Intel Thread

[Zoraptor]

I imagine it will be used a very great deal, at userbenchmark.

[Bartimaeus]

290W for an 8-core 125W TDP CPU seems...not ideal. Guess that's why it has to shut itself down every 25 seconds for 5-10 so it can not go into a nuclear meltdown. Although if you're running a crappy 5 second benchmark like userbenchmark, then I'm sure the results will indeed look great, .

(e): Intel subreddit is calling it "Bulldozer Lake" and the "11700FX". Ouch.

Edited March 6, 2021 by Bartimaeus

[Keyrock]

The new architecture is a legitimate step forward, but it's being held back by the now antiquated 14 nm process. Intel will be in good shape once they get their engineering fabs back on track, until then they can only do so much to overcome AMD's TSMC's manufacturing process lead.

[ComradeYellow]

Yeah I kinda regret being an Intel fanboy for a while but I'm stupid and stubborn at times.  AMD clearly has the snappy crunch power in their CPU's.  Being an Intel fan reminds me of being a fan of American exceptionalism, definitely has the money and power but performance wise.....gonna have to reform if they want to stay on top of the game

Too bad about those shortages though

[Zoraptor]

2 hours ago, Bartimaeus said:

(e): Intel subreddit is calling it "Bulldozer Lake" and the "11700FX". Ouch.

What next, "I was really hoping that Intel would at least be competitive this gen so AMD would be pressured to drop prices"?

2 hours ago, Keyrock said:

The new architecture is a legitimate step forward, but it's being held back by the now antiquated 14 nm process. Intel will be in good shape once they get their engineering fabs back on track, until then they can only do so much to overcome AMD's TSMC's manufacturing process lead.

That's a bit harsh on 14nm- the new chip has major weaknesses which were strengths for previous 14nm offerings. Latency for example, where Rocket Lake is little short of awful compared to Cannon Lake. It's not that 14nm isn't antiquated, but there clearly are architectural problems with the backport despite the IPC gains, hence the poor gaming performance compared to a 9900k even with a theoretical 10%+ clock v clock lift.

Ultimately the situation isn't because of 14nm showing its age so much as Intel's decision to tie architectural advances to process, and then 10nm being so problematic for so long.

[Bartimaeus]

Actually...yes: I literally read that sentiment in the same thread multiple times, . And that's the Intel fan subreddit!

Edited March 6, 2021 by Bartimaeus

[Jozape]

290W is a lot compared to the alternative products, but when the 11700k is pulling that much (using AVX-512) it is destroying everything else in performance, so that's actually not bad. Of course, the 104 degrees Celsius peak (again, using AVX-512) on an expensive air cooler is still horrifying.

225W is what they recorded under more normal heavy loads that use AVX2, which is significantly more common than the irrelevant (for now) AVX-512. That is in line with other 8-core APUs on the same node like the 10700k and 9900k.

[Malcador]

On 3/6/2021 at 6:55 PM, Bartimaeus said:

Actually...yes: I literally read that sentiment in the same thread multiple times, . And that's the Intel fan subreddit!

That place is a bit overrun with the AyyMD crowd.  And again regardless, it's Reddit, go read Anandtech forums or something.  At least there some people do realize CPUs have uses outside gaming

Edited March 8, 2021 by Malcador

[Zoraptor]

Most of the AyyMD crowd is banned from the Intel sub. Rocketlake just doesn't really seem to have any big positives to balance out the obvious negatives. The Intel hype machine is also starting to annoy a lot of Intel loyalists who have been waiting for a new desktop architecture for 5 years (!) and were actually expecting the claimed 19% IPC gain.

But anyway, how many uses does rocketlake beat Zen3 at even excluding gaming though? They're only doing 8 and 6 core versions so they're very much in the 'gaming' market segment, too expensive for the budget office PC type set up and for anything multithreaded it gets brutally murdered by the higher core counts of 5950/5900. Its main advantage is AVX512- which has very few applications while turning the CPU into a blast furnace- and having an iGPU that can be leveraged for some tasks and is pretty decent.

Don't think there's confirmed pricing out (?) but the indications from the Anandtech article is that they will be more expensive than cannonlake but still perform worse than Zen3, and in some uses regress from cannonlake. There is certainly a decent case for some of the cannonlake cpus as a budget option still since their price/ performance is good.

[Malcador]

1 hour ago, Zoraptor said:

But anyway, how many uses does rocketlake beat Zen3 at even excluding gaming though?

Well, other than AVX-512 using some code that an Intel AVX expert optimized probably none.  But the point was scorning Reddit for hardware discussion. 

