If only it wouldn't cost £1000's of pounds to have the decent kits.. I can't believe how expensive ram kits have got.. Here's hoping it goes down a lot at some point... soon...

phillIf only it wouldn't cost £1000's of pounds to have the decent kits.. I can't believe how expensive ram kits have got.. Here's hoping it goes down a lot at some point... soon...

Well thousands is exxagerating. Unless you are refrerring to 64GB and 128GB kits.
For example 4x8GB 4133Mhz CL19 kit costs 664€/591£: geizhals.eu/g-skill-trident-z-black-black-dimm-kit-32gb-f4-4133c19q-32gtzkkf-a1677142.html?hloc=at&hloc=de&hloc=pl&hloc=uk&hloc=eu

Considering I paid nearly 2 years ago for some cherry picked kits £330 for 32Gb (2 kits, 2 x 8Gb kits) that clock to 4000Mhz @ C16 I think (it's been a long time) I think the prices are horrendous... £600 for 32Gb is massively over priced in my eyes...

Not to mention performance improvements only actually improve in synthetic benchmarks where in real world, it barely makes any difference between 2666MHz kit and this 4266MHz...

Ram speeds are quite a waste really for most users I would say.. Getting something middle of the road will be more than likely were most would be best off getting.. Only the Elite benchers would care about the extra few MHz and timings.. I suppose the only way it might help a little more is when you have an APU or something that would actually make more use of the bandwidth.. I'm not so sure I'd spend more so on higher spec'd ram for a higher end CPU..

Wow that's like 3200 CL12.75 latency.

xorbeWow that's like 3200 CL12.75 latency.

without needing 1.9V - 2.1V.

RejZoRNot to mention performance improvements only actually improve in synthetic benchmarks where in real world, it barely makes any difference between 2666MHz kit and this 4266MHz...

In Fallout 4 it makes a BIG difference. In GTA V it makes a good difference. But as far as applications that would actually use bigger pools I am not sure. In games it definitely matters.

It doesn't. Not for the price offset. Paying 1000€ for hyper fast RAM is pure idiocy. Sure, if you piss money every morning straight in the toilet, of course you'll buy best of the best. But if you're a rational gamer (or user in general) you're gonna take same capacity RAM for 300€ and buy a top of the line graphic card for the 700€ that you've just saved from RAM. With RAM, you maybe gain 5fps, 10fps in very best possible situation. With top of the line graphic card, you'll gain 50fps on average, not at very best like with RAM. That's the difference.

Not long ago there was test of RAM from I think 2400MHz up to 4000MHz DDR4 and the biggest gains appeared up to 2666MHz RAM. Then the gain differences became smaller and smaller as the clocks increased. Given how steep is the price increase as you go higher with the clocks, it just doesn't make sense investing more money because gains just aren't worth it.

RejZoRGiven how steep is the price increase as you go higher with the clocks, it just doesn't make sense investing more money because gains just aren't worth it.

That's your personal opinion, not fact though. Even though the gains as measured may show low increases, the apps and games that I personally use tend to have some pretty big differences just from memory increases alone. CPU speed also plays a role as to how much those differences are noticed too. And for me, the price doesn't matter, and for others, the same is true too, as every industry has the same sort of products; ones that offer minor differences but offer hugely inflated prices compared to more "value-oriented" products.

So what I hear you saying is that for you yourself, the price isn't worth it. You'd rather spend your money elsewhere, and that's fine. But that doesn't make it a fact that everyone else needs to accept. Rationality has nothing to do with it. I have extra money, and I'll spend it how I like, because it's extra, and I like to give that extra to certain companies, like G.Skill, and I'll get added performance from it, too.

It is also worth noting that in order for these sticks to run 24/7 at these speeds means that they are a "higher quality" IC, that at lower speeds, will take more punishment, and last longer. That alone, the higher quality of the ICs, is worth paying for, if you actually value your dollar. It really sucks to have one stick die and have to send back an entire kit.

cadavecaEven though the gains as measured may show low increases, the apps and games that I personally use tend to have some pretty big differences just from memory increases alone.

That's pretty much not true. There are multiple reviews on the net including TPU where it's explictely shown that after about 3200 or 3600 MHz the gains are not noticable, and even up to 3200 only in certain games. If you have a different opinion, give some evidence or it's worth nothing. Again there are multiple reviews on the net that clearly state otherwise.

