News China's largest Core i9-14900K gaming cafe has suffered from instability issues since 2023 — the flagship store has 171 gaming PCs with Core i9-149...

[TheHerald]

1) they restricting both by different degree, and by which ppl other than them won't use the same exact settings, so they are NOT representative to anyone other than themselves, if I, say, go opposite, overclock both CPUs, one by 5% and another by 15% and give some failure rate numbers, do you think it is comparable?

They are using them on default settings. The way intel and amd intended them to.


But still, are you saying that it makes for a restricted zen 4 CPU to have such a high failure rate, 5 times higher than alderlake?

2) you mentioned you undervolt your CPU everytime, I didn't say that as I believe in intel stock before the RPL gen. anything not running intel spec by using the intel profiles, is tweaking settings like voltage and power, which is the same, comparing products need to be in stock form, not random tweaks deduced form some random ppl

I am but I never said anything about undervolting here. I said intel increased performance by ~40% at same power within 1 year. AMD did 10% in 2 years. Why is undervolting relevant here?

3) now who is doing the product class matching ? intel and AMD, they compare themselves with the K and X chips, not some haters, maybe intel hate themselves.

The prices. 282$ for the 7700x, 293$ for the 13700

4) You finally got it, it is those matters to YOU, not anyone else on the world or here, so why you bother bashing others saying they lost trust in intel as they provide stock products which is unreliable and have disasterous response to the issue, 2 years after it's launch? you get us all arguing against you is perfectly suggesting the same thing

In your very first post you asked ME why would I still buy Intel. Of course im explaining what matters to me, since that's literally what you asked

 

[YSCCC]

They are using them on default settings. The way intel and amd intended them to.


But still, are you saying that it makes for a restricted zen 4 CPU to have such a high failure rate, 5 times higher than alderlake?


I am but I never said anything about undervolting here. I said intel increased performance by ~40% at same power within 1 year. AMD did 10% in 2 years. Why is undervolting relevant here?


The prices. 282$ for the 7700x, 293$ for the 13700


In your very first post you asked ME why would I still buy Intel. Of course im explaining what matters to me, since that's literally what you asked

1) they are NOT using them on default settings, the default settings for intel RPL are the intel extreme/performance profiles in their spec sheets, and now in the latest bios, all coded as intel specced in, puget is running them even lower in what puget thinks it's better, that's why the now install and forced enforced defaults score something like 36k for 14900k, but puget got it 33k or so. even the PL1 and PL2 isn't the same as the intetl spec, they do it as their secret sauce. you selectively choose to read, they are saying the zen 4 having % higher, which I told you a hundred times it could well be of different denominator, with insufficient sample size and not similar base sample sizes, % makes no sense, luck plays more in those, and yet the restriction secret sauce of both systems arn't published, we don't know how much they changed.

2) it is not increased ~40% more with same power within 1 year, they released them specced in a higher power, that is their spec, and the 40% only appears in limited scenarios, not across the board, and you keep ignoring that by doing so, under their spec, they self destruct.

3) price means what the market is paying them, i.e. consumer value, and nothing more than that. just like in GPU market, you want best bang for buck? never Nvidia


4) I ask you because you literally said you would buy AMD just a while ago, and then you said you will buy Intel now, and if benchmark isn't up to your liking, you will go back to AMD or skip it, which was fine, but the original question was: why you mock the other member of saying at current status he/she won't recommend intel, that was the issue, not you buying whatever, I don't really care about you buying intel amd, mac or go to use relic pentium.

 

[TheHerald]

1) they are NOT using them on default settings, the default settings for intel RPL are the intel extreme/performance profiles in their spec sheets, and now in the latest bios, all coded as intel specced in, puget is running them even lower in what puget thinks it's better, that's why the now install and forced enforced defaults score something like 36k for 14900k, but puget got it 33k or so. even the PL1 and PL2 isn't the same as the intetl spec, they do it as their secret sauce. you selectively choose to read, they are saying the zen 4 having % higher, which I told you a hundred times it could well be of different denominator, with insufficient sample size and not similar base sample sizes, % makes no sense, luck plays more in those, and yet the restriction secret sauce of both systems arn't published, we don't know how much they changed

They are using Pl1 = 125, Pl2 = 253 TAU = 56sec.

Those are the exact settings you can find in Intel's page

https://www.intel.com/content/www/u...-36m-cache-up-to-6-00-ghz/specifications.html

2) it is not increased ~40% more with same power within 1 year, they released them specced in a higher power, that is their spec, and the 40% only appears in limited scenarios, not across the board, and you keep ignoring that by doing so, under their spec, they self destruct.

Yes, it is literally increased by 40% at same power within 1 year. Every review shows that. When using higher power the performance difference is even bigger, approaching 50% according to TPU's numbers in MT workloads. 50 freaking %. Within 1 year. But sure, let's celebrate the 10% amd gave us after 2, just because it says amd on the box. Nah, ill pass, you can celebrate without me.


The only CPUs that self destructed (taking the mobo with them) was the x3ds.

3) price means what the market is paying them, i.e. consumer value, and nothing more than that. just like in GPU market, you want best bang for buck? never Nvidia

Are you telling me that 2 cpus that were launched at the same time, have the same price and follow the same naming scheme aren't competitors? Okay man, sounds silly but sure.

4) I ask you because you literally said you would buy AMD just a while ago, and then you said you will buy Intel now, and if benchmark isn't up to your liking, you will go back to AMD or skip it, which was fine, but the original question was: why you mock the other member of saying at current status he/she won't recommend intel, that was the issue, not you buying whatever, I don't really care about you buying intel amd, mac or go to use relic pentium.

