Question If I get more memory, will games run faster?

[ZöldSzakáll]

I want to get more ram. Now I have 4 gb ddr3 1600mhz ram. I want another 4 gb
My computer
I3-3240
Gt 640 1gb

If I double my ram can I get better fps in games Like rust, rainbow six siege?

 

[compprob237]

I'm not going to sit here and line by line your thread. Aside from your condescending tone trying to "educate me" your anecdotal evidence doesn't trump anyone elses. In many many years working with this I know my statements to be correct. So instead I'm going to tell you you have the best possible resource right here, an account on this forum. Spend any time in the memory or systems areas of this forum and you will see how wrong you are.

Dell has made millions of computers. Guess how many of those systems had "multi channel" kits put in them? Want to know what percentage of those systems had bulk-ordered RAM that was never tested together but yet works just fine? You're wrong. This forum explicitly has people having issues with a system so your sample set is heavily biased.

 

[bignastyid]

Dell has made millions of computers. Guess how many of those systems had "multi channel" kits put in them? Want to know what percentage of those systems had bulk-ordered RAM that was never tested together but yet works just fine? You're wrong. This forum explicitly has people having issues with a system so your sample set is heavily biased.

As a former tech support employee for a large OEM it was not uncommon for partial memory replacements to have problems and more ram having to be sent out to get sticks to work together. Also the systems built with mismatched sets in the factory are tested before leaving the factory as part of QC, so technically they do end up being tested together before being sold. Usually the higher end systems did use matched sets and were also replaced with matched sets.

Mismatched sets can work just fine, but they can also be problematic which is why most recommend just using matched sets as the best practice.

 

[Darkbreeze]

I'm not going to line by line it either, I will however do this. This, which is only ONE example, by a member that is no longer with us (Literally), who probably knew more about memory and memory architectures than anybody, BAR NONE, that frequents this forum aside from probably Pinhedd and Invalid error, that completely refutes just about everything in that diatribe of nonsense.

Just Add More DRAM

JEDEC is a council of electronic-device manufacturers and design firms that sets industry standards to be universally adapted by its members. Because some DRAM manufacturers decided to exceed the JEDEC maximum DDR3-1600 CAS 11 (and, later, CAS 9) by offering tighter timings and higher data rates, mixing DRAM has not been as easy as the council intended.

Simply put, mixing DRAM from different packages is a crapshoot, even when you have two identical packages of the same exact DRAM model. I would like to add that DIMMs that don’t appear to work well together often, but not always, can be helped with voltage and/or timing adjustments. There are a couple of examples in "DDR3 Memory: What Makes Performance Better?" where two of the companies didn’t market 32GB sets of 2400 MT/s DRAM and sent me a pair of matching 2 x 8GB sets. Neither worked initially, and it took minor adjustments for them to run smoothly.

Why is this such a problem? After all, they are the same frequency, timings and voltage.

DRAM is made up (basically) of memory chips that are soldered to a PCB (printed circuit board), driven by electricity. During the course of a DRAM production run on a given model, the manufacturer might be finishing up a large section of PCB that has been cut to the DIMM size but then might switch to a new PCB from a different production lot, which can result in slightly different properties. The same can happen with the physical solder; the manufacturer may change to a completely different kind that has ever so slightly different conductivity properties. Then, there are the dies themselves. When made by the actual chip manufacturer, the chips are "binned" (sorted) according to their quality.

Let’s take a closer theoretical look at this concept. A single production lot may result in, say, 1000 memory chips, which are separated or binned. A manufacturer may classify 200 chips as entry-level, and separate 350 chips that are a little better, 300 chips that are even better and 150 chips that are the best. Then, they sell these chips to different manufacturers.

