PCPer Mailbag #1 - 7/20/2017

Subject: General Tech | July 21, 2017 - 11:12 AM |
Tagged: video, pcper mailbag

Our readers and supporters have asked us to do a mailbag-style video series for a while now and I finally got around to starting it. Welcome to the PCPer Mailbag #1! Thanks to the supporters of the PC Perspective Patreon for making it happen!

Questions addressed this week include:

  • Thoughts on when spinning disks will be replaced by SSDs in lower cost systems
  • Mesh vs Infinity Fabric
  • Will GPU prices return to normal?
  • Short history of PCPer
  • ...and more!


July 21, 2017 | 01:05 PM - Posted by Rob (not verified)

Hey Ryan

I love the site's "Hardware Leader Board" !

Do you think a coin mining system could be added ?
(Hi-End, Power Efficiency, Sweet Spot)

Rob

July 24, 2017 | 09:46 AM - Posted by collie

Coin mining is almost dead in the water..... again.

However the built rigs are always just for the first rush, after that the massive multiple video card rigs are replaced by simple plug and play ASICs, far more efficient in terms of power heat and cost.

July 21, 2017 | 02:23 PM - Posted by Eric (not verified)

Great informal talk. The issue of SSDs vs HDDs is interesting. There is a modern tendency to view electric/electronic devices as superior to mechanical devices. This happens in computers (SSD vs HDD), automobiles (electric vs gas engine), music (electronic pianos/violins/guitars vs traditional) and other areas. The problem for fans of electric/electronic solutions is that mechanical devices are constantly being improved through computer aided design and hybridization with electronics. For example, Volvo recently announced that they will not produce gas-only cars by 2019. But they have been experimenting with kinetic (flywheel) batteries, which could be an alternative to the electric hybrids. Note that there is no electric component.

"In 2013, Volvo announced a flywheel system fitted to the rear axle of its S60 sedan. Braking action spins the flywheel at up to 60,000 rpm and stops the front-mounted engine. Flywheel energy is applied via a special transmission to partially or completely power the vehicle. The 20-centimetre (7.9 in), 6-kilogram (13 lb) carbon fiber flywheel spins in a vacuum to eliminate friction. When partnered with a four-cylinder engine, it offers up to a 25 percent reduction in fuel consumption versus a comparably performing turbo six-cylinder, providing an 80 hp boost and allowing it to reach 100 kilometres per hour (62 mph) in 5.5 seconds."
https://en.wikipedia.org/wiki/Flywheel_energy_storage

So it's theoretically possible to create a purely mechanical hybrid car engine.

Spinning discs will be with us for a very long time.

July 21, 2017 | 04:25 PM - Posted by WayneJetSki (not verified)

This was cool. Please do more of it.

I also love the hardware leader board. Wish there was an update more recently since I have been thinking about doing my first build in the next month. Hope to see some more AM4 motherboards getting reviewed since I will be in the market for one very soon.

July 21, 2017 | 09:12 PM - Posted by Clmentoz (not verified)

3D XPoint and spinning rust. But Optane is a brand of XPoint, same as QuantX(Micron's brand name). So hopefully by the end of 2017 there will be some competition in that market as well to drive the pricing down to more affordable levels relative to NAND. I still trust spinnng rust more for longer term storage than any NAND and Hopefully some XPoint Cache can do wonders for some better all around hybrid XPoint/Spinning rust designs that are better overall than purely NAND based SSDs.

I'd like to see some drives starting with maybe 128GB of XPoint/Cache and at least 1TB of hard drive storage over an NVM/PCIe high bandwidth interface, with options for larger amounts of XPoint if needed. 128GB of XPoint would be enough for OS/Paging swap and also data caching the hard drives most recently/often used/files and user program code.

It would be nice if the OS could treat the memory paging swap space as the prefered method of caching programs/data by utilizing the virtual address space functionality already present in all modern OSs/CPU hardware. So the OS can manage the entire XPoint/Cache as if it were an extention of DRAM with the OS actually stored on an allocated RAM-DISK partition that just so happens to be in a memory area page swap allocated to the XPoint cache and the DRAM. This method with the OS also made to make sure all the actual DRAM space in also mirrored in the XPoint/page swap for fault tolerance and rapid startup via an automatic retention of whatever data was in the DRAM mirrored onto the XPoint/Cache.

The user could also have the option of also having the XPoint contents mirriored out to the hard drive's platters for any OS restore-point functionality/other OS snapshot services and such where needed/required by the OS.

Most home computers currently only use 8 to 16 GB of DRAM, with only a few home systems using 32 and above GB of DRAM, so 128 GB of XPoint should be enough for most Laptop/PCs along with whatever Hard-drive, or NAND storage the user wants for their system.

MY only concern with first generation XPoint is its durability at the present is not that many times greater than SLC NAND so XPoint based DIMMS are going need plenty over provisioning to last as long as DRAM.

