|
calusari posted:Can anyone please link or explain (in simple terms) how to undervolt a 12700k. Step 1: Give it less volts.
|
#
?
Jun 28, 2022 00:05
|
|
|
|
| # ? May 30, 2024 15:15 |
|
|
Went from 3900x to 12700k. Intel is FAR better for various reasons. I am using a DAW studio type computer and Intel is raping AMD in latency out of the box. There are many variables but still, Intel rules the roost for this.
|
|
#
?
Jun 29, 2022 01:28
|
|
|
wow, great post, good to know that intel is forcibly inserting their penis into amd.
|
|
#
?
Jun 29, 2022 01:46
|
|
|
Winifred Madgers posted:Step 1: Give it less volts. To OP wondering about CPU undervolting: it's... kinda tricky. In days of yore, when I was undervolting an AMD FX-8320, it literally involved stepping down the core voltage by small increments and testing until I found a point where it became unstable, then nudging the voltage up slightly, testing again to see if it was stable, and then hopefully calling it a day. This approach may still work on a modern CPU - hop into the BIOS, see where your core voltages lie, and start gently adjusting voltages and seeing what happens. redeyes posted:Went from 3900x to 12700k. Intel is FAR better for various reasons. I am using a DAW studio type computer and Intel is raping AMD in latency out of the box. There are many variables but still, Intel rules the roost for this. Jesus, goon, I get you're excited, but dial it back. Hasturtium fucked around with this message at 03:49 on Jun 29, 2022 |
|
#
?
Jun 29, 2022 03:46
|
|
|
Performance might be stored in the balls, as per the thread title, but come on dude there's a line it's 2022 be cool
|
|
#
?
Jun 29, 2022 04:09
|
|
|
quote:
Im excited to have a good working rig without babysitting the loving thing. Dial back what? Do you understand what a DAW is? Do you measure overall system latency under load conditions? If not, why even comment? (USER WAS PUT ON PROBATION FOR THIS POST)
|
|
#
?
Jun 29, 2022 04:41
|
|
|
redeyes posted:Im excited to have a good working rig without babysitting the loving thing. Dial back what? Do you understand what a DAW is? Do you measure overall system latency under load conditions? If not, why even comment? You could tell us these sorts of things without describing it in terms of sexual assault. This ain't Xbox Live.
|
|
#
?
Jun 29, 2022 05:03
|
|
|
Sorry but finding something that simplifies or eases your everyday workflow is sex talk worthy
|
|
#
?
Jun 29, 2022 05:12
|
|
|
it was the raping, you dolt
|
|
#
?
Jun 29, 2022 05:19
|
|
|
jesus loving christ
|
|
#
?
Jun 29, 2022 05:57
|
|
|
Trolling Thunder posted:it was the raping, you dolt Ah yeah next time I’ll actually read it better
|
|
#
?
Jun 29, 2022 06:22
|
|
|
To cleanse the timeline a bit: https://twitter.com/VideoCardz/status/1542026339707981827 https://twitter.com/9550pro/status/1542003665975709696 Was kinda disappointed with those single core scores at first given the big bump in cache (from 44MB to 68MB) and clocks, but CPU-Z doesn't seem to care about L3 much. The 5800X3D scores lower than the straight 5800X in those tests. So whomst know yet!!
|
|
#
?
Jun 29, 2022 07:25
|
|
|
From what I've read there doesn't seem to be substantial single-threaded improvement over Alder Lake, with the gains going to multi-threaded as you'd expect from adding more E-cores. Will that change by launch time, or is any single-threaded performance gain just going to come from higher clocks?
|
|
#
?
Jun 29, 2022 15:03
|
|
|
I'm curious about the power considering how much Alder Lake needs to hit the last hundreds of mhz. Maybe they'd at least run the E-cores at appropriate voltage as someone mentioned before.
|
|
#
?