JackOneThat's pretty much not true. There are multiple reviews on the net including TPU where it's explictely shown that after about 3200 or 3600 MHz the gains are not noticable, and even up to 3200 only in certain games. If you have a different opinion, give some evidence or it's worth nothing. Again there are multiple reviews on the net that clearly state otherwise.

This is just like saying, that in those games there is almost no difference between an Intel i5 vs i7 or Ryzen. ;)
If a game is shitty codded (most of them are), it doesn't mean that the hardware to play it should be crappier too. Besides all those tests are done in ideal conditions, with no background stuff running so naturally the game can take full resources only for it. However, I always play games with hundreds of stuff opened, like browser window, file mangers, multiple downloads, always on antivirus/malware/idps,etc,etc, all those consuming CPU resources and RAM bandwidth. So yeah, there is a reason to buy stronger hardware for some people...

JackOneThat's pretty much not true. There are multiple reviews on the net including TPU where it's explictely shown that after about 3200 or 3600 MHz the gains are not noticible, and even up to 3200 only in certain games. If you have a different opinion, give some evidence or it's worth nothing. Again there are multiple reviews on the net that clearly state otherwise.

:lovetpu:

Meanwhile, I've been recommending people buy 3200 MHz C14 sticks since they went on sale. Like, you are right, but I quoted a poster saying nothing over 2666 mattered, and at times it does.

I've read nearly all those reviews, and I could write a review with hand-picked apps that show differently like it seems most reviews out there do themselves. You do need to remember, I write reviews for TPU. I know exactly how reviews can be twisted to show one thing or another. So in the end, it really does depends on your usage model. I've gone far further in my testing, however, and it's not truly the raw speed that matters; it's the overall latency. But that's the kicker... OVERALL latency. Very few things are bandwidth-constrained, but latency really kicks some apps into the dirt. It's had a lot to do with how hard CPU cache is thrashed, and this is shown in the opposite way by those Intel chips like i7-4950HQ, with 128 MB of cache. With that CPU, ram really does matter less.

That's why it was interesting enough (I think) for W1zz to do a memory analysis on 8700K... does the added cache help? It does. But also, you'll note that W1zz kept the CPU at stock. So what about if the CPU is OC'd? What about if you clock up the cache as well, or lower it?

Then, let's move over to the Ryzen camp... memory speed matter is a huge way, but why? It's because of what's linked to that same speed. Likewise, Intel chips have many different domains running at different speeds, and when those speeds are played with, interesting things happen.

When I find the time to write a proper review showing those instances where memory matters, I'll do so, but for me it is very important to explain exactly why it matters, and how, in hopes that we can get more interesting discussions about exactly this topic. Then it'll be ME that everyone is linking to, showing the difference. :laugh: Like, that's how I give proof... I write reviews.

[LEFT]

Prima.VeraHowever, I always play games with hundreds of stuff opened, like browser window, file mangers, multiple downloads, always on antivirus/malware/idps,etc,etc, all those consuming CPU resources and RAM bandwidth.

[/LEFT]
:clap:

And thrashing the cache as you switch contexts. So when the CPU needs to swap the data out of cache to that new context, memory latency is really felt.

In first moment I thought C9 or C11 when they say Ultra Low Latency.
I don't know for now, but before some time 2400MHz C10 cost more than 3200MHz kit.
Now Dominator Platinum 2133MHz C10 is 369$, 3000MHz C15 is 279.
That mean timmings/latency are very important in comparison than frequency.
No real improvement in performance if latency/timmings are not same after Frequency increase, only marketing.
Better to launch 3200MHz C10 for normal price than every week higher frequency and higher timmings. 3200MHz is great for every platform, standard for X99/X299 and nice secure frequency for Z270/Z370. And people will see real improvement if they timmings drop extremely.
Example 9-10-9 1T 3200Mhz instead 4000MHz.

@cadaveca
It's not my personal opinion, it's a fact and out of all people, I'd expect you to know it since you helped me with RAM overclocking on few occasions.

www.legitreviews.com/ddr4-memory-scaling-intel-z170-finding-the-best-ddr4-memory-kit-speed_170340
www.techpowerup.com/reviews/Intel/Core_i7_8700K_Coffee_Lake_Memory_Performance_Benchmark_Analysis/
www.anandtech.com/show/8959/ddr4-haswell-e-scaling-review-2133-to-3200-with-gskill-corsair-adata-and-crucial

On Intel, it basically doens't matter. If you can get 3000 or 3200MHz RAM on the cheap (at least few months ago you can often find such RAM being just few € more than 2666 or 2800 where from 3200 upwards it was a lot more to add for basically no real benefits), that's basically what you want and ever need. It's a bit different on Ryzen since it's InfinityFabric speed depends on it. But there, it's also a question of stability at high speeds. Paying ridiculous premium on 4000+ kits is only done by someone who has tons of money and just buys best money can buy whether there are any real benefits or not. For normal users, investing the difference into SSD or better graphic card would benefit WAY more than any RAM, even so fast ever will.

