Re: New PC Build Review
Posted: Mon Jul 29, 2019 10:30 pm
Here earlier, I objected to buying RAM DDR4 @3,000mHz today with the recent drops in RAM prices. Today, I found 16 Gb (2x8 Gb) DDR 4400 mHz for $169 LINK @ https://www.newegg.com/patriot-16gb-288 ... 6820225144
16 Gb (2x8Gb) 4133 mHz for $149
16 Gb (2x8Gb) 4000 mHz for $139
16Gb (2x8 Gb) 3866 mHz for $137
That's all from the same series sold under the Patriot brand. [Also, Don't use any of the above for AMD]
There is a Patriot DDR4 3733 mHz (CAS Latency 17 ms) 16 Gb (2x8 Gb) set at Amazon for $102
But Amazon was listed as the ONLY place to buy it...which just seems weird to me and I hate how Amazon's software is crap at organizing for marketing a heavily segmented brand under one product name like you get when searching PC parts: But here it is DDR4 3733 @ $102 https://www.amazon.com/dp/B07N43CYMS/?tag=pcpapi-20
If you wanna gamble a little on Price-to-Experience RAM, that $102 Amazon price is practically steeling
----------------------------------------------------------------------------------------
Anybody shopping for AMD may get a penalty from the AMD Infinity Fabric(TM) design which optimizes @ near DDR4 3600-3700mHz and from the likelyhood that cheap RAM often markets with a built-in Intel architectural bias which may require a little user frustration, BIOS setting tweeks, several restarts and a couple blue-screens before achieving the advertised speeds and latency, maybe even require a slightly lower-than-advertsied speed setting that you will likely never notice but is statistically likely to impact an application as RAM speed sensitive as P3d [Which is one of the few applications for which I have EVER heard of an end-user reporting notably improved experience after only a RAM speed upgrade] can be however you will never be certain which frame-stutter you see was a hickup in the RAM circuits vs the countless other peculiar hardware-performance sensibilities reported by P3d end-users over the year.
---------------------------------------------------------------------------------------------
RAM's a terrible place to sink ALL of your PC budget for such miniscule R.O.I. improvements likely no more than 1-4% frames per second improvements to the performance Lows & Averages, but RAM is also a terrible place to spend TOO LITTLE especially with a primary intended usage being like P3d which many users appear to max out available settings at whatever highest values are for running at nearest to the perceivable nearly 30fps video our eyes have become accustomed to seeing. But our FPS AVERAGE is not the same reporting measure-of-performace as a Hollywood movie. At which point, our every 1% low moment in the system's application-performance creates a visible stutter. So I'm saying RAM is an objectively terrible place for you to sink your entire build budget, but for GOD's SAKE a program as noticeably "sensitive" as P3d has been, RAM is a terrible place to spend too little. I think you should expect to pay about $10 per gigabyte of RAM, and even then you should check the memory-bandwidth to CAS latency reports on what you are paying for. If you wanna go a step further because those two numbers are too easy to compare, then you can go the extra-extra-step for assuring the end-user of higher performance by comparing the Bandwidth & CAS Latency levels between different modules operating at a given operating voltage.
