At 29:40 you said the power lost through the DLVR amounts to roughly 50W (with 48.8W for a 1.405V - 1.100V = 305mV DLVR voltage drop being shown on-screen for an averaged Vcc_Core and Vcc_Atom actual voltage of 1.1V). However, around 28:33 you showed the DLVR configuration using 11W *LESS* (213W versus 224W) than an equivalent Power Gate configuration (PG-2), when the DLVR and PG-2 configs had their effective P-Core voltage matched to within 33mV. Checking the other numbers at 28:33 also shows something weird: The DLVR configuration has its input VccIA = 1.42V and averaging the p-core and e-core VIDs gives an output Vcc_Core of ~= 1.072V, thus resulting in an effective DLVR voltage drop of 1.420V - 1.072V = 348mV. Since the average Package Power of 213W implies an average current consumption (Icc) equal to 213W ÷ 1.420V = 150A, the power lost in the DLVR can be calculated: 0.348V × 150A = 52.2W, which is very close to that presented at 29:40. How can it be that the same CPU, in the same system, running the same benchmark, at the same clock speed, etc., uses roughly 52.2W + 11W ~= *63.2W* more power and (224W / 1.15V) - 150A = 194.78A - 150A = *44.78A* more current to run the cores when the DLVR is disabled. Surely the small 33mV Vcc_Core delta can't explain such a discrepancy?
Awesome question! I don't have an immediate satisfactory answer. My first thought is to be skeptical about the Package Power reading (hence why I didn't highlight it in the video) in PG mode. It's not impossible that this measurement is not entirely accurate with PG mode enabled since it's not a direct power measurement. A good way to verify if the package power reading is accurate would be to measure the 12V input power or wall socket power. I'll have a look at that later this week. FYI, one reason I'm skeptical about the power reading accuracy is because Intel has indicated the PCU isn't working entirely correctly in PG mode. It seems the PCU doesn't change the calculation method with or without DLVR, so in PG mode you'd still get the high VccIA voltage request. Hence, there's chatter of removing PG mode for ambient use in the near future. Thanks for the question!
skatter, given the issues experience and addressed by intel in raptorlake, is overclocking safe for that and what are the safe voltages now? Would you mind doing a video addressing the new reality?
Those issues were ridiculous and multi layered enough to be completely unique. There's no way of the same ever happening, and they were not tied to overclocking. In fact, you were more likely to damage your CPU if it was cool and nicely tuned If there are any issues, they are bound to be completely different
1.3-1.35 is probably 100% safe for 200 series. 1.4-1.45 is fine for benching, maybe 1.4 ok for daily with appropriate cooling. Unfortunately nobody is doing testing if the "new" limit of 1.55v is safe for 13/14th gen but I would bet that it is not safe for long term use.
All good for overclocking Arrow Lake as far as I know. There's a lot less chatter about degradation or failures leading up to ARL launch then there were with Raptor.
@@SkatterBencher you misunderstood me, I was asking about overclocking raptorlake given the issues addressed by intel, what are the new safe voltages, was it ever a temps issue etc etc etc
benefit of 'boost oc' instead of traditional all cores is that it runs at either lower temp or higher clock perf is proper, = clock speed if oc'd properly, all cores is easier and more certain
Der8auer’s video leaked a slide that had Direct Die cooling that had 5.7 P-Cores and 5.1 E-Cores, the performance was higher even than a 7800X3D in CS. While most reviews today were disappointing, perhaps there’s some ability to OC to get some of the performance back or possible software fixes like Zen 5 post launch. Otherwise it looks like a 9950X3D is the way to go for mixed use cases.
Do you want to use a hardcore OC CPU on a daily basis just to match the performance of a CPU that will consume 100w less? But we'll see in a few weeks, maybe the new bios will help this cpu
@@MrAckers75 285K is a general processor good in everything. The 7800X3D is ONLY good in gaming, for the rest it is rubbish and outdone by CPU's like a 13600K or 7700X already. So a 285K matching the 7800X3D in gaming is a ton doing better in anything else. Because the 285K beats those CPU's that the 7800X3D can't. Plus most games are GPU bound anyway, something as specific as an 7800X3D is useless then. Only good for CS etc. players who want 400fps.
