TR Forums

Chrispy_ wrote:

And yeah, before that was the Intel "Bulldozer" AKA Pentium 4 Willamette/Northwood etc. Good job AMD watched this long-pipeline monster tank horribly and didn't make one of their own. Oh. Nevermind, Ryzen is here, everything's okay again a mere 15 years later....

As I understand it, Intel aimed for a Ghz machine, while AMD aimed for a throughput machine.

They both got to the same (terrible) place, but AMD didn't just go "hey, pentium 4 was good, let's do that, but with way less of a budget."

As for Intel retiring the Core architecture (technically per Intel, Nehalem is technically the first Core microarchitecture part) I don't think it will happen. Not the way we think about it regularly.

We may see something where Core will be skewed either towards purely consumer, with a die reused for server/HPC products which demand lower core counts at maximum single-threaded performance.

And servers goes to a design that ignores trying to hit 3.5ghz+ to increase pref/watt and IPC in the 2.5-3.5Ghz range.

In terms of naming the microarchitectures, I'd just call them with the chip they started with.

P6 uarch = Pentium Pro - Nehalem/Westmere
Sandy Bridge uarch = Sandy Bridge - Kaby Lake (ongoing)

I.S.T. wrote:

They count both Nehalem gens as one due to them being identical. The Core/Core 2 thing is dumb, though.

And yet they make a distinction between Skylake (6th Gen) and Kaby Lake (7th Gen) despite them being identical too; That's marketing logic (or complete lack of it) for you!

jihadjoe wrote:

In terms of naming the microarchitectures, I'd just call them with the chip they started with.

P6 uarch = Pentium Pro - Nehalem/Westmere
Sandy Bridge uarch = Sandy Bridge - Kaby Lake (ongoing)

P6 architecture lineage actually ended with Bainas/Dothan a.k.a Pentium M. Conroe (Core 2) and Yohan (Core "1") had a number of elements that made them significantly different (longer pipe-lines, Netburst's Pre-fetcher and Quad-Pump FSB etc).

Chrispy_ wrote:

I.S.T. wrote:

They count both Nehalem gens as one due to them being identical. The Core/Core 2 thing is dumb, though.

And yet they make a distinction between Skylake (6th Gen) and Kaby Lake (7th Gen) despite them being identical too; That's marketing logic (or complete lack of it) for you!

The distinction was made several years ago, probably by other people. There is the underlying cause for the dumbness.

NoOne ButMe wrote:

As for Intel retiring the Core architecture (technically per Intel, Nehalem is technically the first Core microarchitecture part) I don't think it will happen. Not the way we think about it regularly.

Conroe was the official "Core" architecture from the engineering standpoint. When Nehalem was introduced it was turned into a marketing term. http://www.anandtech.com/show/1998

It's less about retiring the architecture and more about Intel finally getting around to implementing another big change. They can't integrate the IMCs again like they did with Nehalem, but surely there are some other random tricks they can begin introducing. A few percentage points gain in IPC per generation is one thing, but two generations in a row with zero IPC gain is another altogether.

Chrispy_ wrote:

I.S.T. wrote:

They count both Nehalem gens as one due to them being identical. The Core/Core 2 thing is dumb, though.

And yet they make a distinction between Skylake (6th Gen) and Kaby Lake (7th Gen) despite them being identical too; That's marketing logic (or complete lack of it) for you!

Kaby Lake has SGX (hardware-level decryption for protected content) and updated integrated graphics chip on it.

Krogoth wrote:

Kaby Lake has SGX (hardware-level decryption for protected content) and updated integrated graphics chip on it.

https://www.schneier.com/blog/archives/ ... ls_sg.html

And, SGX can be used to stealthily attack itself.

Captain Ned wrote:

Krogoth wrote:

Kaby Lake has SGX (hardware-level decryption for protected content) and updated integrated graphics chip on it.

https://www.schneier.com/blog/archives/ ... ls_sg.html

And, SGX can be used to stealthily attack itself.

Yep, it is one massive low-level backdoor waiting to be exploited.

I wasn't aware that Kaby's IGP was anything other than the addition of a fixed-function decode block.

In other words, a kludge to work around the lack of general shader power of the IGP, and was otherwise just Skylake's IGP, exactly - especially in terms of 3D capabilities and performance.

Anovoca wrote:

derFunkenstein wrote:

The only reason the Pentium brand got demoted to mean "budget" is because Intel ruined the name with Prescott and later Netburst chips. I would expect Core to stick around as a name as long as it's associated with good performance.

Pentium has made a comeback. Once all the people that jumped ship to the i3 for web browsing pcs realized that the Pentium processors were running circles around them in single threaded applications. So i guess if anything maybe expect intel to drop the i3 and load up a new chip with the i9 moniker with even MORE CORZZZZ.

/smug strut

Wow, Intel increased a marketing number by two. Way to go out on a limb there.

derFunkenstein wrote:

Wow, Intel increased a marketing number by two. Way to go out on a limb there.

Gotta justify that $1700 price tag somehow j/k

Agree with the other comments posted here. "Core i" is more of a brand or marketing term now, and encompasses multiple architectures. If we were speaking of "Core i" architecture then it's already been retired/superseded multiple times, but the "Core i" name will likely continue until Intel makes a brand-ruining bungle like it did with Pentium 4.