Jun 29, 2022 19:01
|
|
|
mobby_6kl posted:I'm curious about the power considering how much Alder Lake needs to hit the last hundreds of mhz. Maybe they'd at least run the E-cores at appropriate voltage as someone mentioned before. I think someone here (Paul?) confirmed that the little cores are running on the same 12V line as the big ones. Since the little cores aren’t operating at an efficient voltage, Intel decided to juice the clocks to get what performance they could out of them, and with Raptor Lake they’re literally doubling down on the strategy. In the future - and especially for mobile-targeted chips - I’d expect Intel to adopt BIG.little with appropriate voltages to suit the strengths of each, but that will likely require a socket change to accomplish.
|
|
#
?
Jun 29, 2022 19:56
|
|
|
Hasturtium posted:I think someone here (Paul?) confirmed that the little cores are running on the same 12V line as the big ones. Since the little cores aren’t operating at an efficient voltage, Intel decided to juice the clocks to get what performance they could out of them, and with Raptor Lake they’re literally doubling down on the strategy. In the future - and especially for mobile-targeted chips - I’d expect Intel to adopt BIG.little with appropriate voltages to suit the strengths of each, but that will likely require a socket change to accomplish. yeah I'd read that but taking a look at it further, while they are supplied from the same rail, and while Shamino does note that the e-cores definitely result in voltage droop as you enable them (so they're shared), it's described as a "partial FIVR" (partial fully, lol), so potentially the small core cluster could reduce its own voltage independently (despite being fed off the main rail). I haven't really read anything exhaustive on how it works yet, just trying to piece together some articles. Still though FIVR is a linear regulator so the more you drop, the more you dissipate. It's better than switching transistors with it, but it's still pulling power and turns it to heat, etc, it's basically a big variable resistor. There is a per-core voltage control mode too... and potentially you could reduce the allowed multiplier on the e-cores to push them a little farther into the efficiency range. But yeah maybe they just are clocking them too high to win benchmarks. Alder lake in particular, the e-cores may not end up being worth it, because as Shamino notes, if you disable the e-cores you you can "yeet the ring multiplier" as he puts it. He's getting 5 GHz, I assume with LN2, and and you can supposedly get to like 42x-44x on ambient. Only 8 e-cores potentially isn't enough to make it worth it - yeah it'll be a bit faster in parallel stuff probably, but games/etc still favor punchy cores over more cores. https://www.anandtech.com/show/17206/the-msi-meg-z690-unify-motherboard-review-ddr5/8 https://www.tomshardware.com/how-to/intel-alder-lake-cpu-overclock-guide https://www.overclock.net/threads/maximus-z690-and-alder-lake-modern-cpus-require-modern-overclocking-solutions.1794893/
|
|
#
?
Jun 30, 2022 07:06
|
|
|
I think the capitalization is big.LITTLE, for whatever reason.
|
|
#
?
Jun 30, 2022 07:06
|
|
|
I assume the e-cores are much more efficient on mobile processors where the voltage will be saner and the clocks lower?
|
|
#
?
Jun 30, 2022 07:50
|
|
|
Paul MaudDib posted:yeah I'd read that but taking a look at it further, while they are supplied from the same rail, and while Shamino does note that the e-cores definitely result in voltage droop as you enable them (so they're shared), it's described as a "partial FIVR" (partial fully, lol), so potentially the small core cluster could reduce its own voltage independently (despite being fed off the main rail). I haven't really read anything exhaustive on how it works yet, just trying to piece together some articles. lol no, lmao, how on earth did you ever convince yourself Intel's FIVR technology is linear regulators https://ieeexplore.ieee.org/document/6803344 posted:These 140 MHz multi-phase buck regulators are integrated into the 22nm processor die, and feature up to 80 MHz unity gain bandwidth, non-magnetic package trace inductors and on-die MIM capacitors. The "partial FIVR" thing appears to mean that unlike predecessor chip Tiger Lake, where most supply rails including CPU cores were FIVR, in AL only uncore is powered by FIVR. Uncore is Intel speak for things like last level cache, ring bus or mesh interconnect, and so forth. Finally, the voltage supplied from motherboard to FIVR regulators is fixed and relatively high - 1.8V, iirc. This is because, once again, it's not linear.
|
|
#
?