There not being a new frequency record on helium was expected. but is there any opportunity fir tsmc silicon to overclock at home on water? Roman teased that direct die has some potential.
Yes, with Direct Die, we can see 5.7 all core, or 5.8. I canceled my Apex board after seeing the ring ratio is maxing on unicorn bins at 4.5, and 6ghz is unobtainable.
It's tough because temperature doesn't appear to be the bottleneck for 1T boosts. Even with TVB I can't get even close to 6 GHz stable at 40C. Direct die won't help with that. All-core, there might be more potential. I had 57X stable in a light load and possibly more headroom. However, power will also go up quite quickly
Sadly no, with Direct Die, LM, & Chilled Mora Sugi (with a tray of 150 CPU's) was only able to get 5.8Ghz all core, but the 100mhz was worth a literal 150watts more power consumption. After canceling my Apex X 285K order, I'm going to grab a 9800X3D and just game for 5 years till we're in a better place tech wise.
DLVR seems pointless in consumer desktop chips, but thats such a small part of the market compared to servers and mobile, we kind of just get stuck with it. Glad to see there's a bypass, but I heard Intel wanted to get rid of it? They would have to prove that dynamic overclocks are better the majority of the time, something that thread director was supposed to achieve years ago. I still don't see the point in 2 core boosts, even with such precise voltage control on individual cores, if it comes at the expense of so much extra heat through DLVR. Looking forward to see what you can achieve, although I imagine there's a good chance AMD will be taking up much of your time before long.
I was going to get 13600, an upgraded 2 cores more version of 12700, but I read of the problems 13/14 had. I actually just found out about the problems last week when I watched benchmark videos. I only watch the numbers, so I don't usually know of a GPU or CPU having problems. I just happened to read on a forum if I should buy a 12700 or 13600, and people were talking about the problems. Even people with new CPUs still get those problems even with the BIOS update, so I thought I might as well stay away from those and get a 12700. So I bought it last week but haven't used it yet because the cooler comes today.
It's some MB's pushing too much voltage on 2 boost cores. Keeping voltage static sub 1.4V and you're fine. I've ran my 13700k at 5.6ghz for over a year at 1.28V and no issues. 12th gen is not aggressively clocked, so no crazy voltages...but you can still usually UV. But if you're running non K it's no issue.n
@Greez1337 Alot of chips are and have run on stock and had crashes, which has nothing to do with messing with anything in the bios. Even people with new chips with the bios,these chips are a luck of the draw,just not worth taking the chance
@@Greez1337 no. CPUs themselves were requesting high voltages when they thought they could handle it. You were simply lucky with your CPU or load patterns. If CPU was hot enough or loaded enough it didn't request those voltages
This video is very sensitive to CCP, you mentioned their name many times, I think the youtube background AI for reviewing videos will give this video yellow sign, this means you can't profit from this video....
You're a true OC master
you're a genius! thanks for this vid! i wish you had many more subs!! you're vids are top-notch!
Haha cant wait for the 5 min overclock video. Squeezing all that p/e/x core/ring info in there seems like impossible task.
At 29:40 you said the power lost through the DLVR amounts to roughly 50W (with 48.8W for a 1.405V - 1.100V = 305mV DLVR voltage drop being shown on-screen for an averaged Vcc_Core and Vcc_Atom actual voltage of 1.1V).
However, around 28:33 you showed the DLVR configuration using 11W *LESS* (213W versus 224W) than an equivalent Power Gate configuration (PG-2), when the DLVR and PG-2 configs had their effective P-Core voltage matched to within 33mV.
Checking the other numbers at 28:33 also shows something weird: The DLVR configuration has its input VccIA = 1.42V and averaging the p-core and e-core VIDs gives an output Vcc_Core of ~= 1.072V, thus resulting in an effective DLVR voltage drop of 1.420V - 1.072V = 348mV. Since the average Package Power of 213W implies an average current consumption (Icc) equal to 213W ÷ 1.420V = 150A, the power lost in the DLVR can be calculated: 0.348V × 150A = 52.2W, which is very close to that presented at 29:40.