If you were to go out and buy an 1866 MT/s module from each company, you would likely be getting a different PCB in each, a different solder with various conducting qualities and quite possibly differently graded chips and/or chips from different manufacturers. Several companies are making memory chips, which further adds to the questions about compatibility, and you might start to see why mixing DRAM can be, and often is, problematic. We also noted earlier that most newer lines of DRAM use 4Gb densities, whereas the norm with older lines was 2Gb.

That is found here: https://www.tomshardware.com/reviews/ddr-dram-myths,4155.html#p1 in an ACTUAL article, by a guy who ACTUALLY knows what he is talking about.

Not to mention, my Anecdotal evidence, both from working on and building systems these past 30 years, and working this forum for the last five, suggests that that monologue upstairs is pile of garbage based not on reality, but one persons limited experience on a few systems that didn't happen to encounter any issues. And that basically amounts to the same thing as a grain of sand flicked into the ocean in the grand scheme of things.

 

[InvalidError]

I have multiple anecdotal examples, as well as links specifically to Intel's and AMD's statements, that oppose this predication.

The problem with mixing DIMMs isn't so much that it won't work that you have absolutely no guarantees by anyone that it will as each vendor only qualifies their memory as same-model kits on specific motherboards. The higher the memory clocks and tighter the timings, the more likely you are to run into complications running mixed kits. Even when using all identical model DIMMs, some kits aren't qualified for use beyond one DIMM per channel and have no guarantees that adding a second identical kit or using a quad-channel kit will work at rated clocks and timings if at all. That's before accounting for the CPU manufacturers themselves not officially guaranteeing speeds over 2666MT/s and manufacturers like AMD having different official ratings for each specific DIMM configuration.

Depending on your luck, memory compatibility can turn into a deep rabbit hole.

Two weeks ago, my mother gave me her old laptop which has 4+2GB DDR3. I happen to have a half-destroyed (but still working) laptop which also has 4+2GB DDR3. I decided to swap the 2GB DIMM in my mother's laptop with my old laptop's 4GB so I could have a more usable spare laptop. My mother's laptop had no problem dealing with the two completely different DIMMs, the other laptop wouldn't boot until I switched the 2GB DIMMs between slots.

 

[Rogue Leader]

Dell has made millions of computers. Guess how many of those systems had "multi channel" kits put in them? Want to know what percentage of those systems had bulk-ordered RAM that was never tested together but yet works just fine? You're wrong. This forum explicitly has people having issues with a system so your sample set is heavily biased.

And this right here shows exactly my point.

Bulk ram from the same supplier still came from the same batch, from the same machine, from the same slabs of silicon. So chances are it will work together. But 2 different dimms, of different specs, or even the same model but ordered at a completely different time did not come from the same batch, maybe not even the same machine, heck it may even be different brand chips (Corsair, Gskill, etc use both Samsung and Hynix and sometimes mid production change them). Hence... maybe no work.

I see some others have chimed in with more than enough proof. Me, I literally used to work for Veeco, they make the machines that all these manufacturers use to make DRAM chips. So again, tell me how your anecdotal evidence based on an extremely minor sample set is better than the multitude of proof presented to you? Does this sample set use the same logic you use for testing CPUs against eachother without a GPU bottleneck?

 

[compprob237]

As a former tech support employee for a large OEM it was not uncommon for partial memory replacements to have problems and more ram having to be sent out to get sticks to work together. Also the systems built with mismatched sets in the factory are tested before leaving the factory as part of QC, so technically they do end up being tested together before being sold. Usually the higher end systems did use matched sets and were also replaced with matched sets.

Mismatched sets can work just fine, but they can also be problematic which is why most recommend just using matched sets as the best practice.

This mirrors what I've said. There can be issues, yes, and in general it is not recommended but they can, and do, work together. This post inadvertently supports my assertion. It also proves Intel's and AMD's Flex mode specifications correct. Some further reading on this. Before you even go to cite this link as refutation I will quote what I believe will be cited:

There are no guarantees that any DRAM you buy will play well with what you already have.