I was expecting the 9950x to be much better than it actually landed, so yeah I was going to buy the 9950x 3d but zen 5 have been a disappointment. I changed my mind when presented with evidence.

Im mocking users who would never under any circumstance would suggest Intel no matter what. The internet is full of intel hating BS man. If you can't see it, maybe you are part of the problem.

 

[Pierce2623]

Not saying it isn't without compromise, but being cheaper is a good first step!

That’s fair. I guess my main point is if I’m already spending $2000, I’m going to go ahead and spend the $3000 to buy a high end version of most components in the system. Even the best pre-built machines don’t cool well compared to a normal well planned standard self built system. There’s no way that Alienware prebuilt keeps the GPU and CPU both at 60c with a standard gaming workload like your average self-built system with those components. (And yes even a 13900k/14900k hovers around 60c or lower in an average gaming workload in a decently cooled machine)

 

[YSCCC]

They are using Pl1 = 125, Pl2 = 253 TAU = 56sec.

Those are the exact settings you can find in Intel's page

https://www.intel.com/content/www/u...-36m-cache-up-to-6-00-ghz/specifications.html




Yes, it is literally increased by 40% at same power within 1 year. Every review shows that. When using higher power the performance difference is even bigger, approaching 50% according to TPU's numbers in MT workloads. 50 freaking %. Within 1 year. But sure, let's celebrate the 10% amd gave us after 2, just because it says amd on the box. Nah, ill pass, you can celebrate without me.


The only CPUs that self destructed (taking the mobo with them) was the x3ds.


Are you telling me that 2 cpus that were launched at the same time, have the same price and follow the same naming scheme aren't competitors? Okay man, sounds silly but sure.


I was expecting the 9950x to be much better than it actually landed, so yeah I was going to buy the 9950x 3d but zen 5 have been a disappointment. I changed my mind when presented with evidence.

Im mocking users who would never under any circumstance would suggest Intel no matter what. The internet is full of intel hating BS man. If you can't see it, maybe you are part of the problem.

1) https://www.pugetsystems.com/labs/articles/power-draw-and-cooling-14th-gen-intel-core-processors/
let me quote for you :
Historically, the default settings for these have been PL1=Thermal Design Power (TDP, 125 W), PL2=Maximum Turbo Power (253 W), and Tau=56 s. However, since their 12th generation, Intel has been setting PL1=PL2 (the full 253 W for both) by default on “K” SKU processors.

At Puget Systems, we currently go against Intel default guidance and set TB2 to operate with PL1=125 W, PL2=253 W, and Tau=56 s.

That is against intel default guidance, which, most ppl buy the CPU based on reviews on that exact settings.

2) Apparently intel do think they have issues to have self destructing CPUs for 2 whole freaking years so they have to issue the microcode fixes, maybe in your world a degraded CPU isn't self destructed, but to anyone else it is

3) They can be competitors, but it's a single SKU, not gen on gen, now suddenly you go for the low end of the spectrum, good goal post moving

4) only you said as long as AMD badge on it it is fine, we appreciate the gen on gen leap from 12th to 13th before all those issues come along and paid our money for it (I literally bought the 14900k to have in sockeet upgrade and vote myself back then before the saga arise, coz I consider the leap from 12700k is justifiable to have it last like 2 years longer, did you?)

5) you are mocking Everyone who said they will not recommend intel at the current situation as far as I saw you in the forum, finger pointing to everyone who don't agree on your manually limiting below intel spec power usage and recommend AMD at the meantime, that is the problem, you are the AMD hating BS problem, not ppl hating intel, ppl don't hate intel before the degradation issues, they hate how intel hide the porblem for 2 freaking years and reluctant to deal with it until their 15th gen is around the corner, you can see how many are backing you up against how many just argue with you for the context, if you dont' see the problem... I will skip the last part

 

[YSCCC]

That’s fair. I guess my main point is if I’m already spending $2000, I’m going to go ahead and spend the $3000 to buy a high end version of most components in the system. Even the best pre-built machines don’t cool well compared to a normal well planned standard self built system. There’s no way that Alienware prebuilt keeps the GPU and CPU both at 60c with a standard gaming workload like your average self-built system with those components. (And yes even a 13900k/14900k hovers around 60c or lower in an average gaming workload in a decently cooled machine)

That's fair for us geeks, but TBF a lot of ppl just don't want to spend the time and have more than enough money to pay for the whole system and use their warranty and updoor repair service if anything happens, even a father with spare moeny to spoil the son a bit can have the prebuilt systems work better than trying to learn and deal with issues.

Especially say for the degradataion issue, tons of DIY geeks have been hair pulling and tweaking back and forth thinking it's them doing something wrong so the working machine just don't work, when these happens, the prebuilt could be a stress relief and let them pull the hair. And tbh, from the past track records sens the disasterous 13th and 14th gen intel, a prebuilt running some 80C in gaming workload just survive well beyond like 5 years unless you let the fan clot with tar and spiderwebs, which for a gaming PC is at it's final days of great gaming performance days anyway.

 

[TheHerald]

That is against intel default guidance

That's guidance. It's for reviewers. It's telling them what settings to use. Those are not the defaults. The defaults are in the link.

The rest is going in circles. Enjoy your 10% every 2 years and let's agree to disagree.