How can it be that the same CPU, in the same system, running the same benchmark, at the same clock speed, etc., uses roughly 52.2W + 11W ~= *63.2W* more power and (224W / 1.15V) - 150A = 194.78A - 150A = *44.78A* more current to run the cores when the DLVR is disabled. Surely the small 33mV Vcc_Core delta can't explain such a discrepancy?
Awesome question! I don't have an immediate satisfactory answer.
My first thought is to be skeptical about the Package Power reading (hence why I didn't highlight it in the video) in PG mode. It's not impossible that this measurement is not entirely accurate with PG mode enabled since it's not a direct power measurement. A good way to verify if the package power reading is accurate would be to measure the 12V input power or wall socket power. I'll have a look at that later this week.
FYI, one reason I'm skeptical about the power reading accuracy is because Intel has indicated the PCU isn't working entirely correctly in PG mode. It seems the PCU doesn't change the calculation method with or without DLVR, so in PG mode you'd still get the high VccIA voltage request. Hence, there's chatter of removing PG mode for ambient use in the near future.
Thanks for the question!
@@SkatterBencher Thank you for the thorough answer, it is much appreciated!
skatter, given the issues experience and addressed by intel in raptorlake, is overclocking safe for that and what are the safe voltages now? Would you mind doing a video addressing the new reality?
Those issues were ridiculous and multi layered enough to be completely unique. There's no way of the same ever happening, and they were not tied to overclocking. In fact, you were more likely to damage your CPU if it was cool and nicely tuned
If there are any issues, they are bound to be completely different
1.3-1.35 is probably 100% safe for 200 series. 1.4-1.45 is fine for benching, maybe 1.4 ok for daily with appropriate cooling. Unfortunately nobody is doing testing if the "new" limit of 1.55v is safe for 13/14th gen but I would bet that it is not safe for long term use.
All good for overclocking Arrow Lake as far as I know.
There's a lot less chatter about degradation or failures leading up to ARL launch then there were with Raptor.
@@SkatterBencher you misunderstood me, I was asking about overclocking raptorlake given the issues addressed by intel, what are the new safe voltages, was it ever a temps issue etc etc etc
@@NVMDSTEvil let the people with money to burn find out, give em six months.
Disappointing for P-cores 😢
No no, 5.8 P- 5.4E core on 2nm is inane. 3900 single core geek bench, 25000 mult in cpuz, I am shocked!
@@paulb5125 3nm
@@paulb5125 Where is this ~3900 geek bench score listed? I'm not able to find it but if it can handle that it would amazing
benefit of 'boost oc' instead of traditional all cores is that it runs at either lower temp or higher clock
perf is proper, = clock speed if oc'd properly, all cores is easier and more certain
ah its explained on vid
rather interesting todays bios
Can a memory latency be mitigated through tuning? Intel told us that SoC tile and memory interconnect can be overclocked.
Wait a couple of hours/days. Guarantee he’s just getting the content pieces ready on it.
Yes, I will explain in the MemSS overclocking deep dive video which should go live in 3 days
Did the video cut short?
Der8auer’s video leaked a slide that had Direct Die cooling that had 5.7 P-Cores and 5.1 E-Cores, the performance was higher even than a 7800X3D in CS. While most reviews today were disappointing, perhaps there’s some ability to OC to get some of the performance back or possible software fixes like Zen 5 post launch. Otherwise it looks like a 9950X3D is the way to go for mixed use cases.
Do you want to use a hardcore OC CPU on a daily basis just to match the performance of a CPU that will consume 100w less? But we'll see in a few weeks, maybe the new bios will help this cpu
Der8auer is the only one that had this result, every other reviewer had the 285k very far behind.
Having to overclock the shit out of your new cpu to match an 18 month old cpu isn’t a win
@@MrAckers75 285K is a general processor good in everything. The 7800X3D is ONLY good in gaming, for the rest it is rubbish and outdone by CPU's like a 13600K or 7700X already. So a 285K matching the 7800X3D in gaming is a ton doing better in anything else. Because the 285K beats those CPU's that the 7800X3D can't. Plus most games are GPU bound anyway, something as specific as an 7800X3D is useless then. Only good for CS etc. players who want 400fps.