---------------------------------------------------------------------------------------------------------
Performance sweet-spots appear application-specific and usually limited to finding the point at which the program appears to NOT bet RAM BOTTLENECKED by a specific Application + system-configeration. Settings higher than that one sweet-spot point often creates no significant measures of improvement. There are some exceptions. JUST TODAY, I saw a RAM benchmark video of FAR CRY: NEW DAWN performance that performed best on anything over DDR4-3000 mHz with no significant improvement scaling with higher RAM speeds above DDR4-3000. Immediately after, I saw a RAM benchmark for SHADOW OF THE TOMB RAIDER which objectively peaked with DDR4 3200 CL14 with which BOTH the FPS 1% Lows & 0.1% Lows being successfully lifted significantly above that vital 30 FPS visible-stutter point (0.1% Low of 33fps)
--------------------------------------------------------------------------------------------------------------
TLDR
LONG STORY SHORT: HIGHER SPEED, LOWER CAS LATENCY. IF YOU ARE GOING TO USE THIS SYSTEM FOR THE NEXT 5-YEARS DON'T HAMPER YOUR TOP-END PERFORMANCE WITH UBER-CHEAP VALUE ON RAM that will cost you 1-5% hardware performance on everywhere else invested in the system (or significantly more than 5% if too little RAM performance specific to an application). The RAM prices are cheaper than ever, but the market sweet-spot for reliably higher performance still appears to be around $10+ per gigabyte
16 Gb (2x8Gb) 4133 mHz for $149
16 Gb (2x8Gb) 4000 mHz for $139
16Gb (2x8 Gb) 3866 mHz for $137
That's all from the same series sold under the Patriot brand. [Also, Don't use any of the above for AMD]
There is a Patriot DDR4 3733 mHz (CAS Latency 17 ms) 16 Gb (2x8 Gb) set at Amazon for $102
But Amazon was listed as the ONLY place to buy it...which just seems weird to me and I hate how Amazon's software is crap at organizing for marketing a heavily segmented brand under one product name like you get when searching PC parts: But here it is DDR4 3733 @ $102 https://www.amazon.com/dp/B07N43CYMS/?tag=pcpapi-20
If you wanna gamble a little on Price-to-Experience RAM, that $102 Amazon price is practically steeling
----------------------------------------------------------------------------------------
Anybody shopping for AMD may get a penalty from the AMD Infinity Fabric(TM) design which optimizes @ near DDR4 3600-3700mHz and from the likelyhood that cheap RAM often markets with a built-in Intel architectural bias which may require a little user frustration, BIOS setting tweeks, several restarts and a couple blue-screens before achieving the advertised speeds and latency, maybe even require a slightly lower-than-advertsied speed setting that you will likely never notice but is statistically likely to impact an application as RAM speed sensitive as P3d [Which is one of the few applications for which I have EVER heard of an end-user reporting notably improved experience after only a RAM speed upgrade] can be however you will never be certain which frame-stutter you see was a hickup in the RAM circuits vs the countless other peculiar hardware-performance sensibilities reported by P3d end-users over the year.
---------------------------------------------------------------------------------------------
RAM's a terrible place to sink ALL of your PC budget for such miniscule R.O.I. improvements likely no more than 1-4% frames per second improvements to the performance Lows & Averages, but RAM is also a terrible place to spend TOO LITTLE especially with a primary intended usage being like P3d which many users appear to max out available settings at whatever highest values are for running at nearest to the perceivable nearly 30fps video our eyes have become accustomed to seeing. But our FPS AVERAGE is not the same reporting measure-of-performace as a Hollywood movie. At which point, our every 1% low moment in the system's application-performance creates a visible stutter. So I'm saying RAM is an objectively terrible place for you to sink your entire build budget, but for GOD's SAKE a program as noticeably "sensitive" as P3d has been, RAM is a terrible place to spend too little. I think you should expect to pay about $10 per gigabyte of RAM, and even then you should check the memory-bandwidth to CAS latency reports on what you are paying for. If you wanna go a step further because those two numbers are too easy to compare, then you can go the extra-extra-step for assuring the end-user of higher performance by comparing the Bandwidth & CAS Latency levels between different modules operating at a given operating voltage.
---------------------------------------------------------------------------------------------------------
Performance sweet-spots appear application-specific and usually limited to finding the point at which the program appears to NOT bet RAM BOTTLENECKED by a specific Application + system-configeration. Settings higher than that one sweet-spot point often creates no significant measures of improvement. There are some exceptions. JUST TODAY, I saw a RAM benchmark video of FAR CRY: NEW DAWN performance that performed best on anything over DDR4-3000 mHz with no significant improvement scaling with higher RAM speeds above DDR4-3000. Immediately after, I saw a RAM benchmark for SHADOW OF THE TOMB RAIDER which objectively peaked with DDR4 3200 CL14 with which BOTH the FPS 1% Lows & 0.1% Lows being successfully lifted significantly above that vital 30 FPS visible-stutter point (0.1% Low of 33fps)
--------------------------------------------------------------------------------------------------------------
TLDR
LONG STORY SHORT: HIGHER SPEED, LOWER CAS LATENCY. IF YOU ARE GOING TO USE THIS SYSTEM FOR THE NEXT 5-YEARS DON'T HAMPER YOUR TOP-END PERFORMANCE WITH UBER-CHEAP VALUE ON RAM that will cost you 1-5% hardware performance on everywhere else invested in the system (or significantly more than 5% if too little RAM performance specific to an application). The RAM prices are cheaper than ever, but the market sweet-spot for reliably higher performance still appears to be around $10+ per gigabyte