Is the 3900x overkill?

[neatherback]

I'm currently doing some PRP tests and its running fine, but all the WR holders seem to be using Pentiums and Core 2 Duos, is there a reason for this?

[Xyzzy]

Probably memory bandwidth is an issue. The 3900X has dual-channel memory. All of the fastest systems have quad-channel memory.

[VBCurtis]

What world records are you looking at that are still held by Core2?

Speed records are all held by graphics cards these days.

[axn]

Perhaps he's talking about https://www.mersenne.org/primes/ ?

When it comes to distributed computing, there's no such thing as "overkill". The more you crunch, the greater the chance of discovery.

[SteinFG]

Well, the core 2 duo produced a last WR 7 years ago, there's that. Also, as I remember, some of those people were using a LOT of core2duo computers. So, I'd say that the real overkill was what they did.

[ixfd64]

Originally Posted by axn

When it comes to distributed computing, there's no such thing as "overkill". The more you crunch, the greater the chance of discovery.

You took the words right out of my mouth.

[kriesel]

Originally Posted by neatherback

all the WR holders seem to be using Pentiums and Core 2 Duos, is there a reason for this?

Time and money. Once, long ago, they were the state of the art. The last core 2 duo that found a world record Mersenne prime was in 2013, and the cpu model was introduced in 2008. Universities often run computer gear till it breaks, repair it if it can be done cheaply in-house by someone on staff, and go again. I used to supervise some IT guys, and one was very good at stretching the hardware budget that way. We would sometimes buy pallet-quantity inexpensive used equipment off warranty and scrap out the first ones to break, to maintain the others. It doesn't take the latest and greatest to interface to some lab equipment. It can take large numbers of machines though, for labs, and instructional settings.
https://ark.intel.com/content/www/us...3-mhz-fsb.html
So, "all" is not accurate currently.
The problem of finding the next n world record mersenne primes is very stiff. Primality tests scale as ~p^2.1. It's been estimated that finding successive mersenne primes at the probable spacing increases in effort required 8-fold each. A very rough estimate of how long it will take to finish out the p<10⁹ range is 150 years with evolutionary advances in computing speed. https://www.mersenneforum.org/showpo...5&postcount=11
The hot PC of 30 years ago was a 386/33 with cache. That, and a core 2 duo, is not worth the electricity to run it currently for prime hunting, unless there is space heating benefit.

[Till]

https://nationalinterest.org/blog/bu...changer-127952


Yesterday somebody wrote me that he factored RSA-100 on that machine in 17min, using my PSIQS Java implementation with 30 threads.

[xilman]

Originally Posted by axn

When it comes to distributed computing, there's no such thing as "overkill". The more you crunch, the greater the chance of discovery.

You can never have too much overkill.

No need to restrict yourself to distributed computing.

[dcheuk]

I just ordered a 3950x, 16 cores, that's probably an overoverkill LOL

What mb are you guys using? I'm thinking about going on an matx build in a fratal design meshify mini but there's only one matx board for am4 socket (x570). talking about lack of options.

This is going to be my first ryzen build so I don't have much experience with zen motherboards. I didn't want to go itx because there's only two ram slots and my graphics card, psu and aio is not gonna fit inside.

[Runtime Error]

I thought I could add to this just by pointing to another thread.

Originally Posted by neatherback

I'm currently doing some PRP tests and its running fine, but all the WR holders seem to be using Pentiums and Core 2 Duos, is there a reason for this?

