|
Zotix posted:If I wanted to start piecing together stuff for a 9900k what would be the ram I'd want to look at? I have an old ivy bridge, so I know I'd need to upgrade to ddr4. I'd want some good ram, as I'd be overclocking the cpu. I just haven't done a build in 5 years so the specifics to current chips I'm a bit out of date on. 3000/3200 is the sweet spot right $/performance right now.
|
# ? Sep 19, 2018 20:20 |
|
|
# ? Apr 26, 2024 14:33 |
|
2666 is the new base RAM speed of the 8th and 9th gen CPUs, but if it's what you already have, use it. We're suggesting 3000 and 3200 in the building thread, with 34-3600 only suggested if the kit is on a board's QVL sheet and priced right.
|
# ? Sep 20, 2018 02:15 |
|
I don't know if they're doing it or I'm wasting my time but I don't know if there's super high latency DDR4 or if it matters. Like 2666-19 or something high like that, I don't want to lead someone wrong, but maybe it's not like DDR3? I remember 1600-15 kits for that that were slower than normal 1333-8 kits.
|
# ? Sep 20, 2018 02:58 |
|
Cygni posted:Cannon Lake is never launching on desktop, so you might be waitin' a while! I remember a few months ago Ashraf Eassa was suggesting Intel should scrap 10nm and move on to 7nm, is there any credence to that or was it just conjecture?
|
# ? Sep 20, 2018 06:20 |
|
So where do they go once they hit 7nm?
|
# ? Sep 20, 2018 06:34 |
|
In the short term at the high end, some combination of MCM, bigger dies, lower profits, and/or higher prices.
|
# ? Sep 20, 2018 06:39 |
|
Aeka 2.0 posted:So where do they go once they hit 7nm? Probably the bar to celebrate
|
# ? Sep 20, 2018 06:44 |
|
I think I remember reading that 5nm is a hard wall thermoelectrically with existing technology, and the only thing they can really do short of some revolutionary tech and/or exotic materials is just keep adding cores and cache. This article describes 3nm and smaller, but it seems to be overly optimistic: https://semiengineering.com/transistor-options-beyond-3nm/ BIG HEADLINE fucked around with this message at 07:11 on Sep 20, 2018 |
# ? Sep 20, 2018 07:08 |
|
I'm really out of the loop on how Intel is doing re: CPU technology and I always thought that Intel was the absolute king but this doesn't seem to be true anymore. What's a good and accessible (I'm not a hardware engineer) article I can read to get myself up to speed on what Intel did wrong and what everyone else did right?
|
# ? Sep 20, 2018 07:37 |
|
Boris Galerkin posted:I'm really out of the loop on how Intel is doing re: CPU technology and I always thought that Intel was the absolute king but this doesn't seem to be true anymore. What's a good and accessible (I'm not a hardware engineer) article I can read to get myself up to speed on what Intel did wrong and what everyone else did right?
|
# ? Sep 20, 2018 07:49 |
|
Llamadeus posted:Here's a non-technical overview of Intel's current woes: https://stratechery.com/2018/intel-and-the-danger-of-integration/ Very cool, thanks for that! I think it’s going to be a very interesting next couple of years in the cpu biz. I expect intel to do a few more frantic acquisitions to try and diversify or improve their portfolio.
|
# ? Sep 20, 2018 08:04 |
|
Wasn't one of the tricks in the bag still was gutting x86 or rather large sections devoted to compatibility?
|
# ? Sep 21, 2018 05:41 |
|
incoherent posted:Wasn't one of the tricks in the bag still was gutting x86 or rather large sections devoted to compatibility? No, that's not really an issue for modern processors. Space is taken up by cache and VLIW style execution units inside each core. If you take a look at a processor image with a block diagram over it you'll see that decode is a very small segment of the die.
|
# ? Sep 21, 2018 11:44 |
|
Is there any old simd hardware they can remove?
|
# ? Sep 21, 2018 22:16 |
|
GRINDCORE MEGGIDO posted:Is there any old simd hardware they can remove? Do you see where I'm going with this?
|
# ? Sep 21, 2018 22:26 |
|
Khorne posted:The core adjacent to the iGPU is consistently 5c-10c lower on certain gens of intel CPUs. Add more iGPUs?
|
# ? Sep 21, 2018 22:28 |
|
More cache please. Also gently caress iGpu
|
# ? Sep 21, 2018 23:22 |
|
Llamadeus posted:Here's a non-technical overview of Intel's current woes: https://stratechery.com/2018/intel-and-the-danger-of-integration/ I'm still kinda stunned at how good this is. Normally there's some inconsequential minor point that I can trash, but there's none of those. There's some things I could add, but the biggest would be Marvell.
|
# ? Sep 21, 2018 23:30 |
|
I mean it works, but seeing a CPU being handled that way makes me cringe. https://www.youtube.com/watch?v=zx641tAZFH0&t=223s
|
# ? Sep 22, 2018 23:22 |
|
A cheap vice is the same cost as a good hammer. I don't believe anyone who says they'll never use a vice again. I bought a good vice and it's over 20 years old now and it's so useful to have for random jobs, it's exactly for holding and pressing stuff you can't with your hands. Edit: Alternatively spend 300 dollars on a CPU, can't afford 30 dollars for a vice. craig588 fucked around with this message at 23:40 on Sep 22, 2018 |
# ? Sep 22, 2018 23:34 |
|
craig588 posted:A cheap vice is the same cost as a good hammer. I don't believe anyone who says they'll never use a vice again. I bought a good vice and it's over 20 years old now and it's so useful to have for random jobs, it's exactly for holding and pressing stuff you can't with your hands. Yeah, seriously. Can get a tiny vice that clamps to a table. Even a drat c-clamp would be better than that hammer.