 

[YSCCC]

That's guidance. It's for reviewers. It's telling them what settings to use. Those are not the defaults. The defaults are in the link.

The rest is going in circles. Enjoy your 10% every 2 years and let's agree to disagree.

Just in case you can't read English: Many reviewers and enthusiasts set these to an “extreme” specification, where PL1=PL2 at 253 W and the current limit for the CPU is increased—Intel has added this higher performance profile to the CPU specification, including an elevated ICCMax of up to 400 A.

These are the "extreme" spec, along the "performance" and "baseline" which are all INTEL SPEC. they are using the lower one, so it isn't representative to the failure rate using the highest performance intel spec. it's you keep going in circle man

but yea, let's agree to disagree and whine for both sides usually have 10-20% increase gen on gen, or self destruct getting 40% in one aspect

 

[TheHerald]

Just in case you can't read English: Many reviewers and enthusiasts set these to an “extreme” specification, where PL1=PL2 at 253 W and the current limit for the CPU is increased—Intel has added this higher performance profile to the CPU specification, including an elevated ICCMax of up to 400 A.

These are the "extreme" spec, along the "performance" and "baseline" which are all INTEL SPEC. they are using the lower one, so it isn't representative to the failure rate using the highest performance intel spec. it's you keep going in circle man

Man, if you back to the original puget thread, that's exactly what ive been saying. That as long as you use the 125/253 TAU specification (whether you wanna call it default, baseline or whatever the heck you want) intel chips are more reliable than amd chips (again, according to puget). I never, ever, ever suggested that running these chips at 400 watts makes them more reliable. You are acting like I claimed otherwise, but I never did.

Btw, I really doubt that even with the extreme specifications (400a / 253w) those CPUs degrade. It's the unlocked profiles that cause issue. 253 w is nothing, not enough power per core to cause any issues. At 253w a 13900k pulls lower power (per core) and is running at lower voltage than a 12900k. If the 12900k is reliable, then so is the 13900k with those settings.

but yea, let's agree to disagree and whine for both sides usually have 10-20% increase gen on gen, or self destruct getting 40% in one aspect

The x3d were also self destructing . So clearly it's not because of the 40%.

Again, since 13th gen is 40% faster in MT at ANY wattage, clearly you can lock them to 125w, avoid the self destruct part and still be 40% faster and more efficient than 12th gen.

 

[YSCCC]

Man, if you back to the original puget thread, that's exactly what ive been saying. That as long as you use the 125/253 TAU specification (whether you wanna call it default, baseline or whatever the heck you want) intel chips are more reliable than amd chips (again, according to puget). I never, ever, ever suggested that running these chips at 400 watts makes them more reliable. You are acting like I claimed otherwise, but I never did.

Btw, I really doubt that even with the extreme specifications (400a / 253w) those CPUs degrade. It's the unlocked profiles that cause issue. 253 w is nothing, not enough power per core to cause any issues. At 253w a 13900k pulls lower power (per core) and is running at lower voltage than a 12900k. If the 12900k is reliable, then so is the 13900k with those settings.

Nope for both, I never say you suggested that running them at 400W makes them more reliable, what I said is you suggest that to use the data for CPUs running at a lower performance intel spec profile and expect the reliability will be the same as those using more performing, still intel spec, extreme profile, which is 99% of users will choose when they buy the top chip, you keep banging the wall of twisting what I said.

And after these 2 years you still didn't read what intel finally discover don't you?

It's not actually the high wattage that degrades the CPU, even with unlocked 4096W profile, during normal operations except running cinebench or AVX2 non-stop 24/7, the CPU rarely exceed 200W in normal load like photoshop or gaming, yet still there are widespread cases where ppl faces degradation, which in July, Intel admitted it is their microcode having issues, pumping crazy amount of voltages during transients not even noticable through monitoring software like Hwinfo64. And that, is precisely Intel's issue, the 125/253W is only pointing the finger to mobo to diverge the responsibility.

A lot of affected users only finally find out when they play the UE5 games which is very sensitive to the degradation found out it's not Nvidia or other part's problem but the Intel CPU, the computer works fine except those circumstances which is why you don't see massive returns throughout the year especially for 2023 as ppl keep using the computer blaming the software or Nvidia and the outrage just surfaced, but well, you won't read all those anyway.

 

[TheHerald]

Nope for both, I never say you suggested that running them at 400W makes them more reliable, what I said is you suggest that to use the data for CPUs running at a lower performance intel spec profile and expect the reliability will be the same as those using more performing, still intel spec, extreme profile, which is 99% of users will choose when they buy the top chip, you keep banging the wall of twisting what I said.

But puget also runs AMD at lower performance as well. Still, it fails at higher rates. You are using that argument against intel but you are not using it against AMD. Very curious

It's not actually the high wattage that degrades the CPU, even with unlocked 4096W profile, during normal operations except running cinebench or AVX2 non-stop 24/7, the CPU rarely exceed 200W in normal load like photoshop or gaming, yet still there are widespread cases where ppl faces degradation, which in July, Intel admitted it is their microcode having issues, pumping crazy amount of voltages during transients not even noticable through monitoring software like Hwinfo64. And that, is precisely Intel's issue, the 125/253W is only pointing the finger to mobo to diverge the responsibility.

GREAT, then that clearly demonstrates that the degradation issue has nothing to do with the 40% performance increase. But you keep acting like they only achieved the 40% by degrading the chip, which as you just admitted is clearly not the case.

Just back to my original point from page 2, Intel managed a 40% performance increase within a year, so clearly a 10% increase over 2 years that amd achieved is laughable to me. You find it plenty. It's fine, we don't have to agree.

