Is Intel going to deliver a processor/chipset worth waiting for?

[JustBenching]

I am running a slow eco efficient 14900K (I think now at 5200/4200 MHz), do not feel need to buy anything alse on the market now and honestly I have doubts about what Intel will bring as a next CPU. I will need to see it first to believe.

I do not believe my 14900K would last long with default settings. The insane power draw and frequencies Intel sells some CPUs with should be reminded frequently so that people do not destroy their CPUs for no good reason.

No reason to drop to 5.2 ghz unless you have a really bad 14900k. 5.5 with the ecores at stock is what you should be aiming for. HT off to sweaten the deal

 

[Dr. Dro]

Since i got my 12700k i have just left it as is and not turned off the E-cores. I have not found no detriment to having them enabled. Funny thing is how people who hate them don't accept that eventually all CPU's from Intel and AMD will have some form of BIGlittle core setup, get used to it.

On Alder Lake, having the E-cores enabled will restrict ring frequency to match the E-cores

This was improved on Raptor Lake which does not share this limitation

I do not want any high frequencies to avoid high voltage.

E-cores at such a low frequency do not perform well.

How does per core setting work in norrmal AMD CPus? With Intel you can limit individual P cores and groups of four E cores to different frequencies.

That thing with 20 years CPU life at stock settings is simply not true anymore and INTEL LIES. Run your 14900K at stock settings, load it heavilly, let it run at thermal limit and see what will happen...

I do not believe Intel lies. The only CPU I have that I know for a fact that suffered from electromigration (because its v/f curve decayed noticeably over time) is my 13 year old i7-990X processor. I didn't sell it because I full well knew I put 1.6v through it with inadequate cooling on more than one occasion. CPU still works anyway.

 

[JustBenching]

On Alder Lake, having the E-cores enabled will restrict ring frequency to match the E-cores

This was improved on Raptor Lake which does not share this limitation

Yeap, ive tested that extensively and still ecores on with 4.0 ghz cache is better than ecores off 4.7ghz cache. Game dependant of course, some games that are crazily ST dependant scaled better with the faster cache.

But then you are also using a lot more power cause you cant run the same undervolts anymore with the cache at 4.7.

4.9 / 4.0 / 4.0 needs 1.12v, 4.9 / 4.7 with ecores off needs 1.26 if I remember correctly. It's really not worth it.

 

[dgianstefani]

I do not want any high frequencies to avoid high voltage.

Your loss, it's not "high" voltage, nor is it dangerous.

E-cores at such a low frequency do not perform well.

They perform just fine, they're not designed for foreground processes anyway, so if you notice their "performance" drop, it's because you're running synthetic benchmarks, not normal usage. Besides, you're getting 16 whole cores to run everything except foreground processes, so a few E cores at 3.2 GHz is still faster than having to run those processes together with whatever you're doing in the foreground, shared on a single P core.

How does per core setting work in normal AMD CPUs? With Intel you can limit individual P cores and groups of four E cores to different frequencies.

You don't get per core OC with AMD. That's the point of what I wrote. Because I don't need MT perf at the moment, the 7800X3D works well for me, but I miss per core OC, and it's a shame for most Intel users to not take advantage of it.

That thing with 20 years CPU life at stock settings is simply not true anymore and INTEL LIES. Run your 14900K at stock settings, load it heavilly, let it run at thermal limit and see what will happen...

Believe what you want, the chips perform fine and last a long time at stock.

On Alder Lake, having the E-cores enabled will restrict ring frequency to match the E-cores

This was improved on Raptor Lake which does not share this limitation



I do not believe Intel lies. The only CPU I have that I know for a fact that suffered from electromigration (because its v/f curve decayed noticeably over time) is my 13 year old i7-990X processor. I didn't sell it because I full well knew I put 1.6v through it with inadequate cooling on more than one occasion. CPU still works anyway.

Yep, even with "high" voltages on a static OC, the CPU will by default use higher voltages when boosting to max frequency on stock settings. The chips are validated for this and it's well within tolerances. People overestimate the issues of temperature/voltage on Intel. It's much more of a problem on AMD where you can kill the cache with voltage, or can't get rid of even 65 W worth of heat easily because of the thick stock IHS.

