News All the pains of Intel: from CPU design and process technologies to internal clashes and political pressure

[bit_user]

Rocket lake is still pretty freaking good for my use case, although I have a standard 13900, and I keep wanting to upgrade,

Uh, I think you meant to say Raptor Lake.

shoot when you could get a 265K with Mobo and Ram for around $300 I almost pulled the trigger, but I am really looking for a solid upgrade,

Yeah, going from 32 threads to 20 threads... if you do heavily-multithreaded stuff, that might be more of a side-grade.

Meanwhile, if you simply upgrade the cooling on your i9-13900, then you can increase your power limits and squeeze more performance out of it without voiding any warranty (Intel doesn't consider power-limit tweaking to be overclocking).

even with an old GTX1080 (Although i have a 2080 TI in a box I should install).

IIRC, both are set to stop getting driver updates, starting in a few months. At least the GTX 1080 had a lower TDP, but if you restrict the RTX 2080 Ti to the same TDP, I think it would be more efficient.

we got rocket refresh,

Okay, yeah, you're definitely confusing Rocket Lake (Gen 11) with Raptor Lake (Gen 13-14). Maybe that's why you got such a skeptical reaction.

 

[bit_user]

If you call having to turn down your overclocking a bit defective,

That's a mischaracterization of the issue. Intel issued microcode fixes which did a lot more than restrict OC. CPUs failed which were definitely not OC'd.

https://www.tomshardware.com/pc-com...releases-0x12f-update-to-fix-vmin-instability
​

 

[cyrusfox]

Okay, yeah, you're definitely confusing Rocket Lake (Gen 11) with Raptor Lake (Gen 13-14). Maybe that's why you got such a skeptical reaction.

Yes 100%, too many lakes... Intel needs to do to cpu what they did to GPU, lets give it fantasy names rather than locations which are equally meaningless but easier for me to remember(as well as logical with alphabetical progression). Raptor is what I meant, I did have Rocket lake too though, forgot who I gave that 10850K to, but alder and raptor were quite the upgrade.

IIRC, both are set to stop getting driver updates, starting in a few months.

I am playing items that are 10+ years old or steam, I am good with legacy driver support, as long as productivity apps can still utilize it, good enough for me. Maybe I will switch to a B580 or its successor, I have had a couple opportunities to grab for cheap, assuming Intel doesn't cancel their GPU efforts.

 

[thestryker]

I did have Rocket lake too though, forgot who I gave that 10850K to

That'd be Comet Lake 🤣

IIRC, both are set to stop getting driver updates, starting in a few months.

Pretty sure we have no EOL on Turing yet and I doubt it'll be soon. If anything Pascal losing driver support would be another reason to install the 2080 Ti!

 

[bit_user]

Pretty sure we have no EOL on Turing yet and I doubt it'll be soon. If anything Pascal losing driver support would be another reason to install the 2080 Ti!

Oh, it was Volta, not Turing. I knew there was one after Pascal that was also getting dropped. Okay, so that's only Titan V, but @cyrusfox take heed: it's just one year older than the Turing cards!

https://www.tomshardware.com/pc-com...be-the-last-to-support-gtx-900-and-1000-cards
​

 

[thestryker]

it's just one year older than the Turing cards!

I think Turing being first RTX combined with the last new card release being 2022 will buy it more time than we'd otherwise see.

 

[TerryLaze]

That's a mischaracterization of the issue. Intel issued microcode fixes which did a lot more than restrict OC. CPUs failed which were definitely not OC'd.

​

https://www.tomshardware.com/pc-com...releases-0x12f-update-to-fix-vmin-instability​

​

The link you linked has nothing to do with your quote so I fail to understand why you posted it.

Instability is not the same as degradation, they can be related sometimes but they are not forced to be, fixing a system crashing after days of constantly being on has zero relation to the degradation issue (unless proven otherwise in the future) .
According to the article you posted intel fixed the degradation with update 0x12B.

What you are doing is as if I where saying that ryzen crashing, or outright not booting up anymore, due to incompatible ram is proof that ryzen CPUs still blow up.
They are separate issues.

 

[bit_user]

According to the article you posted intel fixed the degradation with update 0x12B.

This is like the 3rd or 4th microcode fix they've issued for it. They told us the previous one fixed it, too. And some people have seen crashes since they applied 0x21f, on systems that were previously stable. So, I'm not convinced this is the last word on it.

Also, that article has links to their prior patches and announcements. As such, it provides useful context.

 

[TerryLaze]

This is like the 3rd or 4th microcode fix they've issued for it. They told us the previous one fixed it, too. And some people have seen crashes since they applied 0x21f, on systems that were previously stable. So, I'm not convinced this is the last word on it.

Also, that article has links to their prior patches and announcements. As such, it provides useful context.

Again......instability is not the same as degradation.

 

[bit_user]

Again......instability is not the same as degradation.

Accepted. Though, even if the 0x21f microcode merely exposes instability on degraded CPUs, that's still not a positive development and one which might not stand.

 

[stuff and nonesense]

My argument is simple.
AMD did not and could not "beat" Intel alone. TSMC played a major, if not dominant role in AMD's performance improvements relative to their intel counterpart chips in the consumer area. Take away TSMC and AMD would have had Samsung and no 3D cache. And would trail.

AMD had sold off their fabs by the time of Ryzen, GloFlo used a Samsung licenced process for Ryzen 1000.
AMD were looking for better tools to make their processors, GloFlo (an independent fab company) were not going to invest in cutting edge tech, its expensive and didn’t fit their business model.
AMD had a choice between Samsung and TSMC. TSMC was the better option. No idea about what went on between AMD and TSMC but they got together. I can only imagine that there was a degree of information transfer wrt what libraries TSMC had available and in development that influenced how AMD implemented later versions of Zen cores and now Zen IO dies.

It takes 2 to get the best out of a partnership. Over the past 4 generations of Zen silicon that partnership has worked with improvements each release (some greater than others).

AMD are fabless, they need their suppliers to be on the ball. TSMC were at least comparable with Intel at 14nm and better since. TSMC have executed reliably. Intel faltered.

Could AMD have done as well without TSMC? Not really. Their old partner, Glo Flo chose not to compete and Samsung dropped by the wayside and developed in their own direction.
AMD chose the best available toolkit and it’s only got better.