You know what the insanely funny part is? The 13700 is so much faster that even if it degrades every 6 months and you have to wait for 5-10 days for the RMA to go through youll still end up doing more work than if you had the 7700x instead. Let that sink in. It's insane.

 

[YSCCC]

But puget also runs AMD at lower performance as well. Still, it fails at higher rates. You are using that argument against intel but you are not using it against AMD. Very curious

Both are VERY LOW failure rate in their time frame and as I mentioned a dozen times prior, the denominator is vastly different in AMD vs Intel for them, so primary mathematics, if one got unlucky having 1 dead sample out of all their CPUs, the system selling 100 unit of x have a failure rate of 1%, while if Y only sold for 50 unit will have 2%, when the denominator is not vanishingly large, luck have more to be attributed to the failure rate, that's why it isn't comparable, and AMD, while having "horrid" failure rate, do not have such concentrated failure in UE5 or other aspects, so that's a "normal" failure curve, vs the RPL being a general degradation issue, also known as a product failure.

GREAT, then that clearly demonstrates that the degradation issue has nothing to do with the 40% performance increase. But you keep acting like they only achieved the 40% by degrading the chip, which as you just admitted is clearly not the case.

Just back to my original point from page 2, Intel managed a 40% performance increase within a year, so clearly a 10% increase over 2 years that amd achieved is laughable to me. You find it plenty. It's fine, we don't have to agree.

They did have to achieve the 40% MT performance under specific circumstances by pushing it too far and degrading the chip, the volatge is what Intel pumped too close to the safety limit so the transients kills it in order to achieve enough yield to push the chips out as is.

In more plain english, they pushed the cache, core count and frequencies, in order to make them work at the specced frequency, they pump too much voltage into the VID table to make them stable across the board, which in turns the micorcode will result in spikes too hot and kill it. ADL basically use the same microcode as the early RPL, but since it runs much slower, it goes nowhere near the VID table RPL does, and thus making them suiciding. By the fixed microcode, the same RPL chip, just testing default, shows significant performance degradation for example:
"PC Guide tested two different heavily multithreaded benchmarks immediately before and immediately after the BIOS update and found consistent performance degradation of 22–23% in each. "

AKA, if intel release the RPL with the fixed microcode and default enabled bios, the performance is 22-23% slower than they are released, so if basic math is correct, 140% times (100-22)=78%= 109.2%, which is precisely... 9.2% gen on gen performance increase.

You know what the insanely funny part is? The 13700 is so much faster that even if it degrades every 6 months and you have to wait for 5-10 days for the RMA to go through youll still end up doing more work than if you had the 7700x instead. Let that sink in. It's insane.

And for your reference, at release price, 13700k is at 409 vs 7800x at 449 close to 7700x at 299, you know for work what is the insanity? if you get 13700k degrading every 6 months and you need 5 days for RMA, it is NOT making more money for your work, not even games.

Say for working purpose, simplify it as doing video editing for clients. let's say you need 1 hour rendering the daily job in 13700k and 3 hours in 7700x, you ended up taking same days of work submitted to your client, coz you do 8 hours editing, and click the button, let it transcode through your dinner time and family time shows, come back before your sleep or when you woke up, bam you have your work able to submitted to your client. You can shutdown the PC earlier in the 13700k for 2 hours, but when it degrades, assume it's instant working -> f_ up, so you suddenly have a day press the button, go for dinner, come back and the file encoding failed, you ended up instantly send it to intel next morning and get it back a week later, sorry you have your submission delayed for 5 working days. In normal business, you either give a big F you we are sorry but no compensation to your client, or you need to sublet your work and give some compensation to your affected client, which means losing money.

Same for the cover story Chinese Internet cafe, if it runs 30% slower, but still playable framerates, they can still make money at the lower tier machines fares, but sorry you crashed their very tense gaming sessions, you have to refund or give them more free hours to compensationg, which means losing money. That's what matters to >90% of ppl buying high end PC

 

[TheHerald]

Both are VERY LOW failure rate in their time frame and as I mentioned a dozen times prior, the denominator is vastly different in AMD vs Intel for them, so primary mathematics, if one got unlucky having 1 dead sample out of all their CPUs, the system selling 100 unit of x have a failure rate of 1%, while if Y only sold for 50 unit will have 2%, when the denominator is not vanishingly large, luck have more to be attributed to the failure rate, that's why it isn't comparable, and AMD, while having "horrid" failure rate, do not have such concentrated failure in UE5 or other aspects, so that's a "normal" failure curve, vs the RPL being a general degradation issue, also known as a product failure.

Ah, so both zen 3 and zen 4 scored on the top of the failure chart due to pure luck. Okay.

They did have to achieve the 40% MT performance under specific circumstances by pushing it too far and degrading the chip, the volatge is what Intel pumped too close to the safety limit so the transients kills it in order to achieve enough yield to push the chips out as is

No they did not. You are obviously flat out lying. The 13900k is 35 to 40% faster than the 12900k at EVERY wattage. I've already posted proof of that. Even at 125w it's 35% faster. So no, they didn't have to push anything to achieve that result. They could have released both the 12900k and the 13900k at 125w and the performance gen on gen would still be at the 35-40% mark.

