|
Cascade Lake-X is probably going to be mesh architecture, not a ringbus chip, right?
|
#
?
Apr 22, 2018 05:10
|
|
|
|
| # ? Jun 11, 2024 20:36 |
|
|
Is the 'K' designation supposed to stand for something? Seems like it came out so long ago I forgot if it was supposed to mean something.
|
|
#
?
Apr 22, 2018 05:50
|
|
|
Shaocaholica posted:Is the 'K' designation supposed to stand for something? Seems like it came out so long ago I forgot if it was supposed to mean something. Overclockable
|
|
#
?
Apr 22, 2018 05:58
|
|
|
VostokProgram posted:Overclockable No I mean literally. Unless its supposed to be the 'k' in overclockable then thats hella dumb.
|
|
#
?
Apr 22, 2018 06:03
|
|
|
Unlocked frequency multiplier
|
|
#
?
Apr 22, 2018 06:05
|
|
|
Shaocaholica posted:Is the 'K' designation supposed to stand for something? Seems like it came out so long ago I forgot if it was supposed to mean something. Shaocaholica posted:No I mean literally. Unless its supposed to be the 'k' in overclockable then thats hella dumb. The U for ultra low power/ultra high energy efficient happened before K was chosen for unlocked multipliers. U came in with Core 2 Duo, where X was still used for "extreme performance", K comes in with Sandy Bridge Core generation.
|
|
#
?
Apr 22, 2018 06:08
|
|
|
Shaocaholica posted:No I mean literally. Unless its supposed to be the 'k' in overclockable then thats hella dumb.  I don't believe it was ever meant to stand for a word, it's just a designation. O can't be used for anything because it looks too close to 0
|
|
#
?
Apr 22, 2018 06:17
|
|
|
Heh. I mean these things don’t come out of nowhere. A bunch of people probably sat in a room multiple times and sent many an email before deciding to go with ‘k’.
|
|
#
?
Apr 22, 2018 06:22
|
|
|
Shaocaholica posted:Heh. I mean these things don’t come out of nowhere. A bunch of people probably sat in a room multiple times and sent many an email before deciding to go with ‘k’. k
|
|
#
?
Apr 22, 2018 06:28
|
|
|
I like to think it's a reference to Honda K series engines that were popular in riced-out Civics.
|
|
#
?
Apr 22, 2018 08:33
|
|
|
Laslow posted:I like to think it's a reference to Honda K series engines that were popular in riced-out Civics. OK now AMD needs to make a JZ-series CPU that can handle 1000W
|
|
#
?
Apr 22, 2018 21:17
|
|
|
probably no 10nm Intel products shipping this year... 2019 will be interesting if Zen2 on 7nm releases as scheduled https://twitter.com/david_schor/status/988107149036478464 https://twitter.com/TMFChipFool/status/987286420132958208
|
|
#
?
Apr 22, 2018 23:37
|
|
|
I have a coffee Lake i5 (6-core) and am thinking about swapping it out with a Coffee Lake i7 (twelve logical cores.) I started doing some 3D rendering and I think the extra cores would help. Since the i7 is hyperthreaded, should I still get double the performance (FLOPS) versus my existing six-core (no hyperthreading) i5? Also, all I need to do is this, right?: 0. Set UEFI to default (no overclock) 1. Shut down and unplug computer 2. Unbolt cold plate and clean off thermal grease with lint-free alcohol wipe 3. Attach grounding strap 4. Remove i5 and clean off thermal grease, store in original box 5. Install i7 6. Replace thermal grease 7. Re-bolt cold plate (I think the tension is set by springs) 8. Detach grounding strap 9. Reboot computer into UEFI and check settings 10. Do thermal and OC tests, adjust overclocking EDIT: the i7 only appears to have modestly more performance than the i5. https://www.cpubenchmark.net/compare/Intel-i5-8600K-vs-Intel-i7-8700K/3100vs3098 Three-Phase fucked around with this message at 14:27 on Apr 23, 2018 |
|
#
?