Yes, no guarantees but they can work together. I've never disputed this.

It lets you use DRAM of different capacities to enable the exploitation of the motherboard’s multi-channel architecture whenever possible.

Ram can be VERY finicky, and in many situations not using a matched set results in a system that either is not stable or doesn't boot at all. Thats what he was referring to. If you want your best chance of it working and not being stuck with ram that you can't use you need to either buy a matched set, or at a minimum buy the identical DIMM to what you have. For example if you try to run a Samsung and a Hynix DIMM in the same system in most cases it won't work at all.

I see some contradictions. Flex mode lets you use unmatched size sticks which implies they're not matched sets even if they're simply different composition (dual vs single rank). 👍

Where I've personally seen issues is when the sticks deviated from JEDEC standards (faster/lower latency), the IMC of the CPU was very poor quality (a rare few being the cause of repair in the first place), issues with the BIOS (corruption/support) and/or chipset IMC (if it has one like older boards), and finally when there was an issue with the replacement stick itself (DOA/faulty/manufacturing defects that make them more difficult for the IMC to deal with).

Nowhere have I said that you should mismatch sticks. I've clearly stated many times that you should, in fact, buy a "matched" set. Where my contention is with the assertion that they do not work. Yes, we can both agree that there's situations where they can have issues but you must admit that they can work. This is the entire basis of my original argument. Matched kits are suggested but not, necessarily, required.

I'm not going to line by line it either, I will however do this. This, which is only ONE example, by a member that is no longer with us (Literally), who probably knew more about memory and memory architectures than anybody, BAR NONE, that frequents this forum aside from probably Pinhedd and Invalid error, that completely refutes just about everything in that diatribe of nonsense.

That is found here: https://www.tomshardware.com/reviews/ddr-dram-myths,4155.html#p1 in an ACTUAL article, by a guy who ACTUALLY knows what he is talking about.

Quoted article further proves my point. They're not guaranteed to work and even the points made in the article can happen to "matched" sticks. Again:

In fact, most manufacturers only warrant that their kits were tested to work together. Not that they will work together on everything. This is why QVL's exist in the first place as they're tested by the manufacturer to ensure that the kit does work on the system (even if cherry picked by RAM manufacturers for DDR4-3866+ kits).

More specifically in that article it is talking about when the sticks deviate from JEDEC standards.

Because some DRAM manufacturers decided to exceed the JEDEC maximum

When you deviate from the standards is when issues arise. It takes carefully selected ICs to achieve speeds above the JEDEC spec of DDR4-2133. These ICs are all combined per-stick and usually per-batch to ensure they perform similarly but is not necessarily a requirement. This is why any RAM kit installed in a Ryzen system that is sold to run at any speed above DDR4-2133 can have issues. Even then, as I've pointed out, there are further restrictions to these speeds that are explicitly outlined by AMD themselves. For example:

Looking at the chart in the link it appears that four sticks of dual-rank memory, which results in DDR4-1866, will be the hardest to overclock due to the IMC. This is not because the "Ram can be VERY finicky" but because the IMC itself is finicky.

I assume you've seen the threads on this forum about Ryzen and RAM compatibility. Under this assumption you should be readily aware of the issues of perfectly good RAM kits not working with specific CPUs or motherboards. How many of these threads say "Works at DDR4-2133 but not at the rated speed"/"when XMP enabled"? Why is there a stickied thread explicitly outlining how to get your RAM to run at a certain speed or its XMP and explicitly goes in to Ryzen having issues? Is it because the RAM doesn't work with the CPU or is it the CPU's IMC that doesn't work with the RAM? 🤔

Not to mention, my Anecdotal evidence, both from working on and building systems these past 30 years, and working this forum for the last five,

I can dismiss your anecdotal evidence just as readily as you've dismissed mine in your message quoted further below. Yes, mixing RAM is problematic as I've said but it does work and is why Intel and AMD support features specifically for this in their CPU's IMC.

that monologue upstairs is pile of garbage based not on reality, but one person's limited experience on a few systems that didn't happen to encounter any issues. And that basically amounts to the same thing as a grain of sand flicked into the ocean in the grand scheme of things.

