Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

VorpalFish posted:

Would still be a weird move if true. Why shave 2 cores from the top end SKU shortly after introducing 10c to mainstream desktop? If they actually price at $400 it should be plenty competitive against the 5800x I guess? But the supposed leak says i7 under $400 where they could well call the x900 part an i9 so... weird all around if even accurate.

I'd imagine a mix of yield (hence price), thermals, and sales figures for the 10900k. 8c/16t is the "hole" in the AMD lineup right now, with the 5800X being an objectively bad deal. If they can hit $400 or lower with it, it'd be competitive there in a way that, say, a $500 10c/20t wouldn't be against either the 5800X or 5900X.

But yeah, still feels like they should have a 10c and maybe a 12c part in that lineup somewhere.

Adbot
ADBOT LOVES YOU

VorpalFish
Mar 22, 2007
reasonably awesometm

DrDork posted:

I'd imagine a mix of yield (hence price), thermals, and sales figures for the 10900k. 8c/16t is the "hole" in the AMD lineup right now, with the 5800X being an objectively bad deal. If they can hit $400 or lower with it, it'd be competitive there in a way that, say, a $500 10c/20t wouldn't be against either the 5800X or 5900X.

But yeah, still feels like they should have a 10c and maybe a 12c part in that lineup somewhere.

This is supposed to be a backport of Sunny Cove to 14nm++++++++, no? I would think yields would be... pretty good at this point.

Who knows I guess.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

I'd think so, too, but at that point why accept a core deficit vs Ryzen at all? It's very strange, if true.

gradenko_2000
Oct 5, 2010



Lipstick Apathy

It's probably a thermal issue. When AMD was still on 14nm, they topped out at 8/16 for mainstream desktop - they didn't go to 12-core Ryzen 9's until Zen 2. Intel probably threw 10 cores on the 10900 because they needed to have something to compete, but I'm thinking trying to do a 10-core Rocket Lake i9 would either be too warm, or would have to be backed-off too much on clocks.

ConanTheLibrarian
Aug 13, 2004


dis buch is late

Fallen Rib

Presumably the beefier cores require more surface area, so there's the silicon budget to consider too. They probably wanted to make sure their top end part had higher clocks than the previous generation to fend off accusations of generation on generation regression like we saw with the original cove core release.

K8.0
Feb 26, 2004

Her Majesty's 56th Regiment of Foot

Thermal is actually more of a problem the smaller the process gets. AMD hits some pretty hard thermal bottlenecks on 7nm chiplets. If you reduce your surface area, you reduce your cooling capacity. It's going to be a major problem for high performance computing as we drop down to smaller and smaller nodes and try to push them to high clocks.

My guess as to why 8c is that it probably wins them everything that is winnable. If they hit their IPC target, they'll be back in the driver's seat for gaming. However, there's nothing they can do to compete with AMD for highly multithreaded workloads. A 10c part is almost certainly not going to beat the 5950X, and their lead may not even be enough for it to beat the 5900X, especially without being a 500w part or something. It also leaves them room to release a 10c refresh like the 10900k, even if it will also have nothing to really offer vs the 11900k.

Wild EEPROM
Jul 29, 2011


oh, my, god. Becky, look at her bitrate.


They’re gonna save it for hedt because its not like theres any reason to buy intel hedt

Laslow
Jul 18, 2007



HEDT is dead as we know it is dead. Intel needs to rework the X chipsets as they were pre-X58. Basically higher end than the Z, but without using a snowflake/server socket to keep the CPU prices competitive. Maybe use it as the beta test platform for bleeding edge stuff like DDR5/PCIE5, just like they did with X48 with DDR3/PCIE2/UEFI.

If they time it right, they could beat AMD to having a DDR5 product to market. At least that’d give them something compelling, even if on paper.

Basically I’m waiting on DDR5 before my next rebuild, and anything to help that get out the door earlier even by a few months, that’d be swell.

DoomTrainPhD
Feb 12, 2009



Lmao at only 16MB of L3 cache.

Crunchy Black
Oct 24, 2017

CASTOR: Uh, it was all fine and you don't remember?
VINDMAN: No, it was bad and I do remember.




Laslow posted:

HEDT is dead as we know it is dead. Intel needs to rework the X chipsets as they were pre-X58. Basically higher end than the Z, but without using a snowflake/server socket to keep the CPU prices competitive. Maybe use it as the beta test platform for bleeding edge stuff like DDR5/PCIE5, just like they did with X48 with DDR3/PCIE2/UEFI.
See also 754 and 939 though; AMD's socket stack has pretty much always been simpler and more widely compatible. Not that I disagree with you! I can tell you from the inside that the W-SKUs are making the money they expected them to in high-end, long term contract visualization stuff that Intel won't be looking to completely axe it as a product line. Will it probably use the next server socket to save money? Almost assuredly.

Crunchy Black fucked around with this message at 05:45 on Oct 27, 2020

gradenko_2000
Oct 5, 2010



Lipstick Apathy

when did HEDT become A Thing between server-grade and desktop computers, anyway?

