Building a PC, Part VIII: Iterating

mmmmm - 950 Pro

Sequential read speeds top out at 2500MB/s, while sequential writes hit 1500MB/s.

1 Like

@codinghorror, Did you measure power usage? How many watts it approximately your new “iterated” PC requires while idle and while working/playing hard?

After reading your article, I was considering it to -erm- “iterate” my aging Linux server when I found that Asus didn’t add a VGA/HDMI port…

As others pointed already, M.2 slot was available on older chipsets.
The main problems though are:

  1. PCI lane sharing, which is not a real issue as long as you don’t use any PCIe network/sound/whatever cards.
  2. M.2 SSDs run HOT HOT HOT. There are no heat spreaders on these tiny sticks.
1 Like

That’s both true and false; yes, phone’s raw power doubles at a fast rate, but they don’t allow efficient heat dispersal, and are invariably throttled, a fact that phone makers don’t advertize.
For the anecdotal evidence corroborating this fact, do you see any increase in speed when switching apps, going back to the home screen, loading a big game between your Blahphone 3 and your Blahphone 4, or 5, or 6? I certainly don’t.

Absolutely I do. Considering each generation (so far) has been twice as fast in real world terms.

What makes you think that handhelds and smart devices will not hit the same bottleneck soon? Or does Moore’s law apply only to PC processors and they skip mobile processors due to a freak error in Quantum Physics? Just a matter of time, before the SnapDragons and Kirins and Exynoses begin to suffer from the exact same problem (In fact, they already do - the pursuit is on for more cores, less GHz per core, heating problems, BIG.Little and so on).

The iPhone 6S is 10 times more powerful because it started with such a low standard to begin with. A similar argument could be had for Intel CPUs with a 6 generation gap - Compare the current gen SkyLake’s with a Core 2 Quad (6600) etc. - lightyears ahead, without exaggerating, and evidently so when it comes to compute intensive tasks like media encoding or encryption.

Similarly, to debunk the argument that PC processor speeds aren’t doubling every gen - Well, how much is the improvement when going from, say, SnapDragon 801, to 805, then from 805 to the 808, onto the 810. Do you see a doubling with every gen there? Thought not. More like a minuscule improvement, as the synthetics (GeekBench, AnTuTu) depict.

One example for the PC platform, NOT 6 gens apart:

Almost a 170% increase.

As for Skylake, that specifically meets Jeff’s requirements that, to upgrade, there must be a significant performance improvement:

The current latest-and-greatest Intel CPU is Skylake. Like Sandy Bridge in 2011, which brought us much faster 6 Gbps SSD-friendly drive connectors (although only two of them), the Skylake platform brings us another key storage improvement – the ability to connect hard drives directly to the PCI Express lanes.

As for i5 vs. i7, I mostly agree, although I have seen a few gamers (yes, n=1) who claim a 30fps increase. I don’t know enough about games, but there are tons of physics calculations so I would imagine that takes advantage of the extra power of the i7. Some games specifically state that the i7 is their “preferred” CPU, so I would say, it really depends on what you are running.

A brief test of the SM951 in AHCI vs NVMe:

If you already purchased a Samsung SM951 SSD with AHCI, then we can’t recommend the NVMe model for its slight performance boost. It’s faster, but only slightly. Moving from an 850 Pro to the SM951-NVMe is a nice upgrade though, and you will notice snappier performance from the reduced latency.

For new builds you should definitely prefer NVMe but for older builds you are not losing a bunch by having the AHCI version of the SM951.

Tracking performance over the last ~6 years in JavaScript:

Up to today:

image

image

So let’s see:

Kraken ↓ Octane ↑
9900k 725 57k
6700k 740 45k
2600k 1071 31k

So… 1.5x improvement from Sandy Bridge (2011) to Skylake (2015) and 1.25x improvement from Skylake (2015) to latest-iteration Coffee Lake (2019). Total improvement of 1.83x from 2011 to 2019.

Comparing Speedometer 2.0 results in Chrome latest (81) on i7-2600k …

with i9-9900ks

That’s a 2.12x improvement.