RejZoR@cadaveca
It's not my personal opinion, it's a fact and out of all people, I'd expect you to know it since you helped me with RAM overclocking on few occasions.

www.legitreviews.com/ddr4-memory-scaling-intel-z170-finding-the-best-ddr4-memory-kit-speed_170340
www.techpowerup.com/reviews/Intel/Core_i7_8700K_Coffee_Lake_Memory_Performance_Benchmark_Analysis/
www.anandtech.com/show/8959/ddr4-haswell-e-scaling-review-2133-to-3200-with-gskill-corsair-adata-and-crucial

On Intel, it basically doens't matter. If you can get 3000 or 3200MHz RAM on the cheap (at least few months ago you can often find such RAM being just few € more than 2666 or 2800 where from 3200 upwards it was a lot more to add for basically no real benefits), that's basically what you want and ever need. It's a bit different on Ryzen since it's InfinityFabric speed depends on it. But there, it's also a question of stability at high speeds. Paying ridiculous premium on 4000+ kits is only done by someone who has tons of money and just buys best money can buy whether there are any real benefits or not. For normal users, investing the difference into SSD or better graphic card would benefit WAY more than any RAM, even so fast ever will.

Remove the money factor. Not everything has to be based on cost. I don't care about cost, so my opinion differs. Anything CPU-limited benefits from more memory speed, it's just for most, they have highly-overpowered CPUs. In fact, most PC builds are a mess... where I tend to build a perfectly balanced PC which shows these differences. It's not my fault so many people don't know how to build a PC right. You're one of them. My current gaming rig is built with an i5-8400 and 1080's in SLI, and 32 GB of 3600 MHz ram OC'd to 3866 MHz. I run into CPU-limited scenarios, where memory speed really matters.

You also don't seem to understand how easy it is these days to get 4000 MHz. It's going over that where my X299 CPUs, for example (every one of them) needs voltage increases for memory... 4000 MHz is done with stock voltages. So your idea of what runs stable or not is skewed to your own hardware, which you've posted on here with numerous problems. Oh, I wonder why? ROFL.

:lovetpu:

Did you even click the links I kindly provided? There is literally NO difference between speeds. So, why even bother?

Also the "numerous" problems are basically limited to NVIDIA graphic cards. I didn't actually have any other problems with the systems.

RejZoRDid you even click the links I kindly provided? There is literally NO difference between speeds. So, why even bother?

I see FPS increases? And yes, read those. But I also paid attention to the system configurations and not just the results.

[LEFT]

RejZoRAlso the "numerous" problems are basically limited to NVIDIA graphic cards. I didn't actually have any other problems with the systems.

You have problems that no one else does. Those issues you complain about really affects how people look at your posts, and whether they take you seriously or not. I honestly wonder if you do it on purpose to skew ideas in the opposite direction, or if you're simply a standard geek that thinks he knows a lot, but really doesn't. It's probably the latter, but I'll give you the benefit of the doubt and not judge.[/LEFT]

cadaveca:lovetpu:

Meanwhile, I've been recommending people buy 3200 MHz C14 sticks since they went on sale. Like, you are right, but I quoted a poster saying nothing over 2666 mattered, and at times it does.

Well, too much is too much and too less is too less - I think 3200 is about the right clock for optimal performance, as you said yourself.

I've read nearly all those reviews, and I could write a review with hand-picked apps that show differently like it seems most reviews out there do themselves. You do need to remember, I write reviews for TPU. I know exactly how reviews can be twisted to show one thing or another. So in the end, it really does depends on your usage model. I've gone far further in my testing, however, and it's not truly the raw speed that matters; it's the overall latency. But that's the kicker... OVERALL latency. Very few things are bandwidth-constrained, but latency really kicks some apps into the dirt. It's had a lot to do with how hard CPU cache is thrashed, and this is shown in the opposite way by those Intel chips like i7-4950HQ, with 128 MB of cache. With that CPU, ram really does matter less.

