14th Gen "Raptor Lake Refresh"

[welshrat]

I have been keeping an eye and posting a few things from here on the same topic on neogaf ( I visit there for info on ps5 pro news). No idea on the validity of this posts info but thought it interesting if true regarding intel engineer. Specifically the callofduty tag

Post in thread 'Intel still has not fixed the crashes with 13th and 14th gen CPUs' https://www.neogaf.com/threads/inte...13th-and-14th-gen-cpus.1672744/post-269380614

 

[Frenzy]

That's interesting, I've seen some people say they had 12 series out of video memory issues too...

The post further down confirms what I experienced with the ASUS bloatware:

Analysis - Hardware - Intel still has not fixed the crashes with 13th and 14th gen CPUs

According to this data, modern Intel CPUs are even worse at handling light loads than AMD CPUs, so what exactly are you trying to prove? You are aware those are all heavy loads right? None of that represents idle and low burden workloads which is what we're discussing. No one is contesting...

www.neogaf.com

 

[Rroff]

No idea on the validity of this posts info but thought it interesting if true regarding intel engineer. Specifically the callofduty tag

I don't think this is simply an electron migration issue even an accelerated one.

Neither at this point do I think all these CPUs are ******, no one I've talked to has come close to the failure rates a small number are talking of.

EDIT: Aside from the desire to air cool it, one of the reasons I went with the 14700K over the 14900K, despite pairing it up with a high end board, was due to rumours of these issues from people I trust.

 

[Silent_Scone]

Not quite sure what you mean by that, obviously all semiconductors degrade over time, just with a reasonable life expectancy and primarily it is driven by electromigration which is a current based mechanism but voltage (and related thermals) has an effect.

What part are you struggling with? Think of it like so; the VID is fused and the minimum voltage is no longer sufficient after a short burn-in period at a given frequency. Also different workloads produce different swings in current which is why it might be just enough to topple only a small set of bespoke tasks (but it's still enough to become an issue). This ties into users finding that the problem goes away by fixing the ratio to 55x, even after a short spell of using the CPU.

By everything degrades, I mean you don't necessarily need "too much" voltage (or current, primarily). The voltage simply wasn't enough for a given ratio to begin with.

 

[Rroff]

What part are you struggling with? Think of it like so; the VID is fused and the minimum voltage is no longer sufficient after a short burn-in period at a given frequency. Also different workloads produce different swings in current which is why it might be just enough to topple only a small set of bespoke tasks (but it's still enough to become an issue). This ties into users finding that the problem goes away by fixing the ratio to 55x, even after a short spell of using the CPU.

By everything degrades, I mean you don't necessarily need "too much" voltage (or current, primarily). The voltage simply wasn't enough for a given ratio to begin with.

I think you are talking about something slightly different to what I am.

 

[Silent_Scone]

I think you are talking about something slightly different to what I am.

Not really, all you have to do is consider the voltage guardband - or rather whether it's enough.

 

[Robert896r1]

Not really, all you have to do is consider the voltage guardband - or rather whether it's enough.

So the vids are too optimistic in relation to the vdroop being caused by certain workloads?

 

[Dave2150]

Not sure what you are saying there - I'm not saying any voltage over 12th gen will degrade them fast though saying that was talking last night to someone who works in IT for an advertising agency with a lot of systems used for 3D modelling and video rendering and they reckon based on what they've seen for 24x7 full loaded anything over 1.4V, ~300A, ~280 watt is degrading chips, 1.45 for mixed use or 1.5 for gaming, but they are only seeing like 1-2 systems in 30 impacted (which is still pretty high compared to Intel traditionally) and most of those aren't right out the box rather than degrading later.



Sounds like they won't have E cores on the 12P part so will have a bit of room to play with to improve thermals/power.

Yeah I understand that it'll not have E-Cores. Though Intel's P cores are so incredibly inefficient (on Intel '7' process) that even on 13900k, 1400k's that have E cores disabled in EUFI, the 8 P cores still consume hundreds of watts at high clocks.

Adding 4 more P cores will likely be more power and heat than losing 16 E cores.

 

[Bito]

I’ve been running my 13600K in my gaming setup overclocked and undervolted, with a view to getting the 14900K or the 14900KS, once I was comfortable with the new setup.
May now wait and see what Bartlett Lake-S has to bring instead, if it does work on consumer level motherboards.

