Overclock.net - An Overclocking Community - View Single Post - [TPU] AMD Readies Three HEDT Chipsets: TRX40, TRX80, and WRX80
View Single Post
post #41 of (permalink) Old 11-09-2019, 04:11 PM
uBronan
New to Overclock.net
 
uBronan's Avatar
 
Join Date: Aug 2010
Location: netherlands
Posts: 20
Rep: 0
Quote: Originally Posted by Solohuman View Post
Yes indeed, that is at the heart of whether or not gamers upgrade to this new platform. But I suppose hardware vendors like AMD have to put the gear out there first before game devs can take advantage of it.
There are many tasks which will benefit from 8 channel memory but most of them are scientific and are almost all products normal human beings hardly ever encounter.
All these programs are often custom made products which do insane complex calculations and i actually often had to buy these heavy workstations with thousands of dollars costing gpu co-processors some even up to 6 cards in them.
But the most of them which i ordered where made with 2 or 4 AMD or nvidia cards for these tasks, these enterprise computers are for normal tasks hardly any use.
One example of that is/was the long time called not so fast Itanium based workstations, but i disagree with all people calling these slow, they where super machines capable of much more than we gave them credit for.
The problem was that they can not be used by consumers, because the simulation of windows on them did not work well.
Its not the cpu which was bad it was the not so good running os (windows) which held back this superfast cpu, so i am kinda waiting to see how well the latest AMD monster cpu will do in the same tests which have been tested on the Itanium machines as well.

Nevertheless for most tasks normal users do 4 channel memory is in most cases already overkill , but yes 8 channel memory is only needed at the most demanding workstations on the planet.
In the work i did i have been running many of these super workstations ranging from 1 to 4 cpu and all multi core machine which almost always had ECC memory in them often with loads of memory per cpu.
Yes in many cases they where even more powerful than servers which served many hundreds of people working on them.
I actually had to replace a few times a server and used one of the workstations from our labs to run that tasks and often people told me that they did not know what i did but many tasks ran faster.

It was allways hard to explain to them that such a much smaller machine than the actual server was being faster, but also had to explain that this machine actually costed about 10 times more than the normal server.
So at the end they understood that keeping this machines working for their department would become way too expensive.
Its like using a formula one car to go shopping by the supermarket, and also you have a big chance that it will be gone when your return outside .
These are often the TOP in performance for those tasks with the price tag to match.
So totally useless for those which can not even get 8c/16t of these machines stressed ever, these are the ones we never see in the wild.
In most cases these machines run with very special software and will not easily be replaced so they keep running.
On my department i had 4 machines which was still running on os/2 and we all knew replacing them would be very costly so they will keep running as long as the replacement parts i bought will be available.
Even now after i am no longer working there the machine is still doing its work like no time has paste
So use most enterprise keep them running till they die, or contain very confidential information which never will go out of the companies/governments, in my case they where brought to be destroyed by a specialist company to strip and destroy them after i cleared any info from them.

Anyway that does not say these monster machines could not be used for other tasks however in most cases they simply are not going to perform the same as normal use products.
Hell many programs probably will crash if you try to run on them at all, and i have to warn people who think they can run them for the tasks they can run.

Most of the time they are not the most power efficient as well, actually turning them on could pop your breakers often if you try them at home .
Most of them actually have huge power supplies, i actually still got a few of them. One of them being 220 volt 3200 watt which would pop your breaker when its testing if all is ok if any other device is running on the same breaker.
I have used a few times to run a gpu cluster to work on a boinc project untill the power bill arrived .. that made me switch the whole system off

Last edited by uBronan; 11-09-2019 at 04:22 PM.
uBronan is offline