What's that one saying when you assume? You're doing that right now. Perhaps you may not have noticed this part:

these are the ones I remember off the top of my head

You don't know what I do and are clearly assuming that I have very limited experience with computers to the point of being "a grain of sand flicked into the ocean in the grand scheme of things." Very belittling and rude language from a moderator.

Read First: Forum Rules & Styling Posts

Rules of Conduct The Tom's Hardware Rules of Conduct are general guidelines for users within our community. We pride ourselves on being welcoming to users of all levels of expertise, from novice to guru. Follow these simple rules and you'll have a great experience here on the forum. Please...

forums.tomshardware.com

Please don't...
...
• Be rude or impolite. Civility is essential on Tom's, and remember that behind each user is a real person. Personal attacks (ad hominem) and insults are not allowed.

I suggest you try simply stating that you dispute the information instead of resorting to being rude or impolite.

@InvalidError is exactly right. 👍

And this right here shows exactly my point.

Bulk ram from the same supplier still came from the same batch,

Nope. It's likely but not guaranteed.

from the same machine,

LOL

from the same slabs of silicon.

Can be likely but only if the number of sticks purchased in bulk are smaller than or match a single slab of silicon and even then this isn't guaranteed. The orders that I mentioned would amount to multiple slabs of silicon with all the sticks placed in bulk storage packaging after final assembly which further mixes them.

So chances are it will work together.

Yes, so long as they adhere to JEDEC's specifications.

But 2 different dimms, of different specs, or even the same model but ordered at a completely different time did not come from the same batch,

Of course there's no guarantee of this.

maybe not even the same machine,

Did you know that many fabs have multiple machines? More machines results in higher output and having multiple machines running so that other longer processes can be ran in parallel further increases throughput but at the cost of there being more machines.

heck it may even be different brand chips (Corsair, Gskill, etc use both Samsung and Hynix and sometimes mid production change them).

Less likely, especially for the very high end like DDR4-3200C14 that are exclusively Samsung B-Die. Usually when they switch the chips it will receive a different model number. This is especially important for tech support, RMA, and warranty services.

Hence... maybe {it wont} work.

Of course. I never guaranteed they would work. If they conformed to JEDEC spec then it is highly likely to work but there's many other factors, that I've already outlined and have been linked to, that could contribute to them not working.

I see some others have chimed in with more than enough proof. Me, I literally used to work for Veeco, they make the machines that all these manufacturers use to make DRAM chips. So again, tell me how your anecdotal evidence based on an extremely minor sample set is better than the multitude of proof presented to you? Does this sample set use the same logic you use for testing CPUs against each other without a GPU bottleneck?

I can dismiss your anecdotal evidence just as readily as you've dismissed mine. Working for the factory that makes the machines isn't the same as working for the factory that verifies the end product the machines create. There's a multitude of proof here that contradicts what you claim. This proof includes the very posts you've mentioned here that you claim refute what I've said. I'll continue to be a broken record: "There can be issues, yes, and in general it is not recommended but they can, and do, work together." I will point out that this statement does not guarantee that the kit will run at a specific speed or set of timings. It simply states that they can work together.

I'm having fun with this conversation.

 

[Rogue Leader]

I'll continue to be a broken record: "There can be issues, yes, and in general it is not recommended but they can, and do, work together." I will point out that this statement does not guarantee that the kit will run at a specific speed or set of timings. It simply states that they can work together.

Thats not what you said. What you said after making a personal insult to someone else basically saying they don't know what they are talking about when they said practically what you just said above was:

https://www.intel.com/content/www/us/en/support/articles/000005657/boards-and-kits.html#dual
"Rules to enable dual-channel mode
To achieve dual-channel mode, the following conditions must be met:

• Same memory size. Examples: 1 GB, 2 GB, 4 GB.
• Matched DIMM configuration in each channel.
• Matched in symmetrical memory slots.