 

[JustBenching]

That thing with 20 years CPU life at stock settings is simply not true anymore and INTEL LIES. Run your 14900K at stock settings, load it heavilly, let it run at thermal limit and see what will happen...

It's not that I trust intel to care about me but I'm pretty confident they care about their bottomline. Them shipping cpus that would die at stock settings would directly affect their bottom line due to rmas, so I don't think that is the case.

Now if you are talking about no power or Amp limits and running 24 hour synthetic tests like prime 95 and ycruncher, the kind of crap they do on ocnet to test stability then sure, I can see electromigration knocking at your door. I strongly advice not doing that unless you want to RMA soon.

 

[BoggledBeagle]

Believe what you want, the chips perform fine and last a long time at stock.

If you have some money to burn, get yourself a system with 14900K, let it run with high load at a thermal limit with default settings, and count the amount of days or weeks, how long will it last.

 

[dgianstefani]

If you have some money to burn, get yourself a system with 14900K, let it run with high load at a thermal limit with default settings, and count the amount of days or weeks, how long will it last.

You sound a bit like a conspiracy theorist.

If you actually believe the literal tens of millions of CPUs Intel sells, that all run at stock speeds/voltages, typically with underpowered cooling in crappy prebuilts/company PCs, are all killing themselves within "weeks" then... good for you I suppose. I don't normally agree with people who say that tuning a CPU to personal preference is a waste of money "just buy a slower CPU", but in this case, you're losing performance for practically zero gain (some theoretical lifespan prolonging voltage).

How about all those laptops with 13/14900HX CPUs, that manage to fit the same silicon within a laptop chassis/cooling and yet are still going strong?

 

[JustBenching]

You sound a bit like a conspiracy theorist.

If you actually believe the literal tens of millions of CPUs Intel sells, that all run at stock speeds/voltages, typically with underpowered cooling in crappy prebuilts/company PCs, are all killing themselves within "weeks" then... good for you I suppose.

In fact using a crap cooler is the best thing you can do to a 14900k if you don't want to go into the bios. Use an Intel stock cooler and the cpu will limit itself to like 120 watts. Great, now you have the most efficient cpu that mankind was ever blessed with

 

[BoggledBeagle]

If you actually believe the literal tens of millions of CPUs Intel sells, that all run at stock speeds/voltages, typically with underpowered cooling in crappy prebuilts/company PCs, are all killing themselves within "weeks" then... good for you I suppose.

14500 in a powerful office machine at 5 GHz will not fry running excel spreadsheet calculations.

 

[dgianstefani]

14500 in a powerful office machine at 5 GHz will not fry running excel spreadsheet calculations.

You're missing the point, whether intentionally or otherwise, but whatever.

 

[JustBenching]

14500 in a powerful office machine at 5 GHz will not fry running excel spreadsheet calculations.

Just FYI, Im not an expert or anything, but dropping your clockspeeds or lowering your voltages will not really affect degradation. It's the current that causes electromigration (combined with temperatures). Voltages and clockspeeds are kinda irrelevant.

 

[BoggledBeagle]

How about all those laptops with 13/14900HX CPUs, that manage to fit the same silicon within a laptop chassis/cooling and yet are still going strong?

No sane person will load such computer with high load 24/7. If they did, the failure rate would be high.

 

[dgianstefani]

No sane person will load such computer with high load 24/7. If they did, the failure rate would be high.

Can you stop shifting the goalposts? Either respond to the specific points being raised or we can talk about something else.

Most of those laptops with CPUs of that calibre are used for professional workstations or high end gaming machines, where they are indeed loaded for long periods of time.

 

[BoggledBeagle]

I think that a person with money to burn should conduct that test.

 

[JustBenching]

I think that a person with money to burn should conduct that test.

Well, don't have a 14900k anymore, but I can test it on a 12900k. What do you want me to do and for how long?

I've already blasted the chip at 1.65v 114c for 5.6 ghz all core. It's still fine

 

[dgianstefani]

Well, don't have a 14900k anymore, but I can test it on a 12900k. What do you want me to do and for how long?

