8700k Microcode Updates / Planned obsolescence. Cinebench r15, Win11

[Ebony Falcon]

44 minutes ago, DarkSmith2 said:

 

 

It was on a fresh/clean installed system, there was no load/other software running on the CPU during the tests, the test was done multiple times, also after restarts.

3 percent could be anything tho mate 

[Middcore]

1 hour ago, xg32 said:

4% is very significant imho, while ocing we get 1-2% increments.

 

You make OCing sound pretty pointless. 

[DarkSmith2]

14 minutes ago, Ebony Falcon said:

3 percent could be anything tho mate 

it showed constantly same results, there is absolutely no doubt that the microcode update is the culprit of the difference in performance.

Also the system idles at 0% load. My system doesnt randomly spike with "random" load, there is nothing running that could do that.

10 minutes ago, Middcore said:

You make OCing sound pretty pointless. 

Not really, i mean 1-2% per 100mhz seem to be alright in Cinebench, mind you the 8700k has only a 4.3GHz allcore boost without mce (mce is already considered OC).

[Guest]

30 minutes ago, DarkSmith2 said:

Not really, i mean 1-2% per 100mhz seem to be alright in Cinebench, mind you the 8700k has only a 4.3GHz allcore boost without mce (mce is already considered OC).

If you where 4.8ghz and higher all core manual, I'd call THAT overclocking. 

 

What you are doing is still withing the spec envelope. 4.3ghz is snooz fest. (No offence)

[Mister Woof]

58 minutes ago, Middcore said:

 

You make OCing sound pretty pointless. 

It depends.

 

If we're looking at a stock 8700k to a 5ghz overclocked one, there's a significant difference. Of course that's because there's a 700mhz difference, which is not insignificant at all. That's a 16%+ increase in frequency. 

 

In newer chips, it's much smaller. The 10900k will run at around 4900mhz if cooled adequately, and most don't overclock to more than 5200mhz without some exotic shit going on. In that case, you're looking at a much smaller 6% gain in frequency. 

 

EDITED: I cannot maths sometimes

[DarkSmith2]

19 minutes ago, ShrimpBrime said:

If you where 4.8ghz and higher all core manual, I'd call THAT overclocking. 

 

What you are doing is still withing the spec envelope. 4.3ghz is snooz fest. (No offence)

No i totally agree, just technically speaking "intel spec is 4.3Ghz allcore" .. which is by definition "stock" 
MCE exceeds this, therefore is considered overclocking.

[Guest]

2 hours ago, DarkSmith2 said:

No i totally agree, just technically speaking "intel spec is 4.3Ghz allcore" .. which is by definition "stock" 
MCE exceeds this, therefore is considered overclocking.

Eh, kind of. 

I meant anything past 4.7ghz, which we know the cpu is capable of all core with little effort.

[DarkSmith2]

57 minutes ago, ShrimpBrime said:

Eh, kind of. 

I meant anything past 4.7ghz, which we know the cpu is capable of all core with little effort.

yea outside of intel spec its also a stock setting with enabled MCE by some board manufacturers, its just the definition of things....
and how it is implemented / set up.

 

Things have changed since, all CPUs after Coffelake came with way higher allcore boostclocks with much lower room to overclock, maybe just because of this.
What they now fiddle with is powerstate's instead of clocks. I hope some day every board manufacturer will follow the same rules of stock settings, its all kinda fishy to look good in reviews tho.

[Mister Woof]

3 minutes ago, DarkSmith2 said:

yea outside of intel spec its also a stock setting with enabled MCE by some board manufacturers, its just the definition of things....
and how it is implemented / set up.

 

Things have changed since, all CPUs after Coffelake came with way higher allcore boostclocks with much lower room to overclock, maybe just because of this.
What they now fiddle with is powerstate's instead of clocks. I hope some day every board manufacturer will follow the same rules of stock settings, its all kinda fishy to look good in reviews tho.

Eh, at the end of the day what the chips are capable of is what matters. By default all Ryzen chips exhibit MCE-like behavior anyway. It's about time Intel officially just lets things loose.

[DarkSmith2]

17 minutes ago, Mister Woof said:

Eh, at the end of the day what the chips are capable of is what matters. By default all Ryzen chips exhibit MCE-like behavior anyway. It's about time Intel officially just lets things loose.

AMD boost behavior isnt exactly what i would want for Intel CPUs, dont get me wrong, its fine, but it also has downsides to it, the coreclock is switching rapidly, one of the cores will always drop in frequency quite harsh, this + quite alot of load/core swapping just to keep the stability and to reduce heat/power consumption. It also certainly does affect %lows on Ryzen CPUs negatively.

 

On the other hand, all core OC isnt quite viable on Ryzen, you'll burn your chip or have less performance on avg anyways

[YoungBlade]

16 minutes ago, DarkSmith2 said:

AMD boost behavior isnt exactly what i would want for Intel CPUs, dont get me wrong, its fine, but it also has downsides to it, the coreclock is switching rapidly, one of the cores will always drop in frequency quite harsh, this + quite alot of load/core swapping just to keep the stability and to reduce heat/power consumption. It also certainly does affect %lows on Ryzen CPUs negatively.

 

On the other hand, all core OC isnt quite viable on Ryzen, you'll burn your chip or have less performance on avg anyways

Now that Intel is going to be gluing their CPUs together, they might have to go the AMD boost behavior route to make it work. The cores aren't unified anymore. They don't even have identical architectures. Some cores lack certain instruction sets. That thread scheduler is likely going to be pulling the same sort of shenanigans by moving threads between P and E cores and turning off unused cores to save power. It may be more intelligent than what Ryzen does, but the result will probably look similar.