Yes I do and don't take me as being disprespectful as I'm not trying to be that. I just have different information/experience so I wanted to answer on that.

That's why it was interesting enough (I think) for W1zz to do a memory analysis on 8700K... does the added cache help? It does. But also, you'll note that W1zz kept the CPU at stock. So what about if the CPU is OC'd? What about if you clock up the cache as well, or lower it?

Well but those reviews I was talking about also showed that (highly) overclocked Intel CPU's didn't need much more, 3200 is still optimal. They showed scaling up to 4000 I think, but it was 99% irrelevant performance wise. 3200 was still the sweetspot / optimum and what the editor recommended in the end.

Then, let's move over to the Ryzen camp... memory speed matter is a huge way, but why? It's because of what's linked to that same speed. Likewise, Intel chips have many different domains running at different speeds, and when those speeds are played with, interesting things happen.

Ryzen is overstated and hyped regarding memory speeds. It's well known by now that after 2933 or even 2666 the speed is barely scaling, 3200 is more than enough - so it's basically the same as with Intel. As on Ryzen refresh and higher clocks - we will see how much Ram speeds they need, but I bet 3200 will still be enough.

When I find the time to write a proper review showing those instances where memory matters, I'll do so, but for me it is very important to explain exactly why it matters, and how, in hopes that we can get more interesting discussions about exactly this topic. Then it'll be ME that everyone is linking to, showing the difference. :laugh: Like, that's how I give proof... I write reviews.

Yes please do that if you can.

That said, I would still not waste money on needlessly overpowered RAM even if I'd be rich, we simply differ in opinion there. You basically like to waste money, I don't. Many rich people say so themselves "you don't get rich by wasting money" - at least the most wise of them do.

JackOne.That said, I would still not waste money on needlessly overpowered RAM even if I'd be rich, we simply differ in opinion there. You basically like to waste money, I don't. Many rich people say so themselves "you don't get rich by wasting money" - at least the most wise of them do.

:toast:

We'd have many great conversations, I think. Much respect to you, sir! But it's not exactly wasting money... well, OK, it kind of is, but part of being wealthy is not being cheap all the time, it's about splurging to support the things you love, so that they can get better, and THAT is not wasteful. ;)

:rockout:

cadaveca:toast:

We'd have many great conversations, I think. Much respect to you, sir! But it's not exactly wasting money... well, OK, it kind of is, but part of being wealthy is not being cheap all the time, it's about splurging to support the things you love, so that they can get better, and THAT is not wasteful. ;)

:rockout:

I do get what you say, some people like to enjoy expensive sports cars, some like to enjoy expensive PC hardware. :toast:

If these have the same PCB as the existing Trident Z RGB series (and even some non-RGB ones), they're going to be a pain to overclock at 1T timings on Asus and Gigabyte boards.

Good PCB layout (used by most of the non-RGB Trident Z sticks, Corsair Vengeance LPX, others):



Bad PCB layout (used by the Trident Z RGB, some non-RGB Trident Z, others):

Can Coffee Lake IMC even run these speeds without the usual binning lottery ?

I wonder why this is considered "ultra-low". Back in DDR2-400 days, fast memory was running CL2 (10ns latency). Fast forward 10 years, RAM is 10x as fast, but so is CL. So in actual real time, the latency is still ~10ns.
I know there are laws of physics and lowering voltage doesn't make capacitors fill up faster. It's the marketing terms I have a problem with.

bugI wonder why this is considered "ultra-low". Back in DDR2-400 days, fast memory was running CL2 (10ns latency). Fast forward 10 years, RAM is 10x as fast, but so is CL. So in actual real time, the latency is still ~10ns.
I know there are laws of physics and lowering voltage doesn't make capacitors fill up faster. It's the marketing terms I have a problem with.

Actually these sticks are 7.9ns not 10ns - (17/4266*2000=7.9ns).

You're also excluding the peak transfer rate in numbers which these sticks can provide while maintaining that lower latency being 34.1GB/a vs DDR2 400 being a mere 3.2GB/s. That's a factor of 10! Let's also add ontop of the fact that these sticks offer a lower operating voltage of 1.45v vs the DDR2's JEDEC being 1.8v.

Now add all the benefits together (Lower voltage, 10x bandwidth, lower latency) and compare it with your DDR2 400. We are certainly leaps and bounds today from what we could do in the past.