I've already blasted the chip at 1.65v 114c for 5.6 ghz all core. It's still fine

I mean, it's nice of you to offer and I'm sure some people would find the results interesting (and predictable), but Boggled is the one making unusual and unlikely claims, so the burden of proof is on him, not you.

 

[JustBenching]

I mean, it's nice of you to offer, but Boggled is the one making unusual and unlikely claims, so the burden of proof is on him, not you.

I don't mind, for SCIENCE!!

 

[BoggledBeagle]

I've already blasted the chip at 1.65v 114c for 5.6 ghz all core. It's still fine

For how long? These are insane conditions. I think you already damaged the chip somewhat with this.

I do not believe that the chip can survive this for even one hour with high load.

 

[RandallFlagg]

Nah, the 3300X wasn't really all that it was cracked up to be. It just had no additional execution units beyond that single CCX, something that AMD rectified by the 5800X.

This info on Intel is very out of date though and turned out to be inaccurate, Raptor Cove does have physical changes vs. Golden Cove and is an improvement over ADL. Reading back when they thought 14th Gen would be Meteorlake instead of just the laziest rehash this industry has ever seen is pretty funny though.

Intel itself states that ADL and RPL are the same microarchitecture. Just messing around with the cache and interconnects doesn't constitute a new uArch, in the same way that changing from a carb to fuel injection on a 350 v8 doesn't make it a new engine.

Intel Docs List Raptor Lake With Same Microarchitectures as Alder Lake

Details about Intel's next-generation cores revealed.

www.tomshardware.com

 

[BoggledBeagle]

I think a fair test conditions for 14900K would be:

Running it at stock frequencies with 253W power limit, with Cinebecnch R23 load, with cooler fans set so that the CPU runs at or just below the temperature limit.

I would expect the chip last just a few days. Not 20 years.

((( I secretly think that it is hours, not days )))

 

[dgianstefani]

I think a fair test conditions for 14900K would be:

Running it at stock frequencies with 253W power limit, with Cinebecnch R23 load, with cooler fans set so that the CPU runs at or just below the temperature limit.

I would expect the chip last just a few days. Not 20 years.

((( I secretly think that it is hours, not days )))

If I were a betting man I'd wager any amount of my savings that this isn't the case.

 

[BoggledBeagle]

This is a nice opportunity for bidding, I would make a 5 minute time slots to bid on.

Place my bid of $5 on 7h 40-45 Minutes slot, please. The winner takes all, and the experiment conductor should have his cost covered.

 

[dgianstefani]

This is a nice opportunity for bidding, I would make a 5 minute time slots to bid on.

Place my bid of $5 on 7h 40-45 Minutes slot, please. The winner takes all, and the experiment conductor should have his cost covered.

Not sure we're allowed to do this on the forum. If a mod says it's OK I'd be all in.

Running it at stock frequencies with 253W power limit, with Cinebecnch R23 load, with cooler fans set so that the CPU runs at or just below the temperature limit.

Stock settings with 253 W PL enforced, no other changes. Stock voltages etc. Cooler tuned to sit at 100 C, or 99 C so that it doesn't throttle.

Anyone with a custom loop could easily do this by just finding the fan speed equilibrium for 250 W.

I'd be wary of doing this with an AIO though as the pump would probably overheat and burn out.

Ideally a brand new chip that hasn't been fiddled with previously, i.e. no running at 6.2 GHz all core for a year or smth.

Out of curiosity PL1/PL2 etc. what are the stock time limits? I know for laptops they'll boost to one of those for x amount of seconds, then revert to a lower wattage target, but that's because their coolers are limited, I'd assume it's different for desktop.

 

[BoggledBeagle]

Stock settings with 253 W PL enforced, no other changes. Stock voltages etc. Cooler tuned to sit at 100 C, or 99 C so that it doesn't throttle.

Yes. You can do this with an air cooler. You can set the fan speed regulation so that it holds 100°C.

 

[dgianstefani]

Would also need to ensure it was tested on a motherboard and PSU with power delivery that can comfortably supply 250 W continuous, but that's a fairly low bar.