I was thinking about buying an Acer Aspire E 15 E5-575-33BM laptop. I decided not to. Initially, I was attracted by its low price tag for great features and a high rating on Amazon. But as I dug into users’ comments, I thought the ratings would probably start to decline. It seemed the unit might not be very well constructed.
Nonetheless, I did accumulate some notes on that laptop’s hardware upgrade options. This post presents what I learned about updating its CPU and RAM.
CPU Upgrade Options
The first question was whether I could upgrade the processor. The system specifications indicated that the stock CPU was an Intel Core i3-7100U. Intel’s page for this CPU said that it used a FCBGA1356 socket. Wikipedia indicated that FCBGA was a type of BGA socket; MakeUseOf (Smith, 2013) confirmed that BGA CPUs are soldered directly to the circuit board.
Wikipedia noted that certain enterprising souls had invented do-it-yourself means of soldering CPUs to circuit boards. I did not intend to experiment with manually soldering a $281 CPU (see prices for 23 FCBGA1356 CPUs listed at Intel’s Processor Feature Filter). It seemed the more practical route would be to swap in a motherboard with a faster CPU, or simply buy a faster laptop.
Passmark rated the Core i3-7100U at 3915. Only six of those 23 CPUs were priced above $304. (The i3-7100U was $281.) All six were Core i7 CPUs: four at $393 each (namely, the 6500U, 6600U, 7500U, and 7600U) and two at $415 (i.e., the i7-6650U and 6660U). I did not check all of those in Passmark, but I saw that the i7-7600U was rated at 5581. Another possibility was the Core i5-7300U, listed at $281, and rated at 5041.
I assumed that a Core i7 CPU (and probably a Core i5 as well) could impose demands (for e.g., power, heat dispersion) that would be best accommodated on a motherboard designed for it. A search led to various questions and answers along these lines. As a new purchaser not eager to void his warranty, however, I tentatively concluded that a CPU upgrade was not presently a realistic option.
RAM Upgrade Options
To obtain full information on this computer’s memory, I downloaded what appeared to be the correct User’s Manual from the Acer website. It was a general manual for all Acer Aspire E 15 laptops. Unsurprisingly, this manual did not have a Specifications section: it couldn’t, if it was to cover so many different models. A search of the Acer website did not seem to lead to any manuals specifically for this model. I encountered more than one discussion thread expressing frustration at the inability to locate service manuals and the uninformative nature of the user’s manuals for Acer machines.
The Acer webpage for the E5-575-33BM indicated that it came with 4GB of DDR4 SDRAM, upgradeable to a maximum of 32GB in a total of two slots. I expected to add one 16GB stick, possibly keeping the existing 4GB. The webpage did not provide further details about the RAM (e.g., speed; number of pins). A search led to a CNET page indicating that the RAM supplied with the machine was 260-pin SO-DIMM 2133 MHz PC4-17000. That page appeared to say that that was also the maximum speed that the machine could use. That was confusing, because (a) 1 2 Amazon reviewers stated that the supplied 4GB RAM in their E5-575-33BM was DDR4-2400 and (b) Crucial offered allegedly compatible RAM upgrades ranging up to DDR4-2666. I decided to assume that Crucial was right — that the machine could indeed take advantage of increasing speeds up to 2666.
Those Crucial options ranged in price according to their speed and CAS latency (CL). There were four options in the 16GB one-stick offerings: (A) DDR4-2133 CL=15, (B) DDR4-2400 CL=17, (C) DDR4-2400 CL=16 (a/k/a Ballistix Sport), and (D) DDR4-2666 CL=16 (Ballistix Sport). Crucial indicated that speed (e.g., 2666) contributed more to performance than latency (e.g., CL=15). (I could also have explored the relatively arcane topic of RAM timings.) Wikipedia pointed toward an online spreadsheet that permitted user calculation of relevant values. That spreadsheet indicated that, among these four options (A) through (D), the latency measured in nanoseconds was less than 1% lower, for (B) vs. (A); about 5% lower, for (C) vs. (A); and 15% lower, for (D) vs. (A). Meanwhile, at Crucial, the price premiums for 16GB of faster RAM were 0% (from 2133 to 2400), 8% (from 2133 to 2400 Ballistix Sport LT), and 12% (from 2133 to 2666 Ballistix Sport LT). Given that the last figure (12%) translated into a difference of $15, it was tempting to just pay the small premium and get the extra speed.
There was, however, the question of whether the extra theoretical speed made any practical real-world difference. A search led to tests, by TechSpot (Walton, 2016), comparing the effect of different speeds of RAM, ranging from DDR4-2133 to DDR4-4000, on various real-world tasks. In most (e.g., Excel), the faster but more expensive 4000 produced only a modest speed improvement over the 2133. But in a few, the difference between 2133 and 4000 was significant (with smaller differences between e.g., 2133 and 2400). In Photoshop, for example, the DDR4-4000 completed its task 35% faster; in Handbrake, there was a 57% improvement. Summarizing his findings, Walton suggested that it made the most sense to pay for high-end RAM on high-end systems. In another test, ExtremeTech (Hruska, 2015) examined an AnandTech study and concluded that “the benefits of moving from DDR4-2133 to DDR4-3200 are tiny.” This was due to several factors, Hruska said, including “the fact that modern CPUs and software are all designed to hide or diminish the impact of latency.” Techgage (Williams, 2015) agreed: “Memory speed still doesn’t matter much.” More strongly, Tech Buyer’s Guru (Altman, 2016) said, “[W]e no longer recommend DDR4-2800 or DDR4-3000 for X99 systems despite their potentially lower timings because they require more voltage (1.35V) and a higher strap speed, leading to much higher power use at idle.”
For a laptop running on a battery, of course, that question of power use was important. A search led to other indications that, in theory and in users’ experience, all other things held constant, higher voltages meant higher heat. For instance, LI Blog (2016) said that the 1.2 volt standard of DDR4 results in power savings compared to the 1.5 volts of DDR3. Crucial confirmed that its four modules (above) all ran at 1.2V. A Newegg power search for compatible DDR4 RAM modules at the foregoing speeds (i.e., 2133 to 2666) likewise seemed to turn up modules mostly running at 1.2V. Even so, would I be running down the laptop’s battery more quickly if I chose faster RAM, or more of it? A search led to a SuperUser discussion, among others, suggesting that RAM accounts for a very small portion of system power usage, and that having more RAM could actually reduce power usage, if it means less use of hard drives. One participant in that discussion said, “Each memory module uses a mostly fixed amount of power.” Participants in another SuperUser discussion stated that more RAM means more power usage, and that other factors (e.g., the number of apps running, the efficiency of the antivirus software, the backlight level) are far more important in total system power usage. Perhaps most on point for my concerns, Tom’s Hardware (Angelini, 2014) found that two modules of Crucial DDR4-2133 RAM (total: 16GB) used just under 6 watts, while four modules of G.Skill DDR4-2666 (total: 16GB) used just over 6 watts, as did four modules (16GB) of Corsair DDR4-2800. In all three cases, power usage was half as high if half as many modules were used. Angelini noted that, in this study, DDR4 used 25% to 40% less power than DDR3.
Based on these inquiries, I planned to buy one stick of 16GB, rather than a kit of two sticks, 8GB each. Keeping the stock 4GB stick would require only a little more power, but I doubted I would need it, especially since I planned to use the SSD (below) for the paging file. It appeared that speed would have little to no effect on power consumption, and only minor impact on performance, so I thought I would probably look around for a decent compromise on price. With the future in mind, I would go for the faster RAM if there was little price difference.