|
# ? Sep 23, 2018 00:02 |
|
A cheap vice is a lot shittier than a cheap hammer. The only problem is that guide being 3d printed, that's WTF level poo poo. I would have no qualms about whacking a CPU inside a machined delidder with a hammer.
|
# ? Sep 23, 2018 00:03 |
|
The 3D printed delidder works fine.
|
# ? Sep 23, 2018 00:09 |
|
ACTUALLY, a vice is a bad habit a vise is a clamping device (at least in american english) i'm sorry, there was no reason for this post, it's a very common spelling, i just couldn't help myself
|
# ? Sep 23, 2018 00:31 |
|
Cheap vice is relevant, it's the Intel thread
|
# ? Sep 23, 2018 05:55 |
|
spending the time to 3d print a thing, then wacking a $350 fragile electronics part with a hammer on your garage floor to get a completely unnoticable 4% increased overclock this hobby is so dumb
|
# ? Sep 23, 2018 06:03 |
|
Cygni posted:spending the time to 3d print a thing, then wacking a $350 fragile electronics part with a hammer on your garage floor to get a completely unnoticable 4% increased overclock Thank intel for being cheap assholes.
|
# ? Sep 23, 2018 12:24 |
|
I don’t look at it that way, I like to have lower temps so I can set fans at lower speeds and thus lower noise. That matters a lot more to me than overclocking.
|
# ? Sep 24, 2018 02:08 |
|
Qualcomm claims Apple stole its source code and gave it to Intel https://www.cnbc.com/2018/09/25/qualcomm-accuses-apple-of-giving-its-chip-secrets-to-intel.html
|
# ? Sep 26, 2018 08:21 |
|
3peat posted:Qualcomm claims Apple stole its source code and gave it to Intel https://www.cnbc.com/2018/09/25/qualcomm-accuses-apple-of-giving-its-chip-secrets-to-intel.html Well, if they can prove it, that's certainly one way for Apple to lose its trillion dollar company status.
|
# ? Sep 26, 2018 08:22 |
|
Yeah I’m really interested in this cause saying a company stole ip seems like a big deal whether or not it happened.
|
# ? Sep 26, 2018 10:15 |
|
Intel Tock-Ticks Chipsets Back to 22nm
|
# ? Sep 28, 2018 15:08 |
|
It's chipsets, not the CPUs themselves, right?
|
# ? Sep 28, 2018 15:23 |
|
limaCAT posted:It's chipsets, not the CPUs themselves, right? Because of the issues with the 10nm process, they don't have enough capacity on their 14nm process so they are moving the chipset back to 22nm. They have the spare capacity for the 22nm process so they might as well use it, but it just shows to what degree the issues with the 10nm are screwing all their plans up.
|
# ? Sep 28, 2018 15:26 |
|
This means something like +1 or +2 watts for the platform at idle.
|
# ? Sep 28, 2018 15:30 |
|
It's not a huge deal performance wise, there was a chart showing chipsets on 65nm for a long time after CPUs had surpassed that. I think it was from the 975 to the X79 they stayed on 65nm. The bigger surprise is it was worth doing, remaking anything on a new process isn't cheap.
|
# ? Sep 28, 2018 15:52 |
|
craig588 posted:It's not a huge deal performance wise, there was a chart showing chipsets on 65nm for a long time after CPUs had surpassed that. I think it was from the 975 to the X79 they stayed on 65nm. The bigger surprise is it was worth doing, remaking anything on a new process isn't cheap. California's really on their rear end about standby power utilization though, gotta save that 2w of vampire power
|
# ? Sep 28, 2018 16:27 |
|
mystes posted:Because of the issues with the 10nm process, they don't have enough capacity on their 14nm process so they are moving the chipset back to 22nm. They have the spare capacity for the 22nm process so they might as well use it, but it just shows to what degree the issues with the 10nm are screwing all their plans up. The fact that they massively underestimated demand probably plays the biggest part of this. They knew 10 nm wasn't ready for prime time when they moved chipset production.
|
# ? Sep 28, 2018 17:30 |
|
Paul MaudDib posted:California's really on their rear end about standby power utilization though, gotta save that 2w of vampire power Now imagine 2W x at least 10 million people. That’s 20MW of wasted energy.
|
# ? Sep 28, 2018 17:34 |
|
|
# ? Apr 26, 2024 14:33 |
|
latinotwink1997 posted:Now imagine 2W x at least 10 million people. Thats 20MW of wasted energy. Now imagine it doesn't matter since morons run cryptocurrency miners non-stop to the tune of at least 42TW a year for bitcoin alone: https://www.theguardian.com/technology/2018/jan/17/bitcoin-electricity-usage-huge-climate-cryptocurrency California should ban crypto mining first!
|
# ? Sep 28, 2018 17:59 |