FuturePastNow
May 19, 2014




the Pentium Pro was Intel's first HEDT processor

gradenko_2000
Oct 5, 2010



Lipstick Apathy

more intel info!

https://twitter.com/aschilling/stat...814041380245511

https://twitter.com/momomo_us/statu...815529238089740

Cygni
Nov 12, 2005

raring to post



i think im gonna hold off on buying Zen3 until Rocket Lake launches just to see. figure the 8700 in my gaming PC is good enough to push a 3080 at 1440p in the meantime.

Gabriel S.
May 20, 2006

I WILL KEEP IMMIGRANT CHILDREN IN A LOCKBOX, AND THAT PAYS DOWN THE NATIONAL DEBT.

Put this Nazi Scientist fuck on ignore immediately!


I'm suspecting that we won't see availability until February-March? And I am really skeptical of those IPC improvements...

VorpalFish
Mar 22, 2007
reasonably awesometm

Gabriel S. posted:

I'm suspecting that we won't see availability until February-March? And I am really skeptical of those IPC improvements...

Skepticism is good when you're talking about companies talking about their own products, but ice lake and tiger lake both exist in the wild and have been independently tested at about +18% IPC, and given this is supposedly a backport of... Sunny Cove? To 14nm, there's no reason to think they couldn't/shouldn't be able to hit those numbers.

As always wait for independent reviews before making purchasing decisions.

Gabriel S.
May 20, 2006

I WILL KEEP IMMIGRANT CHILDREN IN A LOCKBOX, AND THAT PAYS DOWN THE NATIONAL DEBT.

Put this Nazi Scientist fuck on ignore immediately!


That makes more sense but the whole model naming that Intel is going with is freaking insane. I need a goddamn Visio diagram.

Cygni
Nov 12, 2005

raring to post



https://twitter.com/tum_apisak/stat...027256701616128

21% single core improvement, dang.

movax
Aug 30, 2008



Gabriel S. posted:

That makes more sense but the whole model naming that Intel is going with is freaking insane. I need a goddamn Visio diagram.

Dunno man, I think AMD wins this one with the dumb Zen 2/Zen 3 SKU poo poo they’re pulling.

Ihmemies
Oct 6, 2012




21% single core improvement compared to what? Like what product is X and which product Y is 21% better?

gradenko_2000
Oct 5, 2010



Lipstick Apathy

Ihmemies posted:

21% single core improvement compared to what? Like what product is X and which product Y is 21% better?

The leaked Rocket Lake results are scored 21% higher than the single-core results of an i7-10700K...

... but I'd still take these results with a boatload of salt, because UserBenchmark is a notoriously pro-Intel website, to the point of repeatedly reconfiguring their benchmark tests to suppress the ability of AMD CPUs to look good.

VulgarandStupid
Aug 5, 2003
I AM, AND ALWAYS WILL BE, UNFUCKABLE AND A TOTAL DISAPPOINTMENT TO EVERYONE. DAE WANNA CUM PLAY WITH ME!?


I thought the only thing that makes AMD look
good was blender.

Sent from my iPad
Jun 19, 2000



https://semiaccurate.com/2020/10/29...th-rocket-lake/

quote:

Now on to the really bad, this Rocket Lake part and the ‘new’ Cypress Cove core. We will admit we were wrong when we said that Intel would never backport a 10nm architecture to a 14nm process because that is exactly what Rocket/Cypress is, take Ice Lake and backport it. Also backport the Xe GPUs and slap them on for good measure.

What we should have said is that Intel would never be dumb enough to backport a 10nm architecture that was not designed with process portability in mind to a 14nm process. It would be too big, too inefficient, and simply not competitive with IP blocks designed and tuned for 14nm from the start. A company would need to be desperate and have their backs to the wall so badly that flushing tens of millions of dollars on this port for nothing more than a headline grab and an eventual black eye is worth it. That’s what we should have said but we would have been wrong there too, Intel it seems _IS_ that dumb.

Some Goon
Jan 6, 2013

A golden helix streaked skyward from the Helvault. A thunderous explosion shattered the silver monolith and Avacyn emerged, free from her prison at last.



Doesn't the semiaccurate guy like really hate Intel, or am I getting them mixed up with another semi- site?

WhyteRyce
Dec 30, 2001



I thought Keller was the one pushing for design flexibility and backporting before anyone gets all lol Intel is dumb but I might be extrapolating that from his design portable to other fans talk

Twerk from Home
Jan 17, 2009

This avatar brought to you by the 'save our dead gay forums' foundation.


Some Goon posted:

Doesn't the semiaccurate guy like really hate Intel, or am I getting them mixed up with another semi- site?

Yeah, dude has a semi for AMD.

WhyteRyce
Dec 30, 2001



Some Goon posted:

Doesn't the semiaccurate guy like really hate Intel, or am I getting them mixed up with another semi- site?

He does. Raging hate boner. Have to take what he says with a huge grain of salt. Everyone talks up his big hits (which often aren’t hits its him being wrong and spinning it like 10nm) but ignores his big misses

Cygni
Nov 12, 2005

raring to post



The performance leaks for Rocket Lake honestly look pretty good so he might need to be prepared for another mea culpa.

