Regarding the recent Tech Quickie

[Prysin]

9 minutes ago, dalekphalm said:

While I mostly agree with your points, let's be clear - aesthetics do have value - depending on the person. Some enthusiasts value aesthetics over other things, especially when building a color themed build, etc.

 

So saying aesthetics has no value over performance is disingenuous. It may have no value over performance to you, but you cannot make that blanket statement as a fact.

 

If you do so, you're no better than Linus in this situation.

well, that IS a fair argument, and yes i too do look at aestetics, after i look at performance.

However putting looks ahead of performance when, in the case of memory, you can have the same bloody looks at every performance tier, is just well. dumb.

[0ld_Chicken]

I can't believe I missed this thread all week! Damn work has been murder lately.  

 

Last time I saw this discussion come up (with magetank & prysin) I couldn't believe it. I thought it was total BS and there is no way my 4790k and gtx 980 would benefit from anything faster than 1600 ddr3.  I tried oc to 1866 and didn't get much so I sprung 55$ for a used 2400mhz cl 13 kit on LTT's beloved classifieds. Dropped timing to 11 and ran some tests. I'm at work ATM but I'll post my results when I get home. I got about 10% increase in minimum fps and a slight boost in avg fps for GTA v. After I was sure the used RAM kit was working properly, I sold my 1600 kit for 6$ more than I paid for the 2400 kit (both 2x8 Gskill Ripjaws blue) 

 

I don't think I ever benched anything other than GTA v but I might have and still can. Now I'm running sli 980's so I'd be interested to see the results

[DebatED Nothing]

I feel like this forum needs a moderator who will just find posts that are getting a tad too heated and remind everyone that we're talking about computers here and there's literally no reason to get upset by or angered by someone not agreeing with you or being uninformed. It's just a computer, and ultimately it has no bearing on you at all.

[Prysin]

1 minute ago, DebatED Nothing said:

I feel like this forum needs a moderator who will just find posts that are getting a tad too heated and remind everyone that we're talking about computers here and there's literally no reason to get upset by or angered by someone not agreeing with you or being uninformed. It's just a computer, and ultimately it has no bearing on you at all.

i think we need members to realize that Linus has power. A LOT OF POWER over consumers. He has 2.5 million people following him, and many more viewing his video without subscribing. His word has weight behind it.

 

So when he is saying things, that is wrong and not based in facts, then his actions are at the root of misleading and in the case of Techquickie, teach his viewers false information.

This not only hurts the industry, this hurts the buyer.

 

If you knew that by slapping 2133MHz RAM into your i5 2500k, you could potentially keep on going for another 6-12 months before needing to upgrade, you could have improved your economy. Say you live here in Norway, we got on average 3% interest on savings accounts these days (given the current economic setting). If you were to upgrade from a 2500k to a 6600k today (reasonable step), then the cost of that would be roughly 5000 NOK, or 601 USD (todays exchange rate is 1 USD = 8.3 NOK)...

 

Assume 12 months of savings, that would have increased to 625 USD the year after.

Since ALL HARDWARE drops in price over time, imagine what you can get for 625 USD a year later.

 

So what is the cost of swapping to 2133MHz DDR3?

Assuming you are going with 8GBs of DDR3, or 2x4GB... that is 35-50 USD depending on what aestetics you want.

 

Going with a 6600k would still be faster, but would it be worth it, FPS wise? That comes down to the individual and more importantly, what graphics card you are running. Perhaps all you need for a better gaming experience is a new GPU + faster RAM. A combination that is A LOT cheaper then a full system upgrade.

 

Not to forget the elephant in the room -> Windows 10..... Which is certainly not happy about swapping CPU and motherboard around (ive tried).... As soon as you swap to a "invalidated" setup, you're not considered to be using an activated product. Whilst the installation itself is genuine, you will NOT get updates. Which can be devastating in case there is a Zero Day exploit discovered. Or in the case of a full service pack being released -> you wont get it.

That suddenly adds another 100 USD onto your upgrade budget, well 135 USD here in Norway (hail taxes). This further just drives cost of upgrading up and up and up.

 

This issue however is not tied specifically to Sandy Bridge. It is tied to ALL systems (bar AMD FX and FM2+ products it seems). Meaning as time moves on, systems purchased with slower memory will require an upgrade sooner rather then later.

This incurrs costs on consumers. Costs they could have avoided or atleast pushed back for a while.

 

 

And all this can be directly linked back to the indirect damage that Linuses statements brings with them.

He is teaching people the wrong things, he is encouraging users to spend more, more often, then they have to. And he is doing all this while making money and fame doing so.

 

@LinusTech has crossed several ethical and moral lines with this video. He has thrown aside his responsibility as a public media figure, and gone ahead publicizing a product that is not only factually incorrect, but also detrimental to the PC Gaming consumer base.