Apr 23, 2018 14:15
|
|
|
Intel HT typically gives around 15-20% more performance, iirc.
|
|
#
?
Apr 23, 2018 14:36
|
|
|
Three-Phase posted:Since the i7 is hyperthreaded, should I still get double the performance (FLOPS) versus my existing six-core (no hyperthreading) i5? Processors since the Pentium 1 have multiple redundant computation units ("pipelines") per core for a lot of common tasks, such that if you have (for example) a bunch of non-interdependent integer operations in a row in the same thread, the instruction decoding unit can dispatch them to multiple pipelines to run concurrently instead of having to run them all in series on the same pipeline. If I recall correctly, all HT does is allow the decoder to consider instructions from two threads at once when dispatching instructions to pipelines instead of having to clear out one thread entirely and switch to another. This means that pipelines are less likely to sit empty and looks like two logical processors to the OS and higher levels (the user), but you still have the same number and type of pipelines overall that you do without HT. Thus, like the preceding post says, performance gain can be significant but nowhere near that of just having twice the cores even in the best-case scenario where you're working two threads that mostly have very different mixes of instructions and they can keep different functional units busy. Eletriarnation fucked around with this message at 14:51 on Apr 23, 2018 |
|
#
?
Apr 23, 2018 14:47
|
|
|
Additionally, HT reduces the penalty for context switching by, effectively, doubling the number of thread contexts that can be "loaded" at one time on the processor. Again, its about utilization of the silicon, not additional computational abilities.
|
|
#
?
Apr 23, 2018 15:24
|
|
|
PerrineClostermann posted:Intel HT typically gives around 15-20% more performance, iirc. For workloads that would scale well with more real cores, HT is worth about 30%. It's really quite decent.
|
|
#
?
Apr 23, 2018 17:08
|
|
|
The problem is you have a processor listed as 6-core, and a processor listed as 12-core. “That’s 200% of the performance.” As Atari said back in the 90s about the “64-bit” Jaguar system: “Do the math.” But instead it’s more like 130% depending on the application. To be fair for for a 30% or so price difference it does make sense. But at this point I don’t see swapping from an i5 to an i7 to be worth it.
|
|
#
?
Apr 23, 2018 17:14
|
|
|
Three-Phase posted:I have a coffee Lake i5 (6-core) and am thinking about swapping it out with a Coffee Lake i7 (twelve logical cores.) I started doing some 3D rendering and I think the extra cores would help. you're not going to get double the performance. it's a decent upgrade in some applications (even in some games) and I'd say to get the i7 if you didn't already have a perfectly good processor, but I don't think it's worth your money unless you can sell the i5 for a really dece price. really if you're going to be doing 3d rendering long-haul save up the money for one of the high core count processors or somet, it'll offer a noticeable difference also grounding strap is pussy poo poo and I've never needed one. never broke a piece of computer hardware by handling it
|
|
#
?
Apr 23, 2018 17:21
|
|
|
Generic Monk posted:also grounding strap is pussy poo poo and I've never needed one. never broke a piece of computer hardware by handling it ESD is real, latent failure is real, it's not debatable, there's decades of research and proof. Ground yourself.
|
|
#
?
Apr 23, 2018 17:54
|
|
|
I get ready to handle sensitive electronics by repeatedly going down a plastic slide, and then shuffling over the rug in socks.
|
|
#
?
Apr 23, 2018 17:54
|
|
|
Dr. Fishopolis posted:ESD is real, latent failure is real, it's not debatable, there's decades of research and proof. Ground yourself. i touch a grounded object before i start loving about inside my computer, and I'd probably consider using a strap if i was repairing the things for a living fwiw
|
|
#
?
Apr 23, 2018 18:03
|
|
|
I get zapped by work hardware every day. Or it zaps itself and resets because somebody walked in the room. Dev boards have interesting gremlins 
|
|
#
?
