Rene Lamontagne wrote:
On 2019-08-25 1:58 p.m., Paul wrote:
Rene Lamontagne wrote:
As I said in previous post I gave my new build z390, i7 8700 build to
my son in exchange for his old i7 950 Gigabyte x58 system.
Well I played around with the old box for a week or so then found it
was not to my liking, That board has been rather troublesome for the
last 9 years, So I decided that's enough, Out you go.
Not wanting to spend another $1000+ dollars I did a fair amount of
research and decided to do something I have never done before.

Try an AMD CPU, so I ordered An AMD Ryzen 5 3400G CPU With onboard
Vega 11 graphics and Wraith Spire cooler, an Asus B450 F motherboard
and 16 GB of Corsair 1600 MHZ memory All for abut $450.
Put it all together in the same Coolermaster CM 690 II case and power
supply.

Everything fired up nicely so I installed a fresh copy of Windows-10
ver 1903 and all was good, I had to give Microsoft my 8.1 key to
activate.
I just did a benchmark using Half-Life lost coast test demo and am
getting 256 FPS, So all looks great.
Surely this shoulb be my last build (I hope not) :-)

Rene

I think I just saw an Intel salesman, selling all his INTC stock :-)

Paul


Well I have been using Intel CPUs for about 24 years and was and still
am quite happy with them, This was more of a bang for your buck thing
and my less than former demands, so figured I would give AMD a whirl.
BTW in my previous quote I stated 1600 MHz memory, That was a senior
moment error, it is 3200 MHz memory. :-)

Rene

With the double data rate, a 3200 memory might be receiving a
1600MHz clock. Data is transferred on both edges. You're not that
far off.

There is a faster kind of memory, called QDR, which has
been around for years. There are four transfers per clock
cycle. They're very hard to work with, which is why we
should be glad the computer industry doesn't use them.
One of the guys at work was designing with them, and
the "specs" are full of imaginative budget contributions.
20pS for this, 50pS for that, and somehow the memory is
expected to run without throwing errors. That kind of
memory would work best, soldered right to the motherboard
(which of course, customers don't want).

Eventually, they'll run out of internal dimensions
to make the memory faster. (They use parallelism inside.)

Memory innovation is all about "gradualism". The companies
make their real money, every time a new standard comes out.
Towards the end of a cycle, profits can be thin. Especially
when a new player is available to apply production pressure
(the Chinese make DRAM now, for consumption mainly in China).
Could they have gone from DDR2 to DDR5 and skipped the
others ? We'll never know. The thing is, the memory chips
have some features, that CPUs don't use, even when they
might. It's rather like when Intel adds additional
instructions, and nobody adds the instruction types
to the compiler. Nobody has any time to "absorb the
advantages" of features from other parts of the ecosystem.

Paul