 

And that. That is not cool.

[DebatED Nothing]

37 minutes ago, Prysin said:

i think we need members to realize that Linus has power. A LOT OF POWER over consumers. He has 2.5 million people following him, and many more viewing his video without subscribing. His word has weight behind it.

 

@LinusTech has crossed several ethical and moral lines with this video. He has thrown aside his responsibility as a public media figure, and gone ahead publicizing a product that is not only factually incorrect, but also detrimental to the PC Gaming consumer base.

I agree completely, I just don't think it's worth spending a week getting wound up over. Eventually you've got to step back and just chill out a bit.

[Prysin]

Just now, DebatED Nothing said:

I agree completely, I just don't think it's worth spending a week getting wound up over. Eventually you've got to step back and just chill out a bit.

if nobody calls him out. Nothing will change.

If nothing changes, we all lose.

 

[DebatED Nothing]

1 minute ago, Prysin said:

if nobody calls him out. Nothing will change.

If nothing changes, we all lose.

I think you're still over emphasizing the importance of this "fight". There are many more important things to be annoyed about, and it's not always a good idea to be annoyed about stuff.

[Prysin]

37 minutes ago, DebatED Nothing said:

I think you're still over emphasizing the importance of this "fight". There are many more important things to be annoyed about, and it's not always a good idea to be annoyed about stuff.

k.

Mind leaving my thread now?

You are not contributing to this discussion with any insightful or useful information that will further progress discussion. Rather, you are derailing the topic by assuming i have a personal anger management problem when it comes to Linus's content.

 

I strongly advice you to leave this thread before i permanently add you to my ignore list.

[IMPERIUS]

3 hours ago, Prysin said:

snips from many posts

 

lol we must #RevoltLinus

 

/jk

 

or is it /jk?

[0ld_Chicken]

Framerate improvements in GTA V benchmark ultra settings with 4790k and gtx 980 (both overclocked)

MIN-5.07 fps

MAX-14.61 fps

AVG-5.64 fps

 

I ran the benchmark 4 times at 1600mhz and 4 times at 2400mhz using a 2400mhz cl 13 2x8gb Gskill Ripjaws X  (timings set to 11 I believe but it was a while ago)

[BuckGup]

On 3/27/2016 at 5:57 PM, Suika said:

To be fair, the enthusiast segment typically doesn't use lower or middle end processors, so in many situations a hardware enthusiast wouldn't see the difference.

There we go thank you.

[BuckGup]

3 hours ago, Prysin said:

if nobody calls him out. Nothing will change.

If nothing changes, we all lose.

 

I think you just want attention 

[Prysin]

4 minutes ago, SamStrecker said:

There we go thank you.

actually. what do you consider lower or middle end processor?

Personally, i consider an i3 6100 or FX 8350 a middle end processor. However that would also mean a i5 2500k is now a middle end processor.

The benchmarks i did previously in this thread, was with a simulated i5 4690k.... would you consider a stock 4690k "middle end" ?

 

3 minutes ago, SamStrecker said:

I think you just want attention 

actually, going by your rather measly contribution to this thread, i'd say YOU are just posting for some attention.

[Coaxialgamer]

6 hours ago, IMPERIUS said:

I think digital foundry clearly showed that ram can impact performance a lot and that faster ram can be worth it.

Ofc not like 100$ more for same capacity like dominator platinums but like getting a 2400mhz stick instead of 2133, that jump isn't expensive but gives you a small boost 

they only showed it can help in cases where the cpu is the bottleneck.

[Prysin]

1 minute ago, Coaxialgamer said:

they only showed it can help in cases where the cpu is the bottleneck.

which is more and more often now a days. With DX12. it will get more and more apparent that a "i5 will be all you need"

[Coaxialgamer]

Just now, Prysin said:

which is more and more often now a days. With DX12. it will get more and more apparent that a "i5 will be all you need"

It is worth mentionning the titan X is not MEANT for 1080p. The cpu bottleneck ( and hence the difference in fps ) will go away once you move to higher resolutions . The same thing would happen if i tied to play at 720p or lower . No matter what i do , there will be cpu  bottleneck . Don't forget that bottlenecks change depending on the resolution.

 

Moreover , we have no way to verify that there wasn't a flaw in DF's testing methodology .

People keep misunderstanding the results of a test and forget they apply to nobody . Nobody pairs an i3 with an overclocked titan x. I am also pretty sure nobody is playing at 1080p with a titan X. 

 

So basically people are using this video as "proof" that faster RAM will get you more frames , but they forget this doesn't apply to ANYONE. 

 

Let's take consider my rig here.

I once broke the pins on the back of my 3770k. I got it working , but dual channel stopped working. This means i went all the way to single channel 1600mhz , effectively dividing by two my memory bandwith . 

This should have a crippling effect on my framerates , right ?

