SoylentNews is people
SoylentNews
Over at www.semiaccurate.com they are discussing - in unfavourable terms - the launch by Intel of the 'Comet Lake' series of CPUs:
Intel LogoLast week Intel launched their '10th Gen' CPUs, a self-inflicted wound that shouldn't have been released. If you think SemiAccurate is being a bit harsh here, ask yourself why Intel birthed this debacle in the first place.
[...]
But performance is why we are mocking Intel over this ‘family’. The company split out the Ice and Comet briefings so as to minimize comparisons and uncomfortable questions between the two. As we pointed out in our earlier article on Ice there were vague comparisons between the two ranges but no actual data. Intel even unethically hid the SKUs they were testing.
This time was more of the same, absolutely zero charts comparing this ‘generation’ to the last, or even to itself. All we got was a slide saying, “up to 16% better overall performance vs. previous gen”. That may sound great but, err, 6 cores vs 4 should get one a 50% higher performance or at least something closer to 50% than 0%, right? Don’t forget the faster memory on the new Comet devices which should get you a big chunk of that 16% alone. That level of sleaze is expected but we didn’t expect Intel to be outright unethical. Again. Actually we kinda did but we honestly hoped they wouldn’t do it again.
Yup that 16% max increase when going from a 6C i7-10710U to a 4C i7-8565U was bad. The fact that they compared a 25W 10710U to a 15W 8565U and buried that fact in the fine print is unacceptable. We once again call for the idiots responsible to be fired, not that Intel will do anything, it seems this sort of behavior has become acceptable at big blue. It still isn’t right.
Read the story and see if you agree.
(Score: 0) by Anonymous Coward on Tuesday August 27 2019, @07:22PM (8 children)
Or use AMD which gives you more performance for less price... and once devs get used to developing for manycore processors this advantage is only going to grow. I wouldn't be surprised if many legacy products like Photoshop need to get completely rewritten or even supplanted by new software.
(Score: 1, Funny) by Anonymous Coward on Tuesday August 27 2019, @08:19PM (5 children)
2005 called, they want their excuse back. Seriously, *everyone* knows about SMP these days and if it's not used, it's only because it's not needed.
(Score: 0) by Anonymous Coward on Tuesday August 27 2019, @08:31PM (4 children)
Have you ever refactored old code to be multi-threaded? I mean I still write single threaded code by default even if I know I will want it parallelized later just because it is quicker and easier to debug... Then when a deadline hits that gets left in. Your comment seems pretty ignorant to me.
(Score: 1, Interesting) by Anonymous Coward on Tuesday August 27 2019, @10:26PM (3 children)
There is some stuff that went mutil thread. The rest not so much.
The 'real' speed up for most computers is up to 30% per processor and that falls off pretty quick as you add more and more. For 'stupidly paralizable data problems' the more the merrier. But https://en.wikipedia.org/wiki/Amdahl's_law [wikipedia.org] is hard to work around unless you change the data to match what you are doing. It is along the idea if I have one cent and I add one cent I am 100% richer. But if I have a dollar and add 1 cent I am less than 1% richer. Which is one part of that law. The other one is how the network actually gets in the way.
(Score: 0) by Anonymous Coward on Tuesday August 27 2019, @11:14PM
I don't need any theoretical analogy to understand the advantage of multi-threading... when it is appropriate.
(Score: 0) by Anonymous Coward on Wednesday August 28 2019, @01:42AM (1 child)
Here is one test: https://image.ibb.co/ecH9VL/coretest.png [image.ibb.co]
From 10 minutes to 30 seconds is much more than 30% improvement...
(Score: 1) by Zappy on Wednesday August 28 2019, @08:48AM
Also notice going from 1 to 20 cores gives you almost all the speed-up you're going to get, at least going from 20 to 60 cores didn't make it any slower.
(Score: 2) by JoeMerchant on Tuesday August 27 2019, @09:21PM (1 child)
Still waiting for the ASROCK IBOX-R1000M AMD RYZEN EMBEDDED FANLESS MINI PC announced 3 months ago... AMD can't take the market if they've got nothing to ship to customers who are ready to buy.
🌻🌻 [google.com]
(Score: 1, Informative) by Anonymous Coward on Tuesday August 27 2019, @09:36PM
I looked it up and this let me get to the checkout screen: https://mitxpc.com/products/ibox-r1000m [mitxpc.com]
So not sure what your issue is.