 

[Tetras]

Adding 4 more P cores will likely be more power and heat than losing 16 E cores.

I'd be interested in seeing it, just for curiosity at how it scales, but I thought it was said ages ago that Intel's ring bus doesn't do well beyond 10 cores due to the latency penalty between them.

 

[Rroff]

I'd be interested in seeing it, just for curiosity at how it scales, but I thought it was said ages ago that Intel's ring bus doesn't do well beyond 10 cores due to the latency penalty between them.

I'm not sure if there are any mitigations for P core intercommunication but AFAIK the E cores connect, 4 at a time, to the ring bus as just another core making for a total of 12 on the 14900 and 11 on the 14700. So I assume ditching the E cores and having 12 P cores isn't much different in that respect. With some tweaks I doubt there is actually much difference in performance and power efficiency between 12 P cores and 8 P cores with 16 E cores :s

 

[Troezar]

It doesn't seem that surprising. The closer to the edge you push the more likely things are to fail. Overclockers have known this forever. The difference is now they come preoverclocked from the manufacturer. Intel seem to have found the limit, not ideal when you have a warranty to live up to

 

[jabski]

It doesn't seem that surprising. The closer to the edge you push the more likely things are to fail. Overclockers have known this forever. The difference is now they come preoverclocked from the manufacturer. Intel seem to have found the limit, not ideal when you have a warranty to live up to

will be interesting to see what intel do about the warranty and how the measure if you can claim against it

 

[teejay73]

will be interesting to see what intel do about the warranty and how the measure if you can claim against it

I guess it depends on what their findings are with the raptor lake I7's & I9''s. It's looking like a design fault from what I've read, which, if is, isn't going to be resolved via a RMA replacement, as any other like for like cpu is going to go down the same path. Intel is in a precarious position right now, yet I'd still wager on them trying to brush it under the carpet in some way to keep investors happy. I have the 13700k, which is about 6 months old & has been good so far, but I'm not overclocking & I've undervolted it with a 0.5 offset - if anything my pc is primarily a workstation for my editing, with a bit of casual gaming, so it's not used that hard editing photographs.

 

[Tetras]

will be interesting to see what intel do about the warranty and how the measure if you can claim against it

If your CPU crashes running something like Cinebench at stock settings (I mean, actual Intel stock settings, not what Gigabyte/Asus, etc consider stock settings) then I'd consider that a slam dunk that the CPU doesn't work properly. Hopefully in the future Intel will release a new stepping like MLID suggested and one we can actually trust to be fixed, even if they do lose some performance.

 

[Rroff]

I guess it depends on what their findings are with the raptor lake I7's & I9''s. It's looking like a design fault from what I've read, which, if is, isn't going to be resolved via a RMA replacement, as any other like for like cpu is going to go down the same path. Intel is in a precarious position right now, yet I'd still wager on them trying to brush it under the carpet in some way to keep investors happy. I have the 13700k, which is about 6 months old & has been good so far, but I'm not overclocking & I've undervolted it with a 0.5 offset - if anything my pc is primarily a workstation for my editing, with a bit of casual gaming, so it's not used that hard editing photographs.

I've had a 14700K since release, so 9 months, without any issues so far and runs the stuff fine that some chips are faulting on. Using original Gigabyte profile with a slight tweak for gaming purposes to run upto 6GHz.

 

[Tetras]

I've had a 14700K since release, so nearly 9 months, without any issues so far

It would be good to know when the faulty CPUs are manufactured, so that we can confirm if it is related to their manufacture date or not. If it is related to the date, it could be nothing to do with the CPUs design, but a manufacturing problem.

 

[uscool]

Guess it also depends how much it's used , degradation in server use probably the fastest
Or just occasionally windows error that's ignored seems like a right mess

 

[fail at the internets]

so it seems to be an issue with the single core boost massively spiking the voltage to achieve the maximum listed turbo clock speed frequency.

how do i disable ONLY single core boosting in my motherboard?

 

[Rroff]

Guess it also depends how much it's used , degradation in server use probably the fastest
Or just occasionally windows error that's ignored seems like a right mess

Mine gets a huge range of uses though primarily used for gaming I also chuck stuff like video encoding jobs on it and play some games for hours which have high CPU usage. So far had no problems and no random Windows errors, etc.