Sort of funny how Anandtech is getting grumpy comments from others for their early review of it, although I guess can understand some people who may not bother reading other reviews.   Well and HWUB pulling a YongYea and reading Anandtech's review

[Keyrock]

Kind of a harsh and clickbaity title, but this does confirm the early Anandtech review.

tl;dw: Really hard to see the value of buying this in almost any scenario over Zen3 or Comet Lake. Intel does cut into AMD's lead in production workloads, but AMD still leads, even when core counts are even. This does nothing to move the needle on gaming, even slightly regresses vs Comet Lake in some instances.

It's weird seeing Intel at a process deficit. I've been building my own PCs for over 2 decades and Intel ALWAYS had the process lead, even when AMD had better chips, like the good ol' Athlon 64 days. Intel really needs to get their fab situation in order so they can compete with AMD TSMC on more even footing.

[ComradeYellow]

I see the word "Sand" flying around a lot lately...

[Totally not Gorgon]

Basically just wait a year or so for the chip shortage to subside. 

[Malcador]

https://www.anandtech.com/show/16573/intels-new-strategy-20b-for-two-fabs-meteor-lake-7nm-tiles-new-foundry-services-ibm-collaboration-return-of-idf

2023 planned for their 7nm chip.

[Malcador]

Anandtech's full review - https://www.anandtech.com/show/16495/intel-rocket-lake-14nm-review-11900k-11700k-11600k

The i5 doesn't look too awful, depending on circumstances.

[Keyrock]

Yeah, the 11600K seems to be the one decent value in this launch. The 11700K seems like a poor value and the 11900K makes zero sense whatsoever. I mean, flagship chips are always overpriced, but this is borderline robbery.

[LadyCrimson]

I sure am glad I bought my 9900K.  It's going to last me a long time.  Reminds me of buying the 920 actually - glad I bought that one when I did, too.

Sometimes it's nice to not be a hype-train or benchmark comparing/caring person.  

[Keyrock]

1 minute ago, LadyCrimson said:

I sure am glad I bought my 9900K.  It's going to last me a long time.  Reminds me of buying the 920 actually - glad I bought that one when I did, too.

Sometimes it's nice to not be a hype-train or benchmark comparing/caring person.  

That Bloomfield i7-920 processor was legendary, it had ludicrous staying power at or near the top of performance charts.

[LadyCrimson]

^ I'm pretty certain if I installed Win10 on my old PC, it would still be overall fine for most of my personal gaming needs, even most newer games at medium/60 ... at 1080. Except maybe Cyberpunk. 

But I don't do 1080 anymore.

[Malcador]

https://videocardz.com/newz/intel-iris-xe-dg1-discrete-gpu-from-asus-has-been-tested

Not bad for what it is.

[Keyrock]

Intel's DG2 GPU sampling to customers

I don't expect it to compete with Nvidia's or AMD's high-end, but I'll be interested to see where Intel's discrete graphics stand as they get their feet wet in the market. The optimist in me is thinking RTX 3060 TI or RX 6700 XT level of performance for their top level offering.

[Malcador]

https://www.anandtech.com/show/16886/intel-video-cards-get-a-brand-name-arc-coming-q1-2022

Should use those codenames as the actual product name.

 

[Zoraptor]

I'd argue that their CPUs need a (new) brand name more than their GPUs do. Xe is fine- OK, it reminds you of Blackwater constantly changing their name to escape bad press, but on the positive it also reminds you about a handy currency conversion website. In contrast even Intel engineers and PR people doing presentations have no idea what their CPU branding actually means any more.

[Keyrock]

I don't much care about the branding, they could call it Oomba Goomba for all I care. 

On a serious note, there's an opportunity for Intel to capture a piece of the market that's severely underserved right now, the low to mid range dedicated market. In the current gen we have APUs then there are expensive discrete GPUs, very expensive discrete GPUs, and ludicrously priced discrete GPUs. There's a gap there for discrete GPUs in the $150 - $250 range that offer a bit more oomph than APUs.

[Zoraptor]

Hard for Intel to do that though. The modern TSMC nodes are all expensive and oversubscribed, and that is what they're using so the pressures will be the same as they are for AMD- to use your limited capacity to make the stuff that sells for the most money first. And that isn't budget lines.

At the moment that pricing is still the natural home of the 14nm era Polaris and 16xx cards as it has been for years. They'd still be being produced but there's no/ little GDDR5 available any more (??? IIRC it's almost entirely deprecated but I can't remember the source of that...)