Apr 23, 2018 18:03
|
|
|
Twerk from Home posted:I get ready to handle sensitive electronics by repeatedly going down a plastic slide, and then shuffling over the rug in socks. i put ram in by holding it between 2 tesla coils
|
|
#
?
Apr 23, 2018 18:08
|
|
|
True ballers have heel straps and esd floors 
|
|
#
?
Apr 23, 2018 18:36
|
|
|
Only for gaming at 1080p 144 refresh and pair with a 1070, should I get a 8400, 2600x or a 8700-non k? I'm afraid the 8400 will bottleneck the next round of Nvidia cards if I decide to upgrade the gpu in a year, also afraid Zen 2 will arrive late / not live up to the hype for a later cpu upgrade, and finally afraid Intel will announce hyperthreading to the I5 lineup making the I7 unnecessary for gaming.
|
|
#
?
Apr 23, 2018 18:58
|
|
|
Did this thread discuss the "AVX-512 causes thermal throttling" articles going around? Well, I wish they were that eloquent. "By how much does AVX-512 slow down your CPU? A first experiment." is a response to Cloudflare's On the dangers of Intel's frequency scaling. Both of them are trying to analyze the same thing, both stay an arm's length away from just showing what's going on. The mechanism is really simple. If you're running a lot of wide register instructions, your frequency will be cut down. This can be measured in a lot of different ways, but just a count of the GV transitions would be sufficient to prove out the theory. Nobody goes that "far," they all show Gbps throughput and a benchmark with a single value of "1.48s" across 3 attempted workloads. It's maddening to see folks who should have all the right tools who won't touch the actual theory with a 10 foot pole. Why does everyone suck at this analysis?
|
|
#
?
Apr 23, 2018 21:01
|
|
|
It's also worth noting that Xeon-W is different silicon from the Skylake-SP that Cloudflare is testing on, and clocks down much less on AVX-512 (per InstLatx64). Also, is this processor-wide, or could Cloudflare have improved their results by pinning the OpenSSL processes and the application workload to separate cores?
|
|
#
?
Apr 23, 2018 21:22
|
|
|
Otakufag posted:Only for gaming at 1080p 144 refresh and pair with a 1070, should I get a 8400, 2600x or a 8700-non k? I'm afraid the 8400 will bottleneck the next round of Nvidia cards if I decide to upgrade the gpu in a year, also afraid Zen 2 will arrive late / not live up to the hype for a later cpu upgrade, and finally afraid Intel will announce hyperthreading to the I5 lineup making the I7 unnecessary for gaming. I'd do the 8700 but the 8400 is probably the best bang for buck. I don't think you'll see anything close to a true bottleneck with an 8400 and I think the gains will be mostly marginal. But for +$100 on a cpu you can reasonably expect to use for 5+ years I'd say spend it now if it were my money. Lockback fucked around with this message at 21:33 on Apr 23, 2018 |
|
#
?
Apr 23, 2018 21:30
|
|
|
Otakufag posted:Only for gaming at 1080p 144 refresh and pair with a 1070, should I get a 8400, 2600x or a 8700-non k? I'm afraid the 8400 will bottleneck the next round of Nvidia cards if I decide to upgrade the gpu in a year, also afraid Zen 2 will arrive late / not live up to the hype for a later cpu upgrade, and finally afraid Intel will announce hyperthreading to the I5 lineup making the I7 unnecessary for gaming.
|
|
#
?
Apr 23, 2018 21:35
|
|
|
Paul MaudDib posted:It's also worth noting that Xeon-W is different silicon from the Skylake-SP that Cloudflare is testing on, and clocks down much less on AVX-512 (per InstLatx64). My point is "clocks down" is a giant fuckoff DVFS transition that's tracked by 18 on-chip resources and the host OS, why don't people post that number? Is it misunderstanding the mechanism at play or just laziness?
|
|
#
?