The folks who build dedicated machines for GIMPS tend to optimize differently than those who build good "all around" machines and simply run GIMPS when the machine would otherwise be idle. For instance, George Woltman (who is Prime95 on the forums) started a very interesting thread a few years ago for a dedicated build. To quote him, his analysis looked something like this (keep in mind, this was December of 2015):

Originally Posted by Prime95

This is how I went about deciding my optimal dream build. Let's start with a base line 5 CPU system using overclocked memory:

5 ASRock Z170M-ITX/ac motherboards @130 = 650
5 2x4GB DDR4-3200 @60= 300
5 I5-6600 CPUs (3.3GHz, 65W) @230 = 1150
1 Samsung 850 EVO SSD @90 = 90
4 PicoPSU picoPSU-120 @40 = 160
1 Case, power supply, network switch -- approximate value $$100
Each of the 5 units will consume 65W CPU, 4W memory, 15W(?) mobo or about 425W total. Add in 15% power supply inefficiency for a total of 500W at the wall.
Total cost of 3 year ownership = 2450 parts + 3 * 500 = 3950
Total cost of 4 year ownership = 2450 parts + 4 * 500 = 4450

Now lets guess the throughput of this system using the Haswell data posted earlier. A 2.2GHz Haswell with DDR3-2133 gets 131.8 thoughput. In this system, each CPU will run 50% faster (3.3GHz vs. 2.2GHz) with 50% faster memory (DDR4-3200 vs. DDR3-2133). Thus 131.8 + 50% = 197.7. Actually should be better than that since Skylake CPU is slightly more efficient than a Haswell CPU. But we'll leave the expected throughput number at 197.7

Now lets define a metric to optimize -- expected throughput per dollar (TPD).
3 year TPD = 5 CPUs * 197.7 * 3 years / 3950 = 0.7508
4 year TPD = 5 CPUs * 197.7 * 4 years / 4450 = 0.8885

Let's compare that to a second system built with cheaper motherboards that do not allow overclocking. We will save $60 for each motherboard and $20 for each RAM pair, for a total of $400. Expected throughput for each CPU is 165.4 (that is what a 3.4GHz Haswell gets using DDR3-2133). Now let's look at our TPD metric:
3 year TPD = 5 CPUs * 165.4 * 3 years / 3550 = 0.6989
4 year TPD = 5 CPUs * 165.4 * 4 years / 4050 = 0.8168
Not nearly as good as the previous system.

Now we'll try a cheaper 3.2 GHz CPU in the base system. This saves 25 dollars per CPU. Expected TPD won't go down much probably to 193 or 194.
3 year TPD = 5 CPUs * 193.5 * 3 years / 3825 = 0.7588
4 year TPD = 5 CPUs * 193.5 * 4 years / 4325 = 0.8948
That's better than the first system

How about overclocking? Each K-series CPU will cost $50 more. I assume power draw is proportial to frequency and the square of the voltage. As an example, lets target a 200MHz frequency increase. I'll assume a frequency increase requires a tiny voltage bump. Thus, CPU power goes from 65W to 65 * 3.5/3.3 * (1.17/1.15)^2 or an increase of 7.5W when taking power supply ineffiency. This is less than the 91W TDP listed for K-series CPUs. A 5% increase in throughput is a fairly generous assumption -- 197.7 * 1.05 is 207.6.
3 year TPD = 5 CPUs * 207.6 * 3 years / (3950+5*50+5*7.5*3) = 0.7221
4 year TPD = 5 CPUs * 207.6 * 4 years / (4450+5*50+5*7.5*4) = 0.8561
Not worth the money.

Conclusion: It is best to create an overclocked memory system using the cheaper I5-6500 locked processor.

The optimization problem for his GIMPS dedicated hardware involved maximizing total throughput per dollar spent over several years of operation. The idea is that several "less powerful" machines are able to produce more throughput than one insanely powerful (and expensive) machine. That doesn't mean that fancy processors are necessarily overkill, but they aren't the best bang-for-your-buck if you only care about finding Mersenne primes.

As others have mentioned, Prime95 gets memory bottle-necked if the CPU's cache isn't big enough to handle the FFT size. For instance, I have a machine with an Intel Core i7-8700 processor which has 6 cores, but only 4 of them can be used efficiently with Prime95 because it is a dual channel memory chip with smaller cache. For this application, that specific chip is "overkill" in the sense that it can't use all of its cores for this project. The 3900x is also dual-channel, but perhaps it can use most cores for current test FFT sizes due to its larger cache.