...
The following conditions do not need to be met:

• Same brand
• Same timing specifications
• Same speed (MHz)

The slowest DIMM module populated in the system decides memory channel speed."

That was the beginning of this argument, where you didn't even acknowledge that there could be an issue. Then you posted a monstrous pile of anecdotal evidence backing yourself up asserting that you were right and it will always work.

Now you post something agreeing with EVERYTHING we have said, as if thats what you were saying all along, but it is not. But continue to insist you are right.

This is no longer worth discussing with you. You can keep posting, (although if you continue to derail the thread with this I'm sure an outside Moderator will come along and shut this down), but I'm out.

 

[Rogue Leader]

Actually let me say one last thing. If the OP goes out and buys a DIMM that looks like it matches his, by your recommendation, and it doesn't work with his memory, and the store won't take it back, who is going to pay for it for him? Who will support it, or refund it? You?

Here on this forum we answer questions for real people spending their real hard earned money, some of which can't afford mistakes. So when you post in absolutes, as you did, and they are wrong, you need to expect to be called out on it by others who know that you CAN'T give an absolute answer. Hence where I mentioned to make sure to buy from a store with a good return policy just in case.

This is a responsibility you take on when answering questions here. And a responsibility we take on as moderators to do the best we can to make sure the OP gets the safest answer possible.

 

[Darkbreeze]

This forum explicitly has people having issues with a system so your sample set is heavily biased.

It is, but it also supports the standard recommendation we make against not deliberately trying to use mixed sets. Also, where those samples relate to memory it is OFTEN due to the use of unmatched DIMMs. USUALLY, in those cases, where people are having issues, USING matched DIMMs resolves the problem. This doesn't refute the recommendation, it solidifies it.

In EVERY case? No. Of course not. Often enough to not call it a MYTH? Yes.

So, this assumption that you must purchase a multi-channel kit, or the exact same stick, is an unfounded one based on myths.

Often enough to make it a standard recommendation? Yes.

Nobody has disputed the idea that using mixed memory CAN work. Of course it can. But as you said, on memory that deviates from the JEDEC specifications, which for OUR members, in threads where this even gets brought up are concerned or where questions of incorrect capacity is noted, dual channel operation is not happening but should be or there is a complete failure to POST due to memory, they are USUALLY not instances where somebody is running or trying to run a default configuration, but even if they were, we actually, literally see this problem even without factoring in the compatibility issues found when trying to to run disparate kits WITH their XMP profiles enabled.

Mixed memory can work, of course. Mixed memory can NOT work, for certain. Not intentionally putting yourself into a situation where you MIGHT run into that problem makes more sense than any other scenario. Period. If you have sticks sitting around that might be useful, by all means, try them. If it works, great. If it doesn't, nothing lost. But don't go out and intentionally buy a potential problem for yourself. It's no myth. It happens, a lot. And not just because the kit in question deviates from the JEDEC standards.

As has already been mentioned, and as I have mentioned myself in many thread for which I will provide, once again, a good example, differences from production run to production run even on kits that otherwise seem to have identical specifications, may not run together AT ALL, not even at the default configuration much less at profile specifications.

Differences in secondary and tertiary timings, even when the primary timings are all the same, or differences in ICs, ranks and rows, might all be reasons why the hardware cannot find a way any configuration where everything is happy.

Here is one example of why. Three modules, all with the same part number, all with different configurations. As for anecdotal evidence, I happened to have had sets of all three sticks on hand and could not get any of them to work with the others, even though they were technically ALL the same part number. Not even just to POST at the default baseline JEDEC configuration with no profiles enabled. They simply would not work together, at all. Tried them on three different boards with the same results. A Hero VIII, a Z170X-Gaming 5 a Z270m Extreme4. Also with two different CPUs. A 6700k and an i5-7600k. It should be fairly obvious that given the known problems regarding memory on Ryzen, the recommendation not to mix things up becomes even a bit more imperative.