In more plain english, they pushed the cache, core count and frequencies, in order to make them work at the specced frequency, they pump too much voltage into the VID table to make them stable across the board, which in turns the micorcode will result in spikes too hot and kill it. ADL basically use the same microcode as the early RPL, but since it runs much slower, it goes nowhere near the VID table RPL does, and thus making them suiciding. By the fixed microcode, the same RPL chip, just testing default, shows significant performance degradation for example:
"PC Guide tested two different heavily multithreaded benchmarks immediately before and immediately after the BIOS update and found consistent performance degradation of 22–23% in each. "

AKA, if intel release the RPL with the fixed microcode and default enabled bios, the performance is 22-23% slower than they are released, so if basic math is correct, 140% times (100-22)=78%= 109.2%, which is precisely... 9.2% gen on gen performance increase.

No matter what microcode intel releases the 13900k will be 35 to 40% faster than the 12900k at same power. That's just a fact. If PC guide (whatever that is) says otherwise they are clearly wrong. Whatever you are calculating there is obviously wrong, since if the 13900k is just 9% faster than the 12900k then the 13700k ends up being slower. Yeah, no. You are just wrong. Stop it man.

And for your reference, at release price, 13700k is at 409 vs 7800x at 449 close to 7700x at 299, you know for work what is the insanity? if you get 13700k degrading every 6 months and you need 5 days for RMA, it is NOT making more money for your work, not even games.

MSRPS for both CPUs were similar my man. But in any case, I was using current pricing.

And yes, say I'm doing F@H (folding at home), even if I rma every 6 months ill still end up doing more work with a 13700 than a 7700x. That's how far ahead Intel is over amd.


PS1. Okay I found the page that you quoted the 22% from. That's using a 125w power limit on the 13900k vs unlimited. The 13900k at 125w score 31k points, making it 14,8% faster than the 12900k at half the power. So nothing changed with the microcode, Intel pushed performance by 15% at half the power within 1 year, amd pushed performance by 10% within 2 years with 0 reduction in power. The equation remains the same, amd took some baby steps, intel took a freaking huge leap.

 

[YSCCC]

Ah, so both zen 3 and zen 4 scored on the top of the failure chart due to pure luck. Okay.

They don't disclose the base no. so it is a no conclusion, and it is not the top of the chart, 11th gen is, yet 11th gen intel doesn't have massive issues in consumer feedback, coz it is more likely like zen3/4 or other intel gen being based on DOA luck, not the "it will eventually die due to design flaw"

No they did not. You are obviously flat out lying. The 13900k is 35 to 40% faster than the 12900k at EVERY wattage. I've already posted proof of that. Even at 125w it's 35% faster. So no, they didn't have to push anything to achieve that result. They could have released both the 12900k and the 13900k at 125w and the performance gen on gen would still be at the 35-40% mark.

You flat out do not understand English no? I am saying workload type and you are going to wattage.

For example, if you go benchmarking in very heavily threaded workloads, the extra E cores, higher frequency and cache can have you 40%, but going single core, or gaming, it is single digit performance gain, that is what in english, the gain is workload dependant, nobody every say it's wattage dependant.

Intel Core i9-13900K vs. Core i9-12900K: worth the upgrade? | Digital Trends

Intel's latest flagship, the Core i9-13900K, brings up the core count and offers various upgrades. How does it compare to the Core i9-12900K?

www.digitaltrends.com

No matter what microcode intel releases the 13900k will be 35 to 40% faster than the 12900k at same power. That's just a fact. If PC guide (whatever that is) says otherwise they are clearly wrong. Whatever you are calculating there is obviously wrong, since if the 13900k is just 9% faster than the 12900k then the 13700k ends up being slower. Yeah, no. You are just wrong. Stop it man.

No, they are comparing using the 0x129 bios with intel bios, it is slower than the old microcode for itself. So if they use the performance bios profile it can get 22% slower than the old unlimited profile, which is the 40% performance gain comes from.

Even using extreme profile, in my own 14900k, with old microcode and unlimited wattage in the only run I tested when it released before my undervolting tuning, it got 40k R23 using air cooler, compared to 27k for 12900k, whichis 48% increase, going for the 0x129 microcode bios and intel extreme profile without further undervolting for each step and trial for stability more than just able to run cinebench 10min test, gives me 32-33k marks, which is... 22% more than the stock 12900k. now that is close to the 15% zen 4 vs zen 5, not to say it seems the windows update is pumping up the zen 5 performance also. And mind you, for a normal release, the 13900k/14900k should be released in the 33k marks state using the "fixed" (if it really fixes) microcode and intel extreme profile, which by then is 22% over 12900k "only", and the further undervolting or power limit tweaks to get the 39k marks is where the K SKUs should be, able to let the power users unlock the potential based on their own silicon lottery.

 

[YSCCC]

And yes, say I'm doing F@H (folding at home), even if I rma every 6 months ill still end up doing more work with a 13700 than a 7700x. That's how far ahead Intel is over amd.

That is just folding @ home where your downtime makes no extra cost, and for everyone else not doing F@H or crypto mining and use the PC to really work and earn money, downtime have a huge cost, more rendering time mayebe overnight isnt. that is why freaking longevity and stability is important. Not to say for the more popular SSF builds, RMAing means bloody tear down evey single bit and just to reassemble just when the RMA is back

 

[TheHerald]

You flat out do not understand English no? I am saying workload type and you are going to wattage.

For example, if you go benchmarking in very heavily threaded workloads, the extra E cores, higher frequency and cache can have you 40%, but going single core, or gaming, it is single digit performance gain, that is what in english, the gain is workload dependant, nobody every say it's wattage dependant.

