I was thinking. When I'm benching superpi 1M or Pifast on a heavily overclocked Pentium 3, it takes between 1 and 3 minutes to get the job done. This is quite some time to keep your OC stable.

 

But when benching on a new i5 670 LN2 cooled system, these times go down to as little as a few seconds. This way the record is more getting the allure of a suicide screenshot than a real benchmark Of course, there are still superpi 32M and wPrime1024M available, which helps a lot. These two are now down to ~5 and ~1 minutes, respectively. But eventually they will also go down quite a bit.

 

The discussion points that I want to make:

 

- Is a benchmark such as Pifast still relevant for modern CPU's?

 

- Is the challenge of getting the CPU stable still real for such short bench times? Or is it getting more and more a good shot of good luck and hoping that you get a 'stable' moment while benching?

 

- Are the current set of CPU specific benches future proof?

 

- Are there alternative benches that can induce a more intense load on the modern CPU architectures?

 

Oh yeah, and remember. In 50 years, when graphite is the main resource material for CPU's (up to 1 THz), the silicium age will be known as the stone age of electronics

 

See my attachment for a quick bench envelope over the last years (based on pentium release dates and topscores)

 

EDIT: For some reason I'm missing part of the beginning of my post when looking at it on the forums, but while editing and previewing it looks fine? IE issue?

EDIT2: My post looks fine in Firefox, but I'm still missing the first part in IE. Renderfault?

EDIT3: Correction: wPrime1024M is down to 1m20s with the new 980

Edited March 11, 201016 yr by TRG

Superpi hasn't been "relevant" for 15 years, it doesn't use much of the CPU.

 

I feel we need some longer benchmarks soon, but not yet. Maybe we need to update superpi someday to display .1ms as well:D

WFT! Never!

 

It's getting only meaningless under 1 s...

WFT! Never!

 

It's getting only meaningless under 1 s...

 

comments like that is the reason why i love to bench!!!

Edited March 5, 201016 yr by packet

Superpi hasn't been "relevant" for 15 years, it doesn't use much of the CPU.

 

I feel we need some longer benchmarks soon, but not yet. Maybe we need to update superpi someday to display .1ms as well:D

 

then you just have to run SPI 2M instead of 1M

 

and so on ...

Wow thats seems really interesting. Thnx for sharing ...

I was planning on commenting this a week ago, but have been quite busy. Will reply in the future, this is an interesting topic.

Well I don't know, seems SuperPi1M is very relevant for my setups...;P

 

http://hwbot.org/community/submission/966448_1badmojoe_superpi_k6_2_380mhz_11min_29sec_732ms

 

 

or better yet here is superPi32M

 

http://hwbot.org/community/submission/966450_1badmojoe_superpi_32m_k6_2_380mhz_16h_41min_37sec_484ms

 

 

I am fond of the current de facto standards, I don't need a bench mark that takes a few days to complete. Looking at the not so linear curve in the graph, It does not appear anyone will be breaking under 1sec superpi 1M this decade.

Wprimes' another story if and when better storage management technologies become available.

 

[EDIT ADDITION]

One could point out my not so modern cpu as in the above example and the same can be held true to the not so modern '06-'07 Intel Celeron LGA775 347 and the famous Intel Core 2 E8600 family that lead to the Intel Core i5 670. Which appear to stem from the same pentium family tree.

Is Moores' law of computer scalability coming into play looking into the graph presented? Has alot of refinement and efficiency been incorporated into a possible dead-end Pentiums scalable architecture? How much further can the Pentium be refined? New and technical exciting materials must being developed to see any real answers. I remember when talk of using synthetic Sapphire as a replacement for silicon, but perhaps the Moll hardness of Sapphire is less durable than silica sand, if so, then why not synthetic diamond? Well how about organically engineered tissues, perhaps from an rat brain?

Defiantly something has to change and not when but if, does another overwhelming problem resurfaces its ugly head...storage interface capabilities. Floppy and IDE is dead, SATA is becoming saturated and the only foreseeable time tested integration of an PCI-x storage interface solution for the near term. An prov-en technology yet to hit the main stream, perhaps some of you use this already to have Wprimes' 1sec times within reach.

Edited April 9, 201015 yr by 1BadMoJoe