Well no  . In games ( be it new or old titles ) , as well as more synthetic stuff like firestrike and valley , i saw ABSOLUTELY NO DIFFERENCE in FPS.

 

 

[Prysin]

32 minutes ago, Coaxialgamer said:

It is worth mentionning the titan X is not MEANT for 1080p. The cpu bottleneck ( and hence the difference in fps ) will go away once you move to higher resolutions . The same thing would happen if i tied to play at 720p or lower . No matter what i do , there will be cpu  bottleneck . Don't forget that bottlenecks change depending on the resolution.

 

Moreover , we have no way to verify that there wasn't a flaw in DF's testing methodology .

People keep misunderstanding the results of a test and forget they apply to nobody . Nobody pairs an i3 with an overclocked titan x. I am also pretty sure nobody is playing at 1080p with a titan X. 

 

So basically people are using this video as "proof" that faster RAM will get you more frames , but they forget this doesn't apply to ANYONE. 

 

Let's take consider my rig here.

I once broke the pins on the back of my 3770k. I got it working , but dual channel stopped working. This means i went all the way to single channel 1600mhz , effectively dividing by two my memory bandwith . 

This should have a crippling effect on my framerates , right ?

Well no  . In games ( be it new or old titles ) , as well as more synthetic stuff like firestrike and valley , i saw ABSOLUTELY NO DIFFERENCE in FPS.

 

 

3770k is LGA... no pins on the back. The pins are in the socket.

Dual channel means you get a 64bit bus, rather then 128bit bus. Your bandwidth went down, but your throughput speed remained the same. Thus changes may or may not appear.

 

you see, the thing is, the issue is not always bandwidth. and this is where you are starting to go wrong in your assumption.

Bandwidth only applies to the volume of information that can flow, not the speed said information flows at. Just like some software prefers faster CPUs (higher clock speeds) over raw IPC. The data in question isnt complex, thus it is easy to move about. However the data is essential to the bigger picture, and are thus required to be processed fast in order for everything else to be processed. The longer you need to wait for the memory to find, prep and send the data, the longer the instruction queue for your CPU grows.

The reason memory speed helps is not neccessarily bandwidth itself, but the SPEED at which the information flows. The flow back and forth is higher, this results in less waiting time for the CPU. Less waiting = less queueing = more "in real time" rather then working through a backlog.

Sure bandwidth DOES increase.

 

Incidentally, systems such as Hyper-Threading is designed specifically to help remove the impact of memory latency. By having multiple threads pull in information for the cores to process BEFORE the core actually processes anything, you can reduce the amount of memory dependency you would normally see.

By breaking your DIMMs, you should have seen a slight increase in overall CPU workload, however due to only a handful of games being able to make a i7 sweat, you will likely never see a large impact on a i7.

 

An i5 however does NOT feature Hyper-Threading, thus it is a lot more impacted by memory speed, because it must pull the information itself, and a core may then idle for a few microseconds whilst the info is fetched if the memory itself is not responding fast enough.

By not having any fancy systems in order to help it, the i5s (all of them) is purely at the mercy of the memory speed and bandwidth. Incidentally, this is probably why Digital Foundrys i5 2500k testing showed such large improvements. The 2500k is getting a bit long in the tooth, and due to not having hyper-threading, having lower IPC and clock speeds then modern CPUs, it is also struggling in other departments. So it comes as no surprise that a increase in memory speed can have nearly as large an impact as overclocking the CPU does. After all, some workloads are not complex but essential to the bigger picture, and as thus relies on higher clock speeds in order to complete fast.

 

Bandwidth is mostly a problem when you are trying to push enormous amounts of information back and forth, like with an AMD APU. Intel CPUs iGPU "dodges" some of the bandwidth issues by having a larger L3 GPU cache and probably using a form of hyper-threading to assist the iGPU too. However even intel HD iGPUs will see a boost using faster RAM vs slower RAM.

BUT, back to the APU issue, and it really IS an issue, but that is mostly due to HSA. Which stands for Heterogenous System Architecture. What this does, is that it enables the AMD APU to treat system memory both as standard system memory AND VRAM. By doing so, the memory workload increases exponentially, and the size of the data being moved over the system bus is much higher. This means that for AMD APUs, Bandwidth and speed is important. This is why HBM would work excellent for the iGPU, but its slow speed and high latency would cause the CPU to struggle. Similarily, this is why DDR3 is an issue for the iGPU, because it is inherently bandwidth starved, but the memory has decent speed and low latency, which is desired by the CPU.

 

As for synthetic tests.

