TR Forums

https://www.anandtech.com/show/16535/in ... ocket-lake

Best feature — you might actually be able to buy it.

Downside — you probably won’t want to buy it.

IMO, Rocket Lake is a bit of hot mess. It seems like a compromise on trying to increase the adoption of AVX512 but Intel has to bump the clockspeed on a node process that it is at bleeding limit in order to keep up with Zen2-Zen3 SKUs. The underlying platform is more attractive then the CPU itself. It almost like Intel's version of "Bulldozer" a forward-thinking CPU design that didn't make sense initial depute but later gets vindicated somewhat but it is too late. The design really needs a node shrink to really stand a chance on overtaking Zen2-Zen3.

The architecture was originally intended for 10nm... and appears to be suffering after the back port.

The main problem with 10nm node is that Intel can't reach the clockspeeds that they need yet. That's what prompted the backport in the first place.

Looking at the benchmarks and nothing else, one must think that the 11700K is the 10700K with a couple optimisations and AVX-512. Are there ANY reasons to believe there's a new microarchitecture hiding within, until good guys with X-ray machines and small chainsaws pick the thing apart?

Wirko wrote:

Looking at the benchmarks and nothing else, one must think that the 11700K is the 10700K with a couple optimisations and AVX-512. Are there ANY reasons to believe there's a new microarchitecture hiding within, until good guys with X-ray machines and small chainsaws pick the thing apart?

On the third page "Microbenchmarks" look at the graph with the latency comparison between the 10700K and 11700K. The 11700k has worse latency than the 10700k. If anything, it's backsliding, which does not sound like the same architecture with a couple of tweaks -- it's more like a new architecture that expected to be able to leverage a smaller node.

Basically, Intel bet the farm on 10NM working, and lost. This mess is the result.

The good news for intel is that the processors people want to buy are in short supply. So anybody who needs to buy now might be willing to settle for intel. That could keep intel going strong for the next 12-18 months. that’s a lot more time to pull their head out of their a$$ than they deserve

You'd be shocked at how much Intel still has a rep in the server space, though. Even with AMD leaps and bounds ahead for damn near *EVERY* metric..."nobody got fired for buying Intel" is the new "nobody got fired for buying IBM" regardless of whether it's still true or not.

I don't think that holds as much water as it used to in light of Spectre and Meltdown fiascos. Afterall, they had impacted those the market the most. The security updates/patch to get around these problem has costed those customers millions in terms of performance loss and downtime. Even PHB-types will remember it because it hit the bottom line. I think will have far more trouble AMD delivering enough silicon to sate demand for this market. They are trying to validate Zen-3 based Epycs/Threadrippers as fast as possible.

Sure, you're right, I just made it up and my post was based on imaginary experience.

/s

Intel can lose their lead in server just like IBM did -- but also just like IBM, if they do lose it they will lose it sloooowwwwwlllyyyyy.

Waco wrote:

Sure, you're right, I just made it up and my post was based on imaginary experience.

/s

IIRC, you work in the government sector what happens there doesn't necessary apply to the private sector and vice versa. I have read a number of stories that IT Admins and departments getting burnt by Meltdown/Spectre fiasco when it went down hard. That's something doesn't go away overnight. It put a hard dent into "Nobody got fired for using Intel" mindset.

K-L-Waster wrote:

Intel can lose their lead in server just like IBM did -- but also just like IBM, if they do lose it they will lose it sloooowwwwwlllyyyyy.

Unless, Intel is able to make a significant turn-around soon. It'll happen sooner then you would expect. It isn't just AMD either. You got Apple and ARM trying to make headways in these markets.

Krogoth wrote:

Waco wrote:

Sure, you're right, I just made it up and my post was based on imaginary experience.

/s

IIRC, you work in the government sector what happens there doesn't necessary apply to the private sector and vice versa. I have read a number of stories that IT Admins and departments getting burnt by Meltdown/Spectre fiasco when it went down hard. That's something doesn't go away overnight. It put a hard dent into "Nobody got fired for using Intel" mindset.

I think you're very happily ignoring what I said in my original post.

Despite everything that has happened it still holds true. It took 20 years for the IBM mantra to fade. Intel isn't gone in 5.

Waco wrote:

Krogoth wrote:

Waco wrote:

Sure, you're right, I just made it up and my post was based on imaginary experience.

/s

IIRC, you work in the government sector what happens there doesn't necessary apply to the private sector and vice versa. I have read a number of stories that IT Admins and departments getting burnt by Meltdown/Spectre fiasco when it went down hard. That's something doesn't go away overnight. It put a hard dent into "Nobody got fired for using Intel" mindset.

I think you're very happily ignoring what I said in my original post.

Despite everything that has happened it still holds true. It took 20 years for the IBM mantra to fade. Intel isn't gone in 5.

I think you’re clearly right — there’s tremendous inertia, risk aversion, and herd mentality (kind of all related). But it’s also true that change happens slow then fast.

Intel needs to get back on track before the “fast” part kicks in. It could be close.

Agreed 100%.

If you read that Intel has taken delivery of a dozen or more new EUV lithography systems from ASML, that will be when Intel is back in the game.
https://www.tomshardware.com/news/tsmc-euv-tools-order
Rocket Lake is pretty amazing when you consider that Intel has once again gone back to Ye Olde 14 nm process - the same process node that Broadwell used in 2014.

Krogoth wrote:

K-L-Waster wrote:

Intel can lose their lead in server just like IBM did -- but also just like IBM, if they do lose it they will lose it sloooowwwwwlllyyyyy.

Unless, Intel is able to make a significant turn-around soon. It'll happen sooner then you would expect. It isn't just AMD either. You got Apple and ARM trying to make headways in these markets.

I think you're underestimating the effect of organizational inertia. Getting Finance Departments to change favoured-vendor status is no simple task....

K-L-Waster wrote:

Krogoth wrote:

K-L-Waster wrote:

Intel can lose their lead in server just like IBM did -- but also just like IBM, if they do lose it they will lose it sloooowwwwwlllyyyyy.

Unless, Intel is able to make a significant turn-around soon. It'll happen sooner then you would expect. It isn't just AMD either. You got Apple and ARM trying to make headways in these markets.

I think you're underestimating the effect of organizational inertia. Getting Finance Departments to change favoured-vendor status is no simple task....

It is done much more quicker when favored-vendor has a serious fault that has costed the organization millions in lost revenue through downtime and performance loss.