 

[compprob237]

What you said after making a personal insult to someone else basically saying they don't know what they are talking about when they said practically what you just said above

Thus, dual rank (total of 16) 512Mb composition sticks will work in the system.

This is specifically in reference to the X58 chipset/i7-900s/X54xx/X56xx if taken in context

While it is good practice to purchase the same composition (dual vs single rank), manufacturer of internals for RAM modules, and same manufacturing die it does not necessarily mean that you must purchase a multi-channel kit, specifically, to ensure they work together.

This is the closest my post gets to saying they "always work together". Nowhere did I say it will always work. I did state absolutes but they were not specifically in reference to mixing RAM working. They were true statements in reference to RAM manufacturer warranties and can be considered proper absolutes.

Hey, I'll admit that I was wrong by insulting someone that was actually correct. This is why I left my post unedited. My mistake was skimming the response to OP's question and misreading it. Should I write a formal apology letter and send it to Calvin7? I'm not being sarcastic here. I really did make a mistake reading their post and after re-reading it a few times I'm truly sorry for that mistake. That's exactly why I did not dispute the moderation action.

Actually let me say one last thing. If the OP goes out and buys a DIMM that looks like it matches his, by your recommendation, and it doesn't work with his memory, and the store won't take it back, who is going to pay for it for him? Who will support it, or refund it? You?

The same can be said if you, for instance, suggest anyone buy a product with no return policy/"wont take it back". I even take this in to consideration when I link from third party sellers such as through eBay. The responsible party for the return/refund is the retailer or middle man. eBay and sometimes Newegg (Newegg marketplace), specifically referring to my suggestions, are technically the middle man.

Here on this forum we answer questions for real people spending their real hard earned money, some of which can't afford mistakes.

Yes, I'm completely aware of this. This is why I typically try to exhaust all free avenues of troubleshooting before I attempt to proceed with purchase suggestions for replacement parts. I even go so far as to suggest that they see if they can borrow a friend's working system/component to verify the issue.

So when you post in absolutes, as you did, and they are wrong, you need to expect to be called out on it by others who know that you CAN'T give an absolute answer.

Can you point to the specific statement where I said in absolute form that it will work, guaranteed? The post quoting Intel's "rules to enable dual channel" is not me asserting an absolute. It was merely posting the contents of the article.

Hence where I mentioned to make sure to buy from a store with a good return policy just in case.

Hence why the links I provided to OP were from Newegg and Ebay. Hopefully I'm accurate in assuming you're aware of Newegg's fantastic return policy. Are you aware of eBay's Money Back Guarantee?

eBay Money Back Guarantee means you're protected if the item you ordered didn't arrive, is faulty or damaged, or doesn't match the listing. You'll get your money back.

I now realize that my post about CPU performance testing was mixed up with this thread and the originally intended one. Oh well. 🤷‍♂️


Tell you what. I'll drop this entire conversation just as you are. It's been real fun debating this even if I was making assumptions (assuming message intent), a lot mistakes (like the CPU performance testing bit, misreading Calvin7's post, not properly explaining, etc), or numerous errors in my own judgement (insulting someone I shouldn't have, getting rude with others, continuing to derail a thread that poor OP is likely confused by, and writing unnecessarily dissertations - I tend to over-explain). My original goal was to help OP make the best decision for the lowest cost. I just hope that's understood.

Since I am dropping this off-topic conversation I'm not going to continue to respond to other posts here unless it is related specifically to helping OP. If anyone that I've quoted or mentioned wants to talk to me about what was being discussed here then I'd prefer PMs or perhaps a new thread. Whichever they feel is proper.