Overclocking to match an older CPU is hysterical
There not being a new frequency record on helium was expected. but is there any opportunity fir tsmc silicon to overclock at home on water?
Roman teased that direct die has some potential.
Yes, with Direct Die, we can see 5.7 all core, or 5.8.
I canceled my Apex board after seeing the ring ratio is maxing on unicorn bins at 4.5, and 6ghz is unobtainable.
It's tough because temperature doesn't appear to be the bottleneck for 1T boosts. Even with TVB I can't get even close to 6 GHz stable at 40C. Direct die won't help with that.
All-core, there might be more potential. I had 57X stable in a light load and possibly more headroom. However, power will also go up quite quickly
no OC Benchmarks like SuperPI or did i miss something?
Is there a difference in bus speed between power core and e core?
You mean BCLK? No, they're driven by the same reference clock frequency
Yeah, this is cool but how long until it fries the ring bus? Feeling relly concerned about oc on intel after 13 and 14th gen.
Can the 285K do a 6.0 GHZ and above, ALL CORE OC without use of LN2?
Sadly no, with Direct Die, LM, & Chilled Mora Sugi (with a tray of 150 CPU's) was only able to get 5.8Ghz all core, but the 100mhz was worth a literal 150watts more power consumption.
After canceling my Apex X 285K order, I'm going to grab a 9800X3D and just game for 5 years till we're in a better place tech wise.
@@nerdynumen You have a great plan. Hope you didn't waste $$$ on those 150 waste of sand CPU's.
@@JDD_Tech_MODS I'm NerdyNumen, not Sugi0lover who gets the tray for free.
@@nerdynumen Apologies for the confusion!!!
i think 6ghz allcores is easy on all ultra series, certain if i can guess
Can you overclock euro cores to same speed as Porsche cores.
Sorry it is Entertainment cores and Popcorn cores. Please have your CPU language correct.XP
E-core video goes up tonight ... I've got 6 more scheduled in the next 6 days
DLVR seems pointless in consumer desktop chips, but thats such a small part of the market compared to servers and mobile, we kind of just get stuck with it. Glad to see there's a bypass, but I heard Intel wanted to get rid of it? They would have to prove that dynamic overclocks are better the majority of the time, something that thread director was supposed to achieve years ago. I still don't see the point in 2 core boosts, even with such precise voltage control on individual cores, if it comes at the expense of so much extra heat through DLVR.
Looking forward to see what you can achieve, although I imagine there's a good chance AMD will be taking up much of your time before long.
Overclocking igpu is possible?
Yes th-cam.com/video/yjoI3ZqfYIk/w-d-xo.html
Greetings!
I was going to get 13600, an upgraded 2 cores more version of 12700, but I read of the problems 13/14 had. I actually just found out about the problems last week when I watched benchmark videos. I only watch the numbers, so I don't usually know of a GPU or CPU having problems. I just happened to read on a forum if I should buy a 12700 or 13600, and people were talking about the problems. Even people with new CPUs still get those problems even with the BIOS update, so I thought I might as well stay away from those and get a 12700. So I bought it last week but haven't used it yet because the cooler comes today.
It's some MB's pushing too much voltage on 2 boost cores. Keeping voltage static sub 1.4V and you're fine.
I've ran my 13700k at 5.6ghz for over a year at 1.28V and no issues. 12th gen is not aggressively clocked, so no crazy voltages...but you can still usually UV.
But if you're running non K it's no issue.n
@Greez1337 Alot of chips are and have run on stock and had crashes, which has nothing to do with messing with anything in the bios.
Even people with new chips with the bios,these chips are a luck of the draw,just not worth taking the chance
@@Greez1337 no. CPUs themselves were requesting high voltages when they thought they could handle it. You were simply lucky with your CPU or load patterns. If CPU was hot enough or loaded enough it didn't request those voltages
please overclock core ultra 9 285k to 6,2GHz, make it become core i9 15900K
"Let Him Cook His Sand Souffle"
mb
This video is very sensitive to CCP, you mentioned their name many times, I think the youtube background AI for reviewing videos will give this video yellow sign, this means you can't profit from this video....
What?
You've just met someone from the 50-cent army