Intel Core i9-13900K vs. Core i9-12900K: worth the upgrade? | Digital Trends

Intel's latest flagship, the Core i9-13900K, brings up the core count and offers various upgrades. How does it compare to the Core i9-12900K?

www.digitaltrends.com

I have been talking about MT this whole time. You too actually. I can point you to your previous posts. Now you are moving the goalposts. Why? Man...please, be reasonable.

No, they are comparing using the 0x129 bios with intel bios, it is slower than the old microcode for itself. So if they use the performance bios profile it can get 22% slower than the old unlimited profile, which is the 40% performance gain comes from.

Even using extreme profile, in my own 14900k, with old microcode and unlimited wattage in the only run I tested when it released before my undervolting tuning, it got 40k R23 using air cooler, compared to 27k for 12900k, whichis 48% increase, going for the 0x129 microcode bios and intel extreme profile without further undervolting for each step and trial for stability more than just able to run cinebench 10min test, gives me 32-33k marks, which is... 22% more than the stock 12900k. now that is close to the 15% zen 4 vs zen 5, not to say it seems the windows update is pumping up the zen 5 performance also. And mind you, for a normal release, the 13900k/14900k should be released in the 33k marks state using the "fixed" (if it really fixes) microcode and intel extreme profile, which by then is 22% over 12900k "only", and the further undervolting or power limit tweaks to get the 39k marks is where the K SKUs should be, able to let the power users unlock the potential based on their own silicon lottery.

Bud, the quote you used is from a link using jay2c review (here is the link)

https://www.notebookcheck.net/Asus-...with-23-loss-in-some-benchmarks.873898.0.html

In cinebench which you are quoting the 14900k was actually faster with the microcode update (0x129). Just click on jay2c review and youll see it for yourself

You are scoring 32-33k at 125w. Which is 37.5% faster than the 12900k at 125w (~24k). You are also faster at 125w than the 12900k at 240w. So the gen on gen increase is HUGE, for both performance and efficiency. AMD's performance / efficiency increase after 2 years is no where near. Why are you trying to argue otherwise? I don't get it. It's obviously intel made a huge step forward with 13th gen within 1 year, amd didn't.

That is just folding @ home where your downtime makes no extra cost, and for everyone else not doing F@H or crypto mining and use the PC to really work and earn money, downtime have a huge cost, more rendering time mayebe overnight isnt. that is why freaking longevity and stability is important. Not to say for the more popular SSF builds, RMAing means bloody tear down evey single bit and just to reassemble just when the RMA is back

Of course your downtime costs extra in F@H. You are not doing any work on your downtime man. But even with a downtime for RMA every 6 months, youll end up doing more work with an i7 over an r7. Does that seem reasonable to you? Isn't it absolutely ridiculously insane how big the performance gap is between amd and intel?

 

[YSCCC]

I have been talking about MT this whole time. You too actually. I can point you to your previous posts. Now you are moving the goalposts. Why? Man...please, be reasonable.

MT have a lot of different workload, some only able to efficiently utilize 8 threads, some 16, 32 or even more, so it is software dependant, and not always 40%, so that means it is not only gen on gen 40%, but 40% dependant on workload.

You are scoring 32-33k at 125w. Which is 37.5% faster than the 12900k at 125w (~24k). You are also faster at 125w than the 12900k at 240w. So the gen on gen increase is HUGE, for both performance and efficiency. AMD's performance / efficiency increase after 2 years is no where near. Why are you trying to argue otherwise? I don't get it. It's obviously intel made a huge step forward with 13th gen within 1 year, amd didn't.

Nope, I checked and make sure that was the extreme profile and hwinfo 64 showed it was using 253W PL1=PL2, the 33k is because when it used the built in VID table, during cinebench it spikes to 1.48v, which then W= voltage x current makes it capped the 253W at a low frequency, thus 33k, all I did to make it work is to lower the LLC below intel spec, hard undervolted it to make it cap at 1.38v instead and it is able to run at 39k again. The 1.48v was Intel spec for my sku and 1.38v is my spec fine tuned, not intel's.

The gen on gen improvement for AMD is more mild because
1) core parking issue, which from recent news the coming windows update could bump it up quite abit, we will wait and see

2) I never said it is exceptional, it is fair gain gen on gen and no sane ppl should upgrade every gen, whant I am saying is, the leap on MT only gen on gen only appears in intel since 11th gen, and they basically just stagnated since Sandy bridge through 9700k for the i7 in 7 freaking gen.

https://www.reddit.com/media?url=ht...bp&s=decd8169742c7bc7d378c2a12fe31f72bdfb27e9

and guess what? if puget is selling similar nos. of intel systems, and within brand it is assuming the bias isn't too big, 11th gen failure rate tops everything else by a large margin, and IIRC, 11th gen got qutie a performance hit due to some security bug fix after ADL launched, ADL is a success, but RPL just don't stop degrading prior to 15th gen is released,

As I said, as much RPL gives me performance, it also brings along a troublesome if not painful cycle of following tech news closely just to tune/update it and hope nothing hidden is killing it, it doesn't matter if it is the necessary evil to get the 40% full MT performance, at least it is what it caused to release it in the timeframe with the quality control slip to allow such self degrading VIDs to be built in. A big gen leap is meaningless if it just breaks.

 

[TheHerald]

MT have a lot of different workload, some only able to efficiently utilize 8 threads, some 16, 32 or even more, so it is software dependant, and not always 40%, so that means it is not only gen on gen 40%, but 40% dependant on workload.