Jun 30, 2022 07:52
|
|
|
Rinkles posted:I think the capitalization is big.LITTLE, for whatever reason. I think that one came out of ARM's camp and from what I've heard from people inside there they really do like loving around with naming things. For example, the Advanced RISC Machine's Architecture Reference Manual is published by a company called ARM, making it the ARM ARM ARM.
|
|
#
?
Jun 30, 2022 10:37
|
|
|
Kazinsal posted:I think that one came out of ARM's camp and from what I've heard from people inside there they really do like loving around with naming things. For example, the Advanced RISC Machine's Architecture Reference Manual is published by a company called ARM, making it the ARM ARM ARM. One of the early popular variants was called StrongARM.
|
|
#
?
Jun 30, 2022 12:07
|
|
|
Arivia posted:One of the early popular variants was called StrongARM. That one was DEC's fault. big.LITTLE, on the other hand, is definitely a trademark created by ARM Holdings.
|
|
#
?
Jun 30, 2022 12:20
|
|
|
Paul MaudDib posted:yeah I'd read that but taking a look at it further, while they are supplied from the same rail, and while Shamino does note that the e-cores definitely result in voltage droop as you enable them (so they're shared), it's described as a "partial FIVR" (partial fully, lol), so potentially the small core cluster could reduce its own voltage independently (despite being fed off the main rail). I haven't really read anything exhaustive on how it works yet, just trying to piece together some articles. Saying the e-cores "may not be worth it" is pretty weird. The multicore performance achieved by alder lake would have been impossible without them, and that was the design objective. Intel could not have made a halo product without them. Yes, they could claim to have the fastest gaming CPU again with just 8+0, but gaming was/is not the only aim.
|
|
#
?
Jun 30, 2022 12:45
|
|
|
Paul MaudDib posted:yeah I'd read that but taking a look at it further, while they are supplied from the same rail, and while Shamino does note that the e-cores definitely result in voltage droop as you enable them (so they're shared), it's described as a "partial FIVR" (partial fully, lol), so potentially the small core cluster could reduce its own voltage independently (despite being fed off the main rail). I haven't really read anything exhaustive on how it works yet, just trying to piece together some articles. I have the e-cores active and the ring bus is running happily at 42x. I think 42x is fairly easy even with all cores on? Not sure about 44x.
|
|
#
?
Jun 30, 2022 17:11
|
|
|
Are there any desktop chips coming like the mobile -U chips that have more efficiency cores than performance cores? It's weird to me that most of the desktop lineup has no E-cores, and they're reserved for only the higher end parts at huge TDPs. On the mobile side, I see a whole line-up of things like the the Pentium 8505 and even lower end Celerons packing 5 cores now, and then by the time you're in the mid-range i5s the mobile parts have 10 cores, 2P + 8E. Why aren't we seeing mid or low priced desktop chips with a full host of E cores?
|
|
#
?
Jul 1, 2022 18:31
|
|
|
calusari posted:Can anyone please link or explain (in simple terms) how to undervolt a 12700k. Undervolting or underclocking your CPU is entirely managed by your motherboard. If you have your motherboard's model number on hand, we can help look up your manual and steer you towards the settings. Out of curiosity, what precisely are you trying to accomplish? Are you trying to preserve your processor by taking a little bit of thermal stress off of it? Do you have an overclock in place and you want to take the voltage down a little bit, or do you have running on the stock clock and you just want it cooler regardless? I ask because it is also possible and perhaps more straightforward/stable to underclock the CPU--or even do a blend of the two.
|
|
#
?
Jul 4, 2022 09:42
|
|
|
Twerk from Home posted:Are there any desktop chips coming like the mobile -U chips that have more efficiency cores than performance cores? It's weird to me that most of the desktop lineup has no E-cores, and they're reserved for only the higher end parts at huge TDPs. My guess is that Intel has finite fab capacity and is trying to ensure that they have enough of those -U chips to fulfill laptop OEM orders.
|
|
#
?
