop a cpu and bend a pin on a PGA than one of the contacts in an LGA socket. LGA parts seem pretty hard to gently caress up! It's just that "creak" noise as you bend down the retention arm that inspires that  feeling.
|
|
#
?
Apr 4, 2019 09:01
|
|
|
BIG HEADLINE posted:It's just that "creak" noise as you bend down the retention arm that inspires that Haha oh yeah it’s scary the amount of force that seems to be on that lever. The SPs have hex screws you have to tighten in a specific order to keep the pressure somewhat equal. At least you can’t overtighten on the boards I have seen, they just stop.
|
|
#
?
Apr 4, 2019 09:03
|
|
|
Mmm Socket A.
|
|
#
?
Apr 4, 2019 10:20
|
|
|
Sormus posted:Mmm Socket A. Socket 7/370/A were a nightmare to mount HSFs on, since virtually all of them is clip-mounted by a LOT of force with a flat bladed screwdriver; one slip-up and the mobo is very likely gonna be an expensive paperweight.
|
|
#
?
Apr 4, 2019 11:34
|
|
|
Palladium posted:Socket 7/370/A were a nightmare to mount HSFs on, since virtually all of them is clip-mounted by a LOT of force with a flat bladed screwdriver; one slip-up and the mobo is very likely gonna be an expensive paperweight. If you have a nut driver, you can use that instead, and you won't slip. I have one that's exactly the right size for motherboard stand-offs HalloKitty fucked around with this message at 12:33 on Apr 4, 2019 |
|
#
?
Apr 4, 2019 12:14
|
|
|
I never had a problem with installing heatsinks, but clipped on heatsinks had weight limits. I got significantly worse overclocks once my CPU heatsink exceeded a certain size that all became no problem when I switched to heatsinks that mounted to the motherboard.
|
|
#
?
Apr 4, 2019 14:40
|
|
|
Okay, I figure this is as good a place to ask as any because there's almost a 100% certainty of someone in here using the Corsair PRO series AIOs - in specific, the H150i PRO. I just attached the back bracket to my Aorus Xtreme board, and while I screwed up the first time and noticed I'd mounted the bracket horizontally instead of vertically (so the plastic was impacting the rear part of the socket), so I changed it around, and...there's about a 2mm "gap" between the bottom of the standoff screw and the board itself. Before anyone says it, yes, I'm sure I'm not using the LGA2xxx screws - I'm using the ones for LGA 1151(v2). Just out of morbid curiosity, I unscrewed one of the standoffs and fished out two of the fan washers and lined them up with the hole and *that* made it flush with no movement. But the directions don't tell you to do that. Naturally I'm not too keen on using metal washers to secure the bracket, and it *seems* like the washers are ~1mm in width, so I'd imagine it wouldn't be too difficult to get some hard rubber ~2mm washers at Home Depot. I just don't know why it's got the 'wiggle' it does by default, and I'm certainly not comfortable mounting a water block on a 9900K if there's chance of movement. BIG HEADLINE fucked around with this message at 23:00 on Apr 4, 2019 |
|
#
?
Apr 4, 2019 21:48
|
|
|
Clearly Intel should go back to Slot 1. They can even call it Slot 2 or something.
|
|
#
?
Apr 4, 2019 22:43
|
|
|
Someone did a bunch of early, basic performance measurements of Optane DIMMs https://arxiv.org/pdf/1903.05714.pdf
|
|
#
?
Apr 4, 2019 22:58
|
|
|
27GB/s sequential reads and 9GB/s sequential writes when running as persistent storage, holy moly
|
|
#
?
Apr 4, 2019 23:17
|
|
|
mobby_6kl posted:Clearly Intel should go back to Slot 1. Looks like “pentium(r)!!!”. Mr. Smile Face Hat fucked around with this message at 02:51 on Apr 5, 2019 |
|
#
?
Apr 5, 2019 02:48
|
|
|
Some Optane DIMM pricing is coming out if you're interested in that https://www.tomshardware.com/news/intel-optane-dimm-pricing-performance,39007.html Intel Optane DIMM Pricing: $695 for 128GB, $2595 for 256GB, $7816 for 512GB
|
|
#
?
Apr 7, 2019 20:35
|
|
|
mobby_6kl posted:Clearly Intel should go back to Slot 1. Slot 2 was what Pentium II Xeon used. I kind of miss slotted CPUs too, though.
|
|
#
?
Apr 7, 2019 22:18
|
|
|
Q_res posted:Slot 2 was what Pentium II Xeon used.
|
|
#
?
Apr 8, 2019 02:28
|
|
|
Kinda funny that we’re comin full circle and splitting off I/O and memory back onto a separate die with a more mature process. Bring back the Slot CPU too, damnit!!
|
|
#
?
Apr 8, 2019 03:10
|
|
|
mobile coffee lake refresh numbers were leaked, including comedy TDP ratings for the same old 14nm process https://www.tomshardware.com/news/i5-9300h-i9-9980hk-specs-leaked,39016.html
|
|
#
?