If a workload is only utilizing 30% of your chip you don't use that workload to measure performance differences gen on gen. That's painfully obvious is it not?

Nope, I checked and make sure that was the extreme profile and hwinfo 64 showed it was using 253W PL1=PL2, the 33k is because when it used the built in VID table, during cinebench it spikes to 1.48v, which then W= voltage x current makes it capped the 253W at a low frequency, thus 33k, all I did to make it work is to lower the LLC below intel spec, hard undervolted it to make it cap at 1.38v instead and it is able to run at 39k again. The 1.48v was Intel spec for my sku and 1.38v is my spec fine tuned, not intel's.

Then there is something wrong with your bios. I mean to get 33k your pcores were running at below 5ghz, obviously you don't need 1.48v for 5 ghz.

Are you using your motherboards faulty bios to compare gen on gen performance improvement? I was literally getting 32k on a stock 13900k capped to 125, obviously something is way off on your mobos bios.

There is no performance increase due to Windows. That's kind of click bait nonsense. 1usmus himself (yes, he is the authority) found margin of error differences on zen 5 with the windows update. Stop taking seriously the youtube clickbaiters. They have no freaking clue.

 

[YSCCC]

If a workload is only utilizing 30% of your chip you don't use that workload to measure performance differences gen on gen. That's painfully obvious is it not?

Nope, that's plain out of context, you compare gen on gen on the workload scenarios during what is the relevant workload you will use vs what the chip is released is what is relevant, say if one both game a lot and do some rendering, depending on the rendering software you use, you will reference to that metric for purchase decision, the rest are irrelevant, so even with different benchmarks, the gen on gen performance could vary from 25-40% (which is still very high on the low end). But 40% is relevant to only tiny amount of ppl.

If a workload is only utilizing 30% of your chip you don't use that workload to measure performance differences gen on gen. That's painfully obvious is it not?


Then there is something wrong with your bios. I mean to get 33k your pcores were running at below 5ghz, obviously you don't need 1.48v for 5 ghz.

Are you using your motherboards faulty bios to compare gen on gen performance improvement? I was literally getting 32k on a stock 13900k capped to 125, obviously something is way off on your mobos bios.

There is no performance increase due to Windows. That's kind of click bait nonsense. 1usmus himself (yes, he is the authority) found margin of error differences on zen 5 with the windows update. Stop taking seriously the youtube clickbaiters. They have no freaking clue.

nope, that was the stable bios and have normal 1.2xv on the 12700KF, that was what the intel extreme profile pumps in for AVX, buildzoid's hellish bin get 1.55v for full load also, checked the AC loadline and the VID table and power limit is exactly calling the data built in the 14900k itself, as I said, that is what intel have to do in order to have enough top SKUs out in the wild for release and not the very rare top binned parts, 1.48v was peak though, sustained the default vid gets stable at 4.9-5ghz at 1.45v which is still 0.13v above what it can stay perfectly stable if tweaked well. I am very sure and watched the hwinfo64 throughout the test of the intel extreme profile right out of the gate and it sustained 253W, not 125W. That profile was precisely certified by intel, and then when doing subsequent manual tuning, 39k was back but not under intel VID table anymore, and it is stable for my workload, it might fail under some more demanding long duration workload and it could be why intel is releasing such high VID.

 

[TheHerald]

Nope, that's plain out of context, you compare gen on gen on the workload scenarios during what is the relevant workload you will use vs what the chip is released is what is relevant, say if one both game a lot and do some rendering, depending on the rendering software you use, you will reference to that metric for purchase decision, the rest are irrelevant, so even with different benchmarks, the gen on gen performance could vary from 25-40% (which is still very high on the low end). But 40% is relevant to only tiny amount of ppl.



nope, that was the stable bios and have normal 1.2xv on the 12700KF, that was what the intel extreme profile pumps in for AVX, buildzoid's hellish bin get 1.55v for full load also, checked the AC loadline and the VID table and power limit is exactly calling the data built in the 14900k itself, as I said, that is what intel have to do in order to have enough top SKUs out in the wild for release and not the very rare top binned parts, 1.48v was peak though, sustained the default vid gets stable at 4.9-5ghz at 1.45v which is still 0.13v above what it can stay perfectly stable if tweaked well. I am very sure and watched the hwinfo64 throughout the test of the intel extreme profile right out of the gate and it sustained 253W, not 125W. That profile was precisely certified by intel, and then when doing subsequent manual tuning, 39k was back but not under intel VID table anymore, and it is stable for my workload, it might fail under some more demanding long duration workload and it could be why intel is releasing such high VID.

Man, you cannot possibly think that 1.48v is normal for 5 ghz. It has nothing to do with intel or bins. Your bios is just faulty. My 12900k needs 1.15v and a 14900k needs 1.48? Yeah, not buying it.

Even the worse bin in the history of cpu's won't require 1.48v for 5 ghz. That's absolutely absurd. 5ghz should be stable at less than 1.15 actually. Something is way off with your bios. If your vid shows 1.48v for 5 then what does it show for 5.7? 2 volts? Can you take a picture of your vid table please? There is no freaking way that's correct unless your bios is just terrible.

And you are saying that by undervolting you can only get it stable at 1.35? If that is the case your cpu has heavily degraded. 1.35 at 5 ghz is crazy levels of degradation. I'm doing 4.9 at 1.06v on a 12900k. You need to RMA the chip in this case.

