RAM speed and engine strength

[syzygy]

Rambus is a patent buying company, then sues every other company that remotely uses some method/feature mentioned in the patents.
The company has engineers but they never came up with a good product.
The initial P4 RDRAM fiasco had to be solved/finished by Intel engineers. Later they abandoned RDRAM and moved to DDR.
Rambus is more like e patent fighting company than any other thing.

In the US in the early 1980s there was a video game company named Oddessy. I read they made most of their money by suing other companies claiming they had a copyright on video games in general. Strange.

Magnavox owned patents for video games played on a television screen.
https://en.wikipedia.org/wiki/Magnavox_Odyssey#Lawsuits

The root of the conflict was a pair of patents by Baer -- one which described how the Odyssey showed player-controlled objects, or dots, on a video monitor and described a number of games that could be played with the system, and an earlier one that went into detail about how the Odyssey used that system to have two dots collide with each other and have one bounce off, specifically using a game of ping pong as an example.

The judge ruled that Baer's more general patent for the Odyssey constituted "the pioneering patent of the video game art", and held the defendants' games as infringing the patents. Atari settled early on in the court case with Magnavox, and in return were granted a license in exchange for US$1.5 million and access granted to Magnavox to all technology produced by Atari from June 1976 to June 1977, while the other defendants paid higher penalties.

Magnavox won more than US$100 million in the various patent lawsuits and settlements involving the Odyssey related patents. In addition, they had a large number of licensees of patents, with over one hundred already by the mid 1970s.

And it seems Atari would never have been founded and Pong never been created without the Odyssey having been there to inspire Atari's founder.

The father of video games

[Cardoso]

I forgot to add that Rambus also charged very high license fees for RDRAM which added greatly to the cost of memory.

Have fun with some of these RAMBUS sues:

https://www.reuters.com/article/us-ramb ... XL20111117
https://www.reuters.com/article/ibm-ram ... 2520100824
https://www.pcworld.com/article/212218/article.html
https://gizmodo.com/5024133/rambus-come ... fringement
https://www.cnet.com/news/rambus-sues-n ... ringement/
https://www.eetimes.com/document.asp?doc_id=1126659
https://www.extremetech.com/extreme/826 ... fringement
http://www.theregister.co.uk/2006/01/16 ... es_micron/
https://www.theregister.co.uk/2004/05/0 ... _infineon/
https://www.extremetech.com/extreme/753 ... am-patents
https://www.zdnet.com/article/rambus-su ... ringement/
https://www.computerworld.com.au/articl ... r_patents/
https://www.edn.com/electronics-news/42 ... hip-makers

[Leo]

I find the information interesting. However, I just want to see RAM 5 get made. I basically don't care who does it.

[Cardoso]

I find the information interesting. However, I just want to see RAM 5 get made. I basically don't care who does it.

Come to think of it (RAM prices skyrocketing) I no longer know what's best: Rambus of the rest of the memory makers.
In fact RAM prices is the only thing that is keeping me from upgrading.
I have a core i7 i920 @2.6GHz and 24GB RAM and it's more than time to upgrade.
I would like a Ryzen 2700x and 64GB, to develop the 10man spanish checkers EGTBs. But with these RAM prices it won't happen soon.

[Ozymandias]

RAM prices is the only thing that is keeping me from upgrading.

That's why I ask. Normally I would just buy the "sweet spot" speed and wouldn't bother testing; with DDR3 this was 1600MHz. But DDR4 is behaving differently. I bought 16GB 3 years ago at a high price, because it was relatively new, but after a not so long period of moderate prices, when I came back for more it was already getting more expensive. I've been waiting for more than a year for things to get back to normal, but it doesn't happen, so it becomes more important to know exactly what it is you need, both in terms of size as well as speed.

I'm still waiting on someone to answer the question behind this thread.

[Cardoso]

Well to answer your question we would need several RAM kits to test with.
In chess latency is important.

With Intel systems you won't see a significant difference in KNs between 2133 and 2666.
But with Ryzen I think things are a bit different, the memory controller on Ryzen is very sensitive to raw memory speeds.
Also since you are an early adopter you have to deal with the teething problems of a very young platform. You also got a B350 motherboard wich is less flexible.

But your question was about engine strength differences between different memory speeds.
While you might see a difference in KNs for the same Ryzen cpu but for different RAM kits, I really don't think you wil see an ELO difference.
Your cpu is good and if you could run your memory at 2666 it would be slightly better in KNs but no so much in ELO, not even 1ELO point.
So my advice is forget about the whole thing!
If they release new BIOS updates and you could run your RAM at 2666 please do it, if not be assured you won't be missing anything significant in terms of ELO.

[Rein Halbersma]

I find the information interesting. However, I just want to see RAM 5 get made. I basically don't care who does it.

Come to think of it (RAM prices skyrocketing) I no longer know what's best: Rambus of the rest of the memory makers.
In fact RAM prices is the only thing that is keeping me from upgrading.
I have a core i7 i920 @2.6GHz and 24GB RAM and it's more than time to upgrade.
I would like a Ryzen 2700x and 64GB, to develop the 10man spanish checkers EGTBs. But with these RAM prices it won't happen soon.

With some slicing techniques, 24gb should be plenty for 10pc dbs!
Ed Gilbert built them for American checkers almost 13 yrs ago with only 2gb
http://edgilbert.org/EnglishCheckers/10-pieceBuild.htm

[Cardoso]

I find the information interesting. However, I just want to see RAM 5 get made. I basically don't care who does it.

Come to think of it (RAM prices skyrocketing) I no longer know what's best: Rambus of the rest of the memory makers.
In fact RAM prices is the only thing that is keeping me from upgrading.
I have a core i7 i920 @2.6GHz and 24GB RAM and it's more than time to upgrade.
I would like a Ryzen 2700x and 64GB, to develop the 10man spanish checkers EGTBs. But with these RAM prices it won't happen soon.

With some slicing techniques, 24gb should be plenty for 10pc dbs!
Ed Gilbert built them for American checkers almost 13 yrs ago with only 2gb
http://edgilbert.org/EnglishCheckers/10-pieceBuild.htm

Right, I could slice even further than what I currently do:
I slice not by the leading rank, but by leading square = 784 slices.
But I'm too lazy
But I could use only one machine and to speed up the process I really would like to split the computation in :
5vs5, 6vs4, 7vs3, 8vs2 and 9vs1 at the same time.
What I do is to copy and rename my generator to 5x5.exe, 6x4.exe, 7x3.exe, 8x2.exe and 9x1.exe.
Then run these 5 programs simultaneously and this takes up RAM, both in the capture pass and in the non capture pass. My cacheing scheme is pretty basic: I keep other slices in RAM up until a certain number or RAM usage whatever comes first. When one of those limits is reached I unload those slices and continue. It's simple but it worked great for the 9man database.
Yes with finer slicing I could do it with much less RAM, but the 64GB thing is tempting me to load use the entire 10man into RAM.
Currently my engine supports 2 modes: load the entire DB into RAM or use Memory Mapped files (like Ronald's Syzygy bases).