Apr 8, 2019 11:07
|
|
|
Unless they're the same price and it's just for different markets there seems to be no reason to get any of the upgraded processors. Not even a 5% performance difference?
|
|
#
?
Apr 8, 2019 11:19
|
|
|
Cygni posted:27GB/s sequential reads and 9GB/s sequential writes when running as persistent storage, holy moly
|
|
#
?
Apr 8, 2019 13:31
|
|
|
eames posted:mobile coffee lake refresh numbers were leaked, including comedy TDP ratings for the same old 14nm process Hidden takeaway from this is that 6-core gets cheaper since it's no longer top of the consumer stack. That gets people into the multitasking game a lot easier
|
|
#
?
Apr 8, 2019 13:35
|
|
|
eames posted:mobile coffee lake refresh numbers were leaked, including comedy TDP ratings for the same old 14nm process I don't doubt hitting 5GHz single core boost @ 45W TDP. I do doubt if that is gonna hold for any time more than half a millisecond
|
|
#
?
Apr 8, 2019 14:51
|
|
|
Sidesaddle Cavalry posted:Hidden takeaway from this is that 6-core gets cheaper since it's no longer top of the consumer stack. That gets people into the multitasking game a lot easier AMD must be really excited after seeing the contents of these leaks. Unless zen 2 is a complete flop, but if Intel is standing still and the current ryzens are beating Intel's mid range...
|
|
#
?
Apr 8, 2019 15:24
|
|
|
hobbesmaster posted:AMD must be really excited after seeing the contents of these leaks. TSMC is doing so well compared to Intel that the next two Zen designs could be complete flops and AMD would still be competitive
|
|
#
?
Apr 8, 2019 17:43
|
|
|
Palladium posted:I don't doubt hitting 5GHz single core boost @ 45W TDP. Why? assuming stock clocks and voltages there's absolutely no way a single active Skylake+++...+ core even comes close to 45W thermal power no matter how long it runs at 5 GHz. The single core boost numbers are kinda bullshit for other reasons, but not that. The main other reason: operating systems seldom manage to keep things down to a single core for extended periods of time, so most of the time the cpu's going to select the 2-core multiplier even if, on average, you're only using one core.
|
|
#
?
Apr 8, 2019 18:35
|
|
|
Check out all these balllllzz https://www.anandtech.com/show/14182/hands-on-with-the-56core-xeon-platinum-9200-cpu-intels-biggest-cpu-package-ever
|
|
#
?
Apr 8, 2019 19:57
|
|
|
.
sincx fucked around with this message at 05:55 on Mar 23, 2021 |
|
#
?
Apr 8, 2019 22:49
|
|
|
sincx posted:Love to render my soldered-on $25,000 processor useless when my $2,500 motherboard dies. It's a big iron CPU. If your motherboard is dying, you probably arent replacing it in the first place.
|
|
#
?
Apr 8, 2019 23:46
|
|
|
Sending it out to a rework shop would be a minor cost compared to the cpu cost so that would definitely be an option. We get this done a lot with smaller packages and much less expensive parts/boards. I don’t even know what the max package size they can re-ball though. Should ask the tech who interfaces with the companies who do the work for us. He’d probably freak out over a 5000+ ball BGA  Keeps the layout folks busy though!
|
|
#
?
Apr 9, 2019 00:09
|
|
|
Is there an advantage to bga or is it just cheaper/easier to make for a CPU that's never moving?
|
|
#
?
Apr 9, 2019 02:02
|
|
|
ItBreathes posted:Is there an advantage to bga or is it just cheaper/easier to make for a CPU that's never moving? Better signal integrity on high speed I/Os, more current per ground or power pin, better thermal conductivity into board power planes. A 500W cpu probably doesn’t care so much about the last one; relying on the board as a form of heatsink is only for smaller low power chips. It should care a lot about current per pin though. They have to get >500A into and out of that thing! E: also, sheer pin count. Pretty sure the bga pitch there is much finer than the lga pitch in their socketed chips. Besides power they have an awful lot of DRAM channels to deal with. BobHoward fucked around with this message at 02:29 on Apr 9, 2019 |
|
#
?
Apr 9, 2019 02:26
|
|
|
BobHoward posted:Better signal integrity on high speed I/Os, more current per ground or power pin, better thermal conductivity into board power planes. Also removes issues related to poor pin contact on the socket when you have thousands of pins. AFAIK that's the main reason the AMD TR4 socket uses that weird multi-stage mounting system and one of the reasons why the W-3175X (which barely has a socket to mount into) is OEM-only.
|
|
#
?
Apr 9, 2019 02:37
|
|
|
|
| # ? Jun 8, 2024 19:40 |
|
|
Meanwhile in enterprise land Intel has some slightly more exciting stuff, today was "interconnect day." 400G Silicon Photonics https://twitter.com/david_schor/status/1115389479714344966?s=19 CXL, basically Intel's answer to NVLink https://twitter.com/david_schor/status/1115369189177356288?s=19 https://twitter.com/david_schor/status/1115365952034467840?s=19
|
|
#
?
Apr 9, 2019 03:22
|
|