The main problem with Intel's Optane currently is any support is mostly only provided in the latest Intel hardware so that excludes too many older systems. Micron will be taking a different appraoch/business model with its QuantX/XPoint IP actually being up for licensing by other Drive makers. So third party Hardrive/SSD/DIMM makers can license from Micron and incorporate 3D XPoint into their products.

I'd love to see combo-XPoint/DIMMs for older laptops where there is no PCIe/NVM capability that could be made to have some form of XPoint/DIMM combo DIMMS where the memory interface could be utilized. The combo XPoint/DIMM could be engineered so the combo-DRAM/XPoint DIMM's controller would manage DRAM to XPoint(ON DIMM XPoint) transfers in the background on a DRAM DIMM/background channel so all the OS/file system would have to do is write to the DRAM/Buffer and have the data transfered in the background to the on DIMM XPoint storage pool. This all would be managed by any drivers/firmware/controller on the Combo XPoint/DIMM with the OS made aware of the interface via some form of open standars(JEDEC/Other) based interface.

July 22, 2017 | 04:39 AM - Posted by JohnGR

Millions and Millions of dollars. We finally have proofs on video :p
(just kidding)

Great video. Bring more.

A, probably, stupid suggestion/wish.

Core was the successor to Pentium III in fact, not Pentium 4. Pentium 4's architecture was so bad, that AMD thought it could do better with Bulldozer. And obviously Core was decimating Pentium 4 on IPC.

Could you please find a nice little quad core Phenom II somewhere and do a couple of IPC comparisons with the Ryzen 3 1200 in your Ryzen 3 review? Same frequency, no turbos, a couple of benchmarks, maybe older, that don't sky rocket thanks to the implementation of some modern set of instructions. When Ryzen came out I was thinking that it looked more like the Phenom III we never got than a Bulldozer successor.

July 22, 2017 | 11:28 AM - Posted by Clmentoz (not verified)

With Zen's SMT on and with Zen's SMT off also for some more metrics on how much the SMT in Zen helps total IPC over the the AMD 10h/"K10" CPU micro-arch.

"Zosma[edit]

Four AMD K10 cores chip harvested from Thuban with two cores disabled[20]
45 nm SOI with Immersion Lithography
L1 cache: 64 kB + 64 kB (data + instructions) per core
L2 cache: 512 kB per core, full-speed
L3 cache: 6 MB shared among cores.
Memory controller: dual channel DDR2-1066 MHz (AM2+), dual channel DDR3-1333 (AM3) with unganging option
MMX, Extended 3DNow!, SSE, SSE2, SSE3, SSE4a, AMD64, Cool'n'Quiet, NX bit, AMD-V
Turbo Core
Socket AM2+, Socket AM3, HyperTransport with 2 GHz
Power consumption (TDP): 95 Watt and 125 Watt
First release Unknown, only released to certain OEMs.

Clock rate: 2.7 GHz to 3.5 GHz
Models: Phenom II X4 650T, Phenom II X4 840T, Phenom II X4 960T BE and Phenom II X4 970 BE (E0 stepping)

Deneb[edit]

Four AMD K10 cores
45 nm SOI with Immersion Lithography
L1 cache: 64 kB + 64 kB (data + instructions) per core
L2 cache: 512 kB per core, full-speed
L3 cache: 6 MB shared among cores.
Memory controller: dual channel DDR2-1066 MHz (AM2+), dual channel DDR3-1333 (AM3) with unganging option
MMX, Extended 3DNow!, SSE, SSE2, SSE3, SSE4a, AMD64, Cool'n'Quiet, NX bit, AMD-V
Socket AM2+, Socket AM3, HyperTransport with 1.8 to 2 GHz
Die Size: 258 mm²
Power consumption (TDP): 65, 95, 125 and 140 Watt
First release 8 January 2009 (C2 Stepping)

Clock rate: 2.5 to 3.7 GHz
Models: Phenom II X4 805 to 980 BE except Phenom II X4 840T, Phenom II X4 960T BE and Phenom II X4 970 BE (E0 stepping)"(1)

(1)

"AMD 10h"

https://en.wikipedia.org/wiki/AMD_10h

July 22, 2017 | 02:03 PM - Posted by JohnGR

You can add Athlon II 640T and Athlon II 645T(rare) on those E0 models list. People who still have AM3 motherboards that unlock cores, can try any of those processors and see if they end up with a 5 or 6 core Phenom II. I had two 640T and one 645T. Two could work as 6 cores, one as 5 core.

July 23, 2017 | 05:54 AM - Posted by MEW

Is there somewhere on pcper.com where you show league tables for different hardware?

For example which SSD's Alan rates as the Top 10 best etcetera?

July 27, 2017 | 08:45 AM - Posted by jmc (not verified)

AMD Vega, I am seeing no mention of "Freesync2".
No Freesync2, no buy.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.