Intel Announces 8th Gen Core CPUs: Claims 40% Boost Over Gen 7, More Cores

[Holograph]

40% boost, wow. that should make people happier now

I don't recommend taking that at face value. Their claims of performance increases generally would get a "Pants on Fire" rating if we used the scale from a certain political site.

[bug77]

40% boost, wow. that should make people happier now

Form Arstechnica:

A total of 25 percent of that boost (in the SYSmark benchmark) comes from the doubled core and thread count. The remainder is split evenly between "manufacturing" improvements (which is to say, higher clock speeds) and "design" improvements.

[caligula]

Form Arstechnica:

It's also a synthetic benchmark, not a real world situation.

[smartalgorithm]

Form Arstechnica:

yep, that could be... but that still fine if they adopt a good pricing policy. i don't care how the performance of my computations is boosted, i care if it's boosted or not.
anyway i hope to see more news like these.

[bug77]

It's also a synthetic benchmark, not a real world situation.

I meant: it's not a 40% increase in IPC, it's a 40% increase if you look at one specific thing from one specific point of view. Like all performance increases we've got since Sandy Bridge or even a few iterations before that.

[sdack]

It's also a synthetic benchmark, not a real world situation.

All benchmarks are synthetic. What you probably mean is that it is a benchmark, which tries to give you a singular result, but then its result ends up being misinterpreted. Even PTS, being composed out of many "real world" applications, is still only synthetic and its results have to be interpreted carefully, too. The SYSmark benchmark is useful when one only wants to get a good, first performance number on a CPU and to compare it against previous models, which is also why SYSmark still gets used for this reason.

[torsionbar28]

'cause hating Intel is so trendy.

I guess someone has to buy their over-priced under-performing chips. For the rest of us, we'll go with 'Competitor A'.

[torsionbar28]

All benchmarks are synthetic. What you probably mean is that it is a benchmark, which tries to give you a singular result, but then its result ends up being misinterpreted. Even PTS, being composed out of many "real world" applications, is still only synthetic and its results have to be interpreted carefully, too. The SYSmark benchmark is useful when one only wants to get a good, first performance number on a CPU and to compare it against previous models, which is also why SYSmark still gets used for this reason.

Not sure I follow your logic here.

Synthetic benchmarks are called "synthetic" because they use specialized software designed to stress and measure one specific aspect of a CPU. Integer math, or floating point math, for example, for the Whetstone and Dhrystone benchmarks.

"real world" benchmarks are performed not using specialized software, but rather using the same real applications that an end user runs, performing a similar workload to what an end user would perform. Photoshop, Cinebench, etc. Your example of SYSmark is user-application based, and therefore it falls in the real world category. SYSmark is an aggregate of a multiple real-world applications, so it's an aggregate score, not a singular application benchmark.

To make a sports analogy, a synthetic benchmark would measure how high a player can jump. Another synthetic benchmark measures how fast he can run 100 meters. Another synthetic benchmark measures how many push-ups he can do in two minutes. A real-world benchmark on the other hand, measures his batting average, his free-throw average, or his average number of shots on goal.

Higher synthetic benchmark numbers may or may not translate into higher real world benchmark numbers. How many more frames per second in Crysis 3 will you get from a 5% increase in a CPU's integer math performance? Nobody has a clue. That's why if we're interested in Crysis 3 performance specifically, we run a real-world Crysis 3 benchmark.

So no, benchmarks are most certainly not all synthetic.

[andyprough]

If anyone is looking for the Linux version of the Intel AMT Detection and Mitigation tools: https://downloadcenter.intel.com/dow...?product=23549

Says its only for Ubuntu 16.04, but its just a zip file with a couple of simple tools that worked on my Tumbleweed setup.

[nomadewolf]

WCCFTech welcomes you.

And no, Intel ME doesn't go anywhere. Enterprise wants it, enterprise needs it.

Also, Intel ME can be permanently disabled but bitching about Intel is so cool nowadays, right? As if Zen has finally reached IPC parity with Intel. Oh, wait, it hasn't.

Also, most consumer PCs have it disabled by default (IOW it is present but not active). But who cares? Let's mention it anyways, 'cause hating Intel is so trendy.

Once upon a time, AMD beat Intel in IPC, Multitasking, performance per dollar, performance per watt and price, for years.
Somehow, AMD sales went down instead of up or even maintain...
Later, it came to light that Intel payed, coerced and threatned OEMs to not use AMD processors in their products.
When you have to face 'competition' like Intel is a 'bit' hard to, well, compete...

That said, with Ryzen AMD still managed to bring 53% IPC over it's previous architecture and more than 4 cores to mainstream. Something that Intel didn't manage to do in 10 years.

So hating Intel is not only extremely cool, but also the right thing to do!