Jul 5, 2022 16:59
|
|
|
Potato Salad posted:Undervolting or underclocking your CPU is entirely managed by your motherboard. If you have your motherboard's model number on hand, we can help look up your manual and steer you towards the settings. I am reminded from my time in the trenches for a prominent video card and motherboard mfg I had the same weirdo demanded swap out of of an RMA because he was getting instability trying to undervolt the stock yorkfield era cpus because they wanted a cool room. I doubt that's op's need but undervolters are out there.
|
|
#
?
Jul 6, 2022 06:35
|
|
|
https://twitter.com/videocardz/status/1544901642591633408 Intel matching ARM with the three-tier core design, it seems
|
|
#
?
Jul 7, 2022 07:32
|
|
|
Qualcomm is up to 4 tiers now  Though that's supposedly less to do with power efficiency and more to do with maintaining 32bit support when ARM has mostly dropped it from their reference cores, so they threw in a couple of 32bit cores
|
|
#
?
Jul 7, 2022 11:19
|
|
|
Potato Salad posted:Undervolting or underclocking your CPU is entirely managed by your motherboard. If you have your motherboard's model number on hand, we can help look up your manual and steer you towards the settings. I have the Z690I AORUS ULTRA PLUS DDR4 https://www.gigabyte.com/Motherboard/Z690I-AORUS-ULTRA-PLUS-DDR4-rev-10/sp#sp I have tiny ITX case. My workload is only gaming. Playing Cyberpunk 2077 for a couple hours my hottest P-core gets to 89-91C. I'm curious if I can improve those temps a bit. Is this a good starting point? quote:Turn off Enhanced Boost
|
|
#
?
Jul 7, 2022 13:11
|
|
|
That is indeed a good starting place. in addition to that undervolt, you can also drop the clock a few hundred GHz and REALLY save some power Further, you might actually find it necessary to drop the frequency a little bit as you reach lower and lower undervolts: higher frequencies need higher voltages to remain stable--so as you lower the voltage, you may find that you might also need to lower the frequency a bit as well Potato Salad fucked around with this message at 13:49 on Jul 7, 2022 |
|
#
?
Jul 7, 2022 13:44
|
|
|
I'm trying to run Intel XTU for overclocking in Windows 11. It tells me I have incompatibilities either in the BIOS or in windows but it isn't specific. Anyone know off hand what things to change?
|
|
#
?
Jul 7, 2022 14:36
|
|
|
redeyes posted:I'm trying to run Intel XTU for overclocking in Windows 11. It tells me I have incompatibilities either in the BIOS or in windows but it isn't specific. Anyone know off hand what things to change? Do you have HyperV enabled?
|
|
#
?
Jul 7, 2022 15:26
|
|
|
BurritoJustice posted:Do you have HyperV enabled? Nope, just checked.
|
|
#
?
Jul 7, 2022 15:55
|
|
|
Cygni posted:https://twitter.com/videocardz/status/1544901642591633408 gently caress everything, we're doing five-tier core design!
|
|
#
?
Jul 12, 2022 18:07
|
|
|
From a qualification sample (there's no change in performance expected from this to retail)  and in case you were hoping for less power consumption, lmao:  🔥🔥🔥🔥🔥
|
|
#
?
Jul 18, 2022 03:25
|
|
|
I've seen my 12900K draw 175-180W in Horizon: Zero Dawn according to MSI Afterburner's CPU wattage meter on its overlay, wonder why.
|
|
#
?
Jul 18, 2022 04:21
|
|
|
Dr. Video Games 0031 posted:From a qualification sample (there's no change in performance expected from this to retail) Those are decent gains right?
|
|
#
?
Jul 18, 2022 04:26
|
|
|
|
| # ? May 30, 2024 15:15 |
|
|
I'm guessing the cache is showing in the min FPS gains, there's way more of it and the uncore can now run at 4.7 with the E-cores enabled.
|
|
#
?
Jul 18, 2022 04:29
|
|