 

[Pierce2623]

That's fair for us geeks, but TBF a lot of ppl just don't want to spend the time and have more than enough money to pay for the whole system and use their warranty and updoor repair service if anything happens, even a father with spare moeny to spoil the son a bit can have the prebuilt systems work better than trying to learn and deal with issues.

Especially say for the degradataion issue, tons of DIY geeks have been hair pulling and tweaking back and forth thinking it's them doing something wrong so the working machine just don't work, when these happens, the prebuilt could be a stress relief and let them pull the hair. And tbh, from the past track records sens the disasterous 13th and 14th gen intel, a prebuilt running some 80C in gaming workload just survive well beyond like 5 years unless you let the fan clot with tar and spiderwebs, which for a gaming PC is at it's final days of great gaming performance days anyway.

You’re definitely not wrong. I just built my first PC as a 9 year old in 1994 and it only took like an hour. If a 9 year old can build a computer in an hour way back when they were MUCH LESS STANDARDIZED then surely any numpty can do it in an hour now? My last pc without a radiator only took about 20-30 minutes. Obviously adding radiators slows things down a bit even if it’s an AIO but most people don’t use them. Heck I’ve considered switching back to air coolers but Raptor Lake and Zen4 quickly disabused me of that notion.

 

[YSCCC]

Man, you cannot possibly think that 1.48v is normal for 5 ghz. It has nothing to do with intel or bins. Your bios is just faulty. My 12900k needs 1.15v and a 14900k needs 1.48? Yeah, not buying it.

Even the worse bin in the history of cpu's won't require 1.48v for 5 ghz. That's absolutely absurd. 5ghz should be stable at less than 1.15 actually. Something is way off with your bios. If your vid shows 1.48v for 5 then what does it show for 5.7? 2 volts? Can you take a picture of your vid table please? There is no freaking way that's correct unless your bios is just terrible.

And you are saying that by undervolting you can only get it stable at 1.35? If that is the case your cpu has heavily degraded. 1.35 at 5 ghz is crazy levels of degradation. I'm doing 4.9 at 1.06v on a 12900k. You need to RMA the chip in this case.

you do have reading difficulty.. that was peak voltage recorded in hwinfo, I dun know if it's that +0.8v spike as buildzoid experienced in his, average it is around 1.4-1.42v in default, intel extreme profile.

and that I can get it stable at 1.32v at all core 5.7, I didn't stupidly tune down the multicore so in multi core it stayed stable at 5.7 and all core 5.3-5.4 at R23 at 39k score. don't twist my word.

 

[YSCCC]

You’re definitely not wrong. I just built my first PC as a 9 year old in 1994 and it only took like an hour. If a 9 year old can build a computer in an hour way back when they were MUCH LESS STANDARDIZED then surely any numpty can do it in an hour now? My last pc without a radiator only took about 20-30 minutes. Obviously adding radiators slows things down a bit even if it’s an AIO but most people don’t use them. Heck I’ve considered switching back to air coolers but Raptor Lake and Zen4 quickly disabused me of that notion.

It actually didn't matter much, I did similar at 12 years old when I got the budget to do so, and back then the most time consuming part was formatting the harddisk and installing windows which takes forever... but still, assembling it is one thing, go do the settings, driver installation is something quite a lot of ppl just don't bother to learn and do themselves, I personally was a game freak and building myself is fun yet it gives more C/P ratio so learning was good, but for those who themselves don't really interested in such learning buying prebuilt with a small change in their bank does make sense, it's just like paying garage to do the checking and spark plug/oil change make sense for a lot of ppl, not us, but we couldn't rule out that value to them. So personally I will always build my own, but for those who have no clue of trouble shooting anything and have spare money I always just suggest them to buy prebuilt and forget the trouble, call those 24hr hotline but not call me 24hr is nice.

 

[TheHerald]

you do have reading difficulty.. that was peak voltage recorded in hwinfo, I dun know if it's that +0.8v spike as buildzoid experienced in his, average it is around 1.4-1.42v in default, intel extreme profile.

and that I can get it stable at 1.32v at all core 5.7, I didn't stupidly tune down the multicore so in multi core it stayed stable at 5.7 and all core 5.3-5.4 at R23 at 39k score. don't twist my word.

You literally said the default vid at 4.9-5 ghz is 1.45v, how am i twisting your words?

If you are scoring 33k at 253w default then there is obviously something wrong, that's all im saying.

 

[YSCCC]

You literally said the default vid at 4.9-5 ghz is 1.45v, how am i twisting your words?

If you are scoring 33k at 253w default then there is obviously something wrong, that's all im saying.

I only tested once at the intel default setting at 0x129, yea we all know something is wrong, but that is what INTEL said it should run at for that particular 14900k, which can do all those compression test, pass 30min R23 at the undervolted for 1.2xv all core 5.4 39k mark setting if I undervolt it myself, that is intel who build those VID inside, not me nor the mobo. and what buildzoid have, averaging 1.3x v after his tuning in 0x129, still have transient peaks at 1.55v which cannot be measured by the software monitor, that is what the degradation is. I don't know if it was intel's new code logic doing the initial training making all those crazy voltage or it is CEP kicking in, or let step aside, at 0x125 and 0x129, quite some reported 36k instead of 39k before undervolting, that is still a 8% performance drop.

So, after 2 years, cooporating with the board vendors and enforce their settings, and still I have to dail in all those constrains to get what we saw it should perform in all those reviews with a massive hike from 27k of 12900k to 39k for 14900k, don't you think those who experienced all these should feel distrust in intel and say we would not recommend anything intel in short term?