In this era of fake news, folks rarely get past the headlines and delve into the numbers, so let's do just that.
1. Headline: Increased RAM speed has no effect on Gaming Performance
As for testing, tell me what ya wanna prove and I'll give ya a test that proves either side of the argument.
Memory Speed has no effect - testing with Metro 2033 proves it
Memory Speed has significant effect - testing with F1 proves it (11%)
http://techbuyersguru.com/gaming-ddr4-memory-2133-vs-26663200mhz-8gb-vs-16gb?page=3
6700k w/ 98 TI and 16GB
Crysis 3 - 2133 = 61.7 / 2666 = 62.1 / 3200 = 64.0
The first comeback ya get when posting those results is ... yeah but the increase is insignificant, no one will notice the difference. I must say that is certainly a valid argument at just 3.7%. So lets consider a moderately priced $1200 build ...
does it make sense to invest in DDR4-3200 ? Yes, the 3.7% increase in speed is somewhat insignificant, however what is many times more insignificant is the the increase in system price at only 0.8%. In other words, your return on investment is 4.6 to 1. Let's look at some other games, but I first want to note the absence of games like F1 which are known to be memory intensive in the test
http://techbuyersguru.com/gaming-ddr4-memory-2133-vs-26663200mhz-8gb-vs-16gb?page=3
BF4: 82 w/ 2133 to 84 w/ 3200
Ryse, Son of Rome: 78.3 to 81.1
Grid Autosport: 119 to 122.4
Grand Theft Auto V: 61.3 to 61.6 (minimum speed increases from 40.7 to 43.4 = 6.4%)
The Witcher 3: 53.5 to 55.9 (47 to 49)
RoTR went from 63.5 w/ 2133 to 63.7 w/ 2666 and dropped to 62.7 w/ 3200.
On average (all 7 games tested):
2133 = 65.7 avg fps (54.4 min)
2666 = 66.5 avg fps (54.9 min)
3200 = 67.2 avg fps (56.1 min)
Doing the math....
Going from 2133 to 2666 for $5 get you a 1.2% performance increase in avg fps (0.9% min) for a 0.4% increase in system cost. That's of ROI of 3.0 to 1 on avg fps and 2.3 to 1 on min fps
Going from 2666 to 3200 for $4 get you a 1.05% performance increase in avg fps (2.2% min) for a 0.3% increase in system cost. That's of ROI of 3.5 to 1 on avg fps and a pretty significant 7.3 to 1 on min fps
So while the performance gains might be considered "insignificant", it's clear that the cost bump is much more insignificant being several orders of magnitude larger. The move from 2666 to 3200 (3.5 / 7.3) clearly has a better ROI than the move from 2133 to 2666 (3.0 / 2.3).
2. Headline: DDR4 over 1.35 volts is "scary scary".
RAM up to 3200 uses the same 1.35 volts. So "ooh scary scary voltage thing" is a non-issue. But,
according to Intel, they have no issue with 1.4 and permit up to 1.5 on DDR4 for XMP. Exceeding the maximum voltage cited in the JEDEC specification is "not a thing" and has not been since this whole business started with Sandy Bridge.
Again, this comes from the fact that the JEDEC spec only goes up to 1.35 volts. But for as long as we have had XMP, higher voltages have been ... and remain to be fully supported and warranted by Intel, MoBo Manufacturers and RAM manufacturers. As long as Intel is listing DDR4-4266 RAM on their "Certified Compatible" RAM list, and allowing up to 1.5 volts for XMP, there is no issue here... it's is most certainly "perfectly fine".
http://www.legitreviews.com/what-is-the-safe-voltage-range-for-ddr4-memory-overclocking_150115
Legit Reviews contacted Intel about the safe voltage range on DDR4 memory and we received this response
1.5v is the absolute max we allow for XMP certifications. However, good DDR4 memory will run at 1.35v up to 3200.
Despite thousands upon thousands of warning posts on this topic, I have yet to see a single post stating "I blew my IMC running my RAM at the rated voltage on my memory package". I have seen forum posts from the major MoBo vendors and personally received e-mails from Asus, MSI and Gigabyte stating that they have no issue with running any compatible RAM at the rated voltage on the package and up to 0.075 above that. I have seen folks running 1.65v XMP rated Mushkin RAM (supposedly limited to 1.5) at 1.94v w/o issue .... I have 3 boxes in this room, running at 0.050 to 0.075 above rated voltage for 3 to 6 years and these machines are all on 24/7. If this was a problem, RAM manufacturers, MoBo manufacturers and Intel would not be listing this voltages on their published spec sheets, QVLs, packaging, etc as then they are on the hook for replacing anything that dies.
If Intel is willing to go on record saying 1.7v is just fine for XMP with DDR3 and 1.5v is just fine for XMP with DDR4, I don't see what all the fuss is about. Does it have an impact on component life ? Yes, all overclocking does .... do we care ? If we did, no one would be buying K series processors or using XMP ... we just don't care because we don't care how long it might have lasted if the stuff wasn't in that dusty PC in the corner in the spare room, a drawer some where .... or in a landfill.
Facebook
Twitter
Instagram