Apr 23, 2018 21:38
|
|
|
Laslow posted:I doubt we'll see HT in i5's any time soon. It's really the only thing that differentiates i5's and i7's. It's only used in Pentium's and i3's to make their dual core SKU's acceptable. HT as an i7 thing is a heinous profit center for Intel that AMD is pointedly aiming sharp sticks at. They can continue to get away with it if they price i5 aggressively enough and i7 continues to have significant gaming performance gap with AMD's flagship, but at some point they're going to have to transition to "i7 has more cores".
|
|
#
?
Apr 23, 2018 21:39
|
|
|
Craptacular! posted:HT as an i7 thing is a heinous profit center for Intel that AMD is pointedly aiming sharp sticks at. They can continue to get away with it if they price i5 aggressively enough and i7 continues to have significant gaming performance gap with AMD's flagship, but at some point they're going to have to transition to "i7 has more cores".
|
|
#
?
Apr 23, 2018 22:04
|
|
|
JawnV6 posted:You get one frequency for all of the core slices. So if Core0 is flipping every transistor and C1 thru C3 are ice cold, you can still get stepped down. It's a little more fine grained than that, i.e. you'd expect more throttling if C1 had a nontrivial load. Few people are cpu architects, and software engineers often only barely understand microarchiteture let alone know what dvfs means
|
|
#
?
Apr 24, 2018 02:05
|
|
|
Malcolm XML posted:Few people are cpu architects, and software engineers often only barely understand microarchiteture let alone know what dvfs means Paul MaudDib posted:It's also worth noting that Xeon-W is different silicon from the Skylake-SP that Cloudflare is testing on And 'dvfs' was my approachable phrasing for geyserville...
|
|
#
?
Apr 24, 2018 02:58
|
|
|
JawnV6 posted:Is this really a comparison between a workstation CPU with half a ring and the server equivalent?? Nevermind the higher order whiffs, that's just... I don't have words. Does Xeon-W really have half a ring? I thought it (same silicon as SKL-X) both had mesh. What's the implication of that in terms of AVX clockdown and/or whatever test they were doing? I just meant that Xeon-W clocks down much less under load, so since the second test was done on Xeon-W silicon it would show less impact. What does geyserville mean? DVFS actually made more sense there.
|
|
#
?
Apr 24, 2018 03:47
|
|
|
wrong thread
Proud Christian Mom fucked around with this message at 04:40 on Apr 24, 2018 |
|
#
?
Apr 24, 2018 04:34
|
|
|
Proud Christian Mom posted:https://apnews.com/a58d824cd3304caba36f9c82d67b5801?utm_campaign=SocialFlow&utm_source=Twitter&utm_medium=APWestRegion AMD's stock price is totally going to bump off of this
|
|
#
?
Apr 24, 2018 04:36
|
|
|
Paul MaudDib posted:What does geyserville mean? DVFS actually made more sense there.
|
|
#
?
Apr 24, 2018 05:09
|
|
|
|
| # ? Jun 11, 2024 20:36 |
|
|
JawnV6 posted:I was a little cavalier reading earlier, missed the implication of: Reminder most software engineers have no idea that cpus are vastly more complicated than what they are taught. Most think that the in order scalar cpus of the 80s are basically it. One of the most famous answers on stack overflow explains saturation branch prediction. In 2008. In 2018 there's a tiny perceptron on amd cpus doing that. Now incorporate multiple caching hierarchies, NUMA on chip with the ring bus/mesh, NUMA off chip, cache coherence protocols, out of order execution with functional units, DVFS and other modern cpu features and Joe Javascript's head explodes. Now I am surprised that cloudflare didn't bother to read the manual on avx-512 but not that surprised. They are a golang shop, a language created by people who stopped looking at improvements after the 80s. the principal of least surprise says that AVX-512 shouldn't cause massive frequency changes since it's billed as just another ISA extension though
|
|
#
?
Apr 24, 2018 14:35
|
|