[DarkSmith2]

42 minutes ago, YoungBlade said:

Now that Intel is going to be gluing their CPUs together, they might have to go the AMD boost behavior route to make it work. The cores aren't unified anymore. They don't even have identical architectures. Some cores lack certain instruction sets. That thread scheduler is likely going to be pulling the same sort of shenanigans by moving threads between P and E cores and turning off unused cores to save power. It may be more intelligent than what Ryzen does, but the result will probably look similar.

[Guest]

2 hours ago, DarkSmith2 said:

yea outside of intel spec its also a stock setting with enabled MCE by some board manufacturers, its just the definition of things....
and how it is implemented / set up.

 

Things have changed since, all CPUs after Coffelake came with way higher allcore boostclocks with much lower room to overclock, maybe just because of this.
What they now fiddle with is powerstate's instead of clocks. I hope some day every board manufacturer will follow the same rules of stock settings, its all kinda fishy to look good in reviews tho.

Powerstates on a timer. I'm not keen for that stuff. 

 

I do miss the days of a CPU that has a stock frequency without boosting gimmicks. Overclock it, get X% higher scores/fps and everyone was happy.

 

Now (at least for AMD) tweak the PBO. Or XFR. Or SenseMi offset and skew. Or mess with PPT and EDC... to accomplish basically..........well....... nothing. Under-volting seems more common these days instead of making sure we have adequate cooling.... 

 

I'm keeping my eye on Intel 12th gen. Its looking pretty good so far (taking leaks with a grain of salt).

[DarkSmith2]

59 minutes ago, ShrimpBrime said:

Powerstates on a timer. I'm not keen for that stuff. 

 

I do miss the days of a CPU that has a stock frequency without boosting gimmicks. Overclock it, get X% higher scores/fps and everyone was happy.

 

Now (at least for AMD) tweak the PBO. Or XFR. Or SenseMi offset and skew. Or mess with PPT and EDC... to accomplish basically..........well....... nothing. Under-volting seems more common these days instead of making sure we have adequate cooling.... 

 

I'm keeping my eye on Intel 12th gen. Its looking pretty good so far (taking leaks with a grain of salt).

For my usecase, 1080p gaming, i dont really see a GPU i could match a 12th gen CPU up with. Nvidia 30series cards are crippled at low resolutions, no point in trying to get more performance out of them, 3080ti/3090 doesnt seem to scale pretty well, 6900xt is running circles around those at 1080p for alot of games i play... 

I mean look at this CoD 2019 Multiplayer FPS on an empty map: (5950x + rx6900xt, 1080p ULTRA)


You cant get this out of an 30series card, like at all. In this game in Multiplayer small maps a 2080ti is as fast as or faster than a 3090 and this is not because of CPU bottlenecks but because of GPU architecture, Warzone BR might get boosted, but i highly doubt that a 12900k would increase the performance meaningfully in Multiplayer even with a 6900xt.

This is with my 8700k + 2080ti :

 

Most people with 30 Series cards cant even touch those FPS at 1080p or 720p...

So no matter what i do, i probably have to wait for new GPUs to come by anyways.

[freeagent]

3 hours ago, DarkSmith2 said:

AMD boost behavior isnt exactly what i would want for Intel CPUs

I like Intel better for that.. They have finer consistency when they slap cores together.. they are all usually awesome. AMD only gives you a few good ones, that's why they boost like they do. I think its a bit lame. That is my biggest gripe with AMD ATM.. all cores are not equal with them. With Intel, if you have good cooling chances are you are getting all cores over 5GHz.. not with AMD you aren't.. unless you freeze it 

 

As for microcode updates.. I literally watched them steal GFlops from me. Back when Specter/Meltdown was causing a massive shitstorm for them. The timing was right.. just got a 3770K, tested overclocks with Linpack, GFlops were dropping, cores needed more volts for the same clocks, and I still lost 100MHz off the tippy top by the time they were done. This was over the course of about 6-8 months or so.. Ahh well.. its not like it just sucks now or anything, she still rips.

[DarkSmith2]

6 minutes ago, freeagent said:

I like Intel better for that.. They have finer consistency when they slap cores together.. they are all usually awesome. AMD only gives you a few good ones, that's why they boost like they do. I think its a bit lame. That is my biggest gripe with AMD ATM.. all cores are not equal with them. With Intel, if you have good cooling chances are you are getting all cores over 5GHz.. not with AMD you aren't.. unless you freeze it 

 

As for microcode updates.. I literally watched them steal GFlops from me. Back when Specter/Meltdown was causing a massive shitstorm for them. The timing was right.. just got a 3770K, tested overclocks with Linpack, GFlops were dropping, cores needed more volts for the same clocks, and I still lost 100MHz off the tippy top by the time they were done. This was over the course of about 6-8 months or so.. Ahh well.. its not like it just sucks now or anything, she still rips.

yea they had the pressure of the shitstorm.. so they probably had to do something radical. But it just went the wrong way. Spectre and Meltdown arent as dangerous as people tried or still try to imply... the recent microcode update thats within win11 is just crippling more yet again for unknown reasons.. i mean with the 8700k spectre and meltdown alone isnt even that big, but the microcode update seems to be really nasty.