Synthetic tests are meant to isolate, test and measure. This is why you should look at the detailed breakdown to Firestrike scores. If you had looked at the Physics score which is 100% CPU bound, you would probably have seen a increase or decrease in score, albeit very low (because its not a lot of data being loaded), change. I have personally done this test with my AMD FX 8320, although iv'e later found out that AMDs FX and FM2+ platform may have a system bus or IMC issue which causes memory bandwidth to be radically starved from the get go. I have yet to do thorough testing with my i7 4790k, but i plan to do so eventually.

 

 

Your whole argument is that "memory doesnt matter because i don't notice if my 1600MHz DDR3 is an issue". Yet have you even TRIED to eliminate the possibility by using faster memory?

Have you even tried to overclock your current memory?

 

 

Another issue is the increasing amount of complex physics in games. Physics is mostly done by the CPU, and can be increadibly complex and also be essential to the draw-call process. This means in raw physics work, faster memory should help by a notable amount. I plan to test this with Just Cause 3 which has an extensive usage of physics based effects in the game. I may also test this with The Witcher 3, using hairworks on full setting as this will be very taxing even for my i7 4790k.

[BuckGup]

1 hour ago, Prysin said:

actually. what do you consider lower or middle end processor?

Personally, i consider an i3 6100 or FX 8350 a middle end processor. However that would also mean a i5 2500k is now a middle end processor.

The benchmarks i did previously in this thread, was with a simulated i5 4690k.... would you consider a stock 4690k "middle end" ?

 

actually, going by your rather measly contribution to this thread, i'd say YOU are just posting for some attention.

I didn't make the thread calling Linus out on a small problem that most it doesn't apply to and blow it up into a huge deal

[Prysin]

Just now, SamStrecker said:

I didn't make the thread calling Linus out on a small problem that most it doesn't apply to and blow it up into a huge deal

No you didnt, you also didnt answer my question as to what you consider "low or middle end processor".

 

Considering your signature says you own a i5 4670k, i'd be interested in hearing what you think about the 4690k....

[BuckGup]

5 minutes ago, Prysin said:

No you didnt, you also didnt answer my question as to what you consider "low or middle end processor".

 

Considering your signature says you own a i5 4670k, i'd be interested in hearing what you think about the 4690k....

It is a better cpu. I don't know what you want to hear. It overclocks better.

[Prysin]

2 minutes ago, SamStrecker said:

It is a better cpu. I don't know what you want to hear. It overclocks better.

My original question was:

 

Would you consider the 4690k to be a "low" end, "middle" end or "high" end CPU?

[BuckGup]

Just now, Prysin said:

My original question was:

 

Would you consider the 4690k to be a "low" end, "middle" end or "high" end CPU?

In what time frame. When it first came out or now? It is high middle range but there are things like the 5960X that destroy it

[Coaxialgamer]

Just now, Prysin said:

-snip-

First of all , i did manage to break the pins on the back of my 3770k. If you don't belive me , just look here :

 

 

Also , there are two factors contributing to memory "speed" : bandwith ( directly linked to bus width ( so 64 bit = 128/2 bit  , hence memory bandwith is divided by two ), and latency , measured in cyles and/or nanoseconds  . I typically run my memory at cl7 ). There is no other factor at play here. Yes , the DMI bus can have an impact , but let's stay away from that for now ).Here's a quick rundown of ram types and their respective bandwith .

 

 

Now , i would like you to answer one thing: what exactly makes you certain that DF's video is valid proof ? For all we know , there testing procedure could be flawed . 

And i would still like to mention that they show a situation that never happens. I doubt anyone pairs an i3 with a titan X. Also , even if someone were to pair an i5 with a titan X , they wouldn't be playing at 1080p . Of course the cpu is going to be the bottleneck ( and in that case , faster ram WILL help ). If you listen to the beginning of each of those videos , they point out they are using a titan X at 1080p to create a cpu bottleneck . Same thing happens if you try to game at 720p today . The cpu WILL be the bottleneck .

 

Why does the video matter if it doesn't apply to anyone .

 

If you were to use a typical rig with an i3 or an i5 , along with a 380/960 ( the cards people are ACTUALLY buying , not the titan X ), you will see that there is almost no difference .

 

To sum it up : why should linus care ( or anyone else for that matter ) if faster RAM makes a difference in scenarios that don't apply to anyone ?

[BuckGup]

1 hour ago, Prysin said:

My original question was:

 

Would you consider the 4690k to be a "low" end, "middle" end or "high" end CPU?

i3 and pentium is low i5 and lowest tier i7 are middle and extreme edition i7s and xeons are high. So middle

[Prysin]

1 hour ago, Coaxialgamer said:

snip

you're like a LP record with a massive scratch. Tell me when you are ready to move on so we can discuss something. Not just me sitting here listening to you moaning about one out of three fucking videos and two long ass forum posts/ threads not being to your liking.

 

Go read the OP again. You're just going in circles anyway.

 

@MageTank

See if you can make him understand, cuz i don't have the time or patience to explain this for the n'th time.