Whether it’s enough to handle Zen3 is a diff story, but we will see!

Twerk from Home
Jan 17, 2009

This avatar brought to you by the 'save our dead gay forums' foundation.


Cygni posted:

The performance leaks for Rocket Lake honestly look pretty good so he might need to be prepared for another mea culpa.

Whether it’s enough to handle Zen3 is a diff story, but we will see!

A competitive AMD causing Intel to pull out all the stops and put Hyper-threading on everything, release non-GPU versions for a few bucks cheaper, and up core counts has been really great for consumers, especially after 4 years of mostly stagnant CPU market.

VorpalFish
Mar 22, 2007
reasonably awesometm


This is a pretty dumb take. Backporting is expensive and time consuming but if they either don't have enough 10nm capacity or aren't able to get good enough yields for 8c+ dies on 10nm it's literally what they have to do. An 18% IPC uplift will give them a product that can compete with zen3 if they price aggressively enough, even if it is less efficient.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

Twerk from Home posted:

A competitive AMD causing Intel to pull out all the stops and put Hyper-threading on everything, release non-GPU versions for a few bucks cheaper, and up core counts has been really great for consumers, especially after 4 years of mostly stagnant CPU market.

Now when is Intel gonna up L3 cache?

I'd be real interested to see if something like Broadwell's eDRAM would be of any benefit. It was pretty compelling back then, IIRC, just maybe expensive to produce and slapped on DOA chip series.

gradenko_2000
Oct 5, 2010



Lipstick Apathy

DrDork posted:

Now when is Intel gonna up L3 cache?

I'd be real interested to see if something like Broadwell's eDRAM would be of any benefit. It was pretty compelling back then, IIRC, just maybe expensive to produce and slapped on DOA chip series.

anandtech did a recent retrospective on Broadwell and I think they found it still punches way above its weight due to the cache

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

gradenko_2000 posted:

anandtech did a recent retrospective on Broadwell and I think they found it still punches way above its weight due to the cache

Yeah, it was a pretty interesting way to add performance in some workloads. I wonder how fast they could push it based on a 14++++ process vs the 22nm one Broadwell used?

If Intel wanted to re-take the gaming crown (or at least compellingly break away from AMD's near-parity), something like a 11700k(f) with eDRAM and overclocking headroom for a good price might be a solid offering. Which I assume is why we won't see one anytime soon.

movax
Aug 30, 2008



So I can see back-porting not sounding completely loving insane because Intel owns the PDKs themselves and hopefully they interact with their RTL designers such that everyone's using their standard cell libraries outside of the hand-tuned register file / sense amps / analog bits and what not. At the same time, these are tiny, tiny processes and I wonder what team is being hit doing the physical layout and verification work in moving those blocks / designs to the 14 nm PDK, and what work that team is therefore not doing, unless they've staffed up for that specific job which in and of itself has its own set of problems.

Gabriel S.
May 20, 2006

I WILL KEEP IMMIGRANT CHILDREN IN A LOCKBOX, AND THAT PAYS DOWN THE NATIONAL DEBT.

Put this Nazi Scientist fuck on ignore immediately!


Dismissing release dates, when do we expect Rocket Lake to actually be commercially available?

With all of the recent delays, I'm thinking a paper launch in January but nothing until late February.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

Gabriel S. posted:

With all of the recent delays, I'm thinking a paper launch in January but nothing until late February.

Yeah, I think Feb-Mar has been the general consensus. Zen3 might have lit a bit of a fire under them, but I don't know if that'd be enough to actually drag the dates left any.

Cygni
Nov 12, 2005

raring to post



Gabriel S. posted:

Dismissing release dates, when do we expect Rocket Lake to actually be commercially available?

With all of the recent delays, I'm thinking a paper launch in January but nothing until late February.

March at best, imo.

Gabriel S.
May 20, 2006

I WILL KEEP IMMIGRANT CHILDREN IN A LOCKBOX, AND THAT PAYS DOWN THE NATIONAL DEBT.

Put this Nazi Scientist fuck on ignore immediately!


DrDork posted:

Yeah, I think Feb-Mar has been the general consensus. Zen3 might have lit a bit of a fire under them, but I don't know if that'd be enough to actually drag the dates left any.

Freaking lame. Releasing new PC parts in the dead of winter would have been perfect timing.

Arzachel
May 12, 2012


gradenko_2000 posted:

anandtech did a recent retrospective on Broadwell and I think they found it still punches way above its weight due to the cache

I just mentioned this in the AMD thread but Anandtech uses maximum supported memory frequency and JEDEC timings which is probably going to favor Broadwell.

Adbot
ADBOT LOVES YOU

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

Arzachel posted:

I just mentioned this in the AMD thread but Anandtech uses maximum supported memory frequency and JEDEC timings which is probably going to favor Broadwell.

Yeah, on a platform-as-a-whole perspective I don't think anyone would be recommending Broadwell anytime soon. But by keeping the performance lift from RAM somewhat controlled (it's still DDR3 vs 4 for newer platforms), I think it does give a bit better a sense of how interesting and strong those CPUs were.

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply