WCCFTech RAM affects FPS in FO4....bullshit

S

SomeGuy133

2[H]4U
Joined
Apr 12, 2015
Messages
3,447
http://wccftech.com/fallout-4-performance-heavily-influenced-by-ram-speed-according-to-report/

2 things.
1. No mention of the true latency of the RAM
2. They are using a 970...

I am wondering if the issue is the 970s 3.5 GB of RAM. Is Fallout 4 pushing video memory to the RAM becuase its maxing all 3.5/4 GBs? It is extremely rare for a game to be affected by RAM speeds today and I am call BS on this simply because they are using a 970 and that thing as VRAM issues out the ass.


What are your guy's thoughts?

EDIT: correction...my dumbass missed a paragraph. They supposedly tried this with a Titan X as well to see if it was reproducible and it was.

http://www.eurogamer.net/articles/digitalfoundry-2015-the-best-pc-hardware-for-fallout-4-4023
 
yeah my guess is the rams having an effect due to the engine automatically dumping and reloading textures to system memory constantly while ignoring how much vram the graphic card has available.
 
The site they refer to is digital foundry (http://www.eurogamer.net/articles/digitalfoundry-2015-the-best-pc-hardware-for-fallout-4-4023). I do like their reviews and they are quite detailed in their analysis, however I am skeptical in their theory of being extremely cpu limited and that faster RAM is a bottle neck.

What I did notice is that when god rays are shown in complex areas like through trees/bushes, my frame rate does drop and frame pacing becomes a serious issue [frame rate is still in the high 50's, but the engine is terrible for anything that isn't 30fps or 60fps. I wish more people realized the importance of frame pacing and less on high average fps. [Core i52500k @ 4.0, gtx 980, 16gb DDR3 @ slow 1600mhz)
 
correct me if I'm wrong, if RAM speed actually affects performance, shouldn't the slow RAM speed FPS eventually match up with the fast RAM speed FPS if you stay in a static position or is wccftech experiencing a memory leak?
 
Corsair already proved during the Battlefield 4 days that ram speed affects frame rate to a great degree. If your video card doesn't have a lot of VRAM then it makes sense that it's going to texture swap more. From watching Fallout 4 streams, there seems to be an obscene amount of interactive objects with 3D models. So it makes sense that the engine would continuously flush and swap textures to keep the relevant ones in VRAM. I don't see what's wrong with that.

In short, the better your system build is overall, the better your experience is going to be while gaming. Added this chart from the Corsair article to show that running faster ram speeds is like skipping a generation of video cards. Or getting high end performance from your midgrade video card.

BF41920.png
 
Last edited:
cageymaru said:
Corsair already proved during the Battlefield 4 days that ram speed affects frame rate to a great degree. If your video card doesn't have a lot of VRAM then it makes sense that it's going to texture swap more. From watching Fallout 4 streams, there seems to be an obscene amount of interactive objects with 3D models. So it makes sense that the engine would continuously flush and swap textures to keep the relevant ones in VRAM. I don't see what's wrong with that.

In short, the better your system build is overall, the better your experience is going to be while gaming. Added this chart from the Corsair article to show that running faster ram speeds is like skipping a generation of video cards. Or getting high end performance from your midgrade video card.
BF41920.png
Click to expand...

You forgot tightened timings
 
cageymaru said:
Corsair already proved during the Battlefield 4 days that ram speed affects frame rate to a great degree. If your video card doesn't have a lot of VRAM then it makes sense that it's going to texture swap more. From watching Fallout 4 streams, there seems to be an obscene amount of interactive objects with 3D models. So it makes sense that the engine would continuously flush and swap textures to keep the relevant ones in VRAM. I don't see what's wrong with that.

In short, the better your system build is overall, the better your experience is going to be while gaming. Added this chart from the Corsair article to show that running faster ram speeds is like skipping a generation of video cards. Or getting high end performance from your midgrade video card.

BF41920.png
Click to expand...

but this happened on a Titan as well with 12 GB of VRAM....if its not using all 12GB and dumping stuff into RAM that is a series design flaw.
 
The thing is, back in the way early PC gaming days, memory speed made a huge difference in gaming. Then for awhile there 1600mhz ram was enough and having the 1866 and 2100 speeds didn't do much.

Looks like some games are coming around wanting faster memory speeds.....interesting.

Video memory has nothing to do with System memory when it comes to speed of the System memory, If you look at the BF4 benchmarks (which even I am not sure of since it was release from corsair) it does help.

Looking at the FO4 benchmarks concerning memory speed to me needs more investigation, Maybe [H] can try it out? Shouldnt be that hard to bench.

Either way interesting....been awhile since System Memory speed even helped with PC gaming.
 
DASHlT said:
The thing is, back in the way early PC gaming days, memory speed made a huge difference in gaming. Then for awhile there 1600mhz ram was enough and having the 1866 and 2100 speeds didn't do much.

Looks like some games are coming around wanting faster memory speeds.....interesting.

Video memory has nothing to do with System memory when it comes to speed of the System memory, If you look at the BF4 benchmarks (which even I am not sure of since it was release from corsair) it does help.

Looking at the FO4 benchmarks concerning memory speed to me needs more investigation, Maybe [H] can try it out? Shouldnt be that hard to bench.

Either way interesting....been awhile since System Memory speed even helped with PC gaming.
Click to expand...

the point is if the issue is coming from the swapping of VRAM with memory and you have 12GB of VRAM and only 4GB is filled thats a design flaw.
 
SomeGuy133 said:
the point is if the issue is coming from the swapping of VRAM with memory and you have 12GB of VRAM and only 4GB is filled thats a design flaw.
Click to expand...

why? so you think RAM only works for texture storing?.. also why you think swapping have to be directly between RAM and vRAM?.

in the RAM are stored all the Meshes, Models, Maps, great part of the game itself its stored in the RAM to allow faster load times, faster transition between areas even in the same map, you know what its a bad design flaw? when you have game with little utilization of RAM, because it suffer so badly from popping and constant model dumping, Dragon Age inquisition its a good example of that, that game suffer of BADLY texture popping and models refreshing due the low RAM utilization. you turn fast the camera and plop a lot of popping.. you can see people reappearing just there in front of your eyes, if you are running in a horse you see the game filling texture just in front of you.. and that have nothing to do with vRAM all its RAM management. this is specially bad if you have the game stored in HDD..

Faster RAM and Larger portion of RAM always help with this kind of open word games, and it always favored Skyrim I don't why people take it as something strange with Fallout 4, skyrim used a crap 2GB RAM limit (which was later updated up to 4GB) which caused severe bugs, texture popping, slow texture filling and slow model refresh.. im one of those guys who always say the more RAM and the faster then better if it allow faster load times, greater and longer FOV, greater and longer distant terrain meshes, more and better FPS.. etc.. Im a supporter of this kind of utility for every game possible as this is a truly reason to buy faster RAM, more than just couple of synthetic benchmarks.

as DASHIT said, once RAM were very very important, im happy that's happening again.
 
Yeah, the problem with Fallout is there's so much stuff out there in the world, textures might get completely bumped out by the time you come back to an area, so you have to reload them. This may happen even if you have a 4, 6, or 8GB graphics card!

As far as the bandwidth dependency goes, if you have to reload those textures back into VRAM because YOU ARE IN THE AREA AND MIGHT NEED THEM AGAIN, you probably don't wait until the frame when the texture is needed to request it - you likely prefetch the texture. That's a lot of extra load on main memory bandwidth, as there's a lot of loads that may not be used. And since PCIe 3.0 x16 (bidirectional) is just now starting to exceed the bandwidth of dual-channel DDR3, we're starting to see the benefits of faster ram again.

I've seen this on many recent game benchmarks. I'll be sure to recommend fast ram in any future system builds.
 
Last edited:
Araxie said:
why? so you think RAM only works for texture storing?.. also why you think swapping have to be directly between RAM and vRAM?.

in the RAM are stored all the Meshes, Models, Maps, great part of the game itself its stored in the RAM to allow faster load times, faster transition between areas even in the same map, you know what its a bad design flaw? when you have game with little utilization of RAM, because it suffer so badly from popping and constant model dumping, Dragon Age inquisition its a good example of that, that game suffer of BADLY texture popping and models refreshing due the low RAM utilization. you turn fast the camera and plop a lot of popping.. you can see people reappearing just there in front of your eyes, if you are running in a horse you see the game filling texture just in front of you.. and that have nothing to do with vRAM all its RAM management. this is specially bad if you have the game stored in HDD..

Faster RAM and Larger portion of RAM always help with this kind of open word games, and it always favored Skyrim I don't why people take it as something strange with Fallout 4, skyrim used a crap 2GB RAM limit (which was later updated up to 4GB) which caused severe bugs, texture popping, slow texture filling and slow model refresh.. im one of those guys who always say the more RAM and the faster then better if it allow faster load times, greater and longer FOV, greater and longer distant terrain meshes, more and better FPS.. etc.. Im a supporter of this kind of utility for every game possible as this is a truly reason to buy faster RAM, more than just couple of synthetic benchmarks.

as DASHIT said, once RAM were very very important, im happy that's happening again.
Click to expand...

there example was of a static position and your still missing the point from what someone else said earlier.
 
Wonder how 1866 real tight timing vs 2400 with looser timings would fare. Basically wondering I should upgrade my ram
 
This is why I always try to spend enough money that I am not getting screwed with RAM. Doesn't need to be the fastest but one or two steps down at most.
 
{NG}Fidel said:
This is why I always try to spend enough money that I am not getting screwed with RAM. Doesn't need to be the fastest but one or two steps down at most.
Click to expand...

yea only people that are running into trouble are IB and SB users who have DDR3 from 1066-1600mhz days. Or people who were unlucky and not know yet about RAM and true latency values. Live and learn and all that jazz. I remember not knowing what each value was in DDR 1 days but I lucked out and bought something good :D
 
Yikes IB or SB with that slow RAM would just be a no go for me.
Even my old 1366 is at 2000.
 
Tgrove said:
Wonder how 1866 real tight timing vs 2400 with looser timings would fare. Basically wondering I should upgrade my ram
Click to expand...

Should be similar. You could probably loosen the timings on your RAM and overclock it to 2400 if you wanted to.
 
Man, your telling me those stupid expensive RAM sticks that I bought 5 years ago might actually make a difference now ? Good to know:D.
 
Nahawand said:
Man, your telling me those stupid expensive RAM sticks that I bought 5 years ago might actually make a difference now ? Good to know:D.
Click to expand...

yea but would 16GB make a bigger difference...I don't know how to get by with 8 GB. I have a netbook with 8GB and it gets memory errors all the time because I run out. I haven't had 8GB of RAM ever since the 8GB laptop sticks came out and were no longer 400 bucks each. like 2011 or 2012.
 
SomeGuy133 said:
yea but would 16GB make a bigger difference...I don't know how to get by with 8 GB. I have a netbook with 8GB and it gets memory errors all the time because I run out. I haven't had 8GB of RAM ever since the 8GB laptop sticks came out and were no longer 400 bucks each. like 2011 or 2012.
Click to expand...

To be honest, I don't know what you guys do with your machines but to this date I have never ever hit the 8GB limit on my machine, but all I do is browser the internet watch the occasional movie and play games, like running Fallout 4 with only steam and couple more programs in the background barely breaks the 4GB mark.
 
Nahawand said:
To be honest, I don't know what you guys do with your machines but to this date I have never ever hit the 8GB limit on my machine, but all I do is browser the internet watch the occasional movie and play games, like running Fallout 4 with only steam and couple more programs in the background barely breaks the 4GB mark.
Click to expand...

browser is what kills me. I ussually have 3 windows with like 20 tabs and i hit the limit on the netbook all the time. But I also do other things too but browser eats RAM. If i play a game and have browser open i can get like .5GB from 16GB of RAM. When I had 28GB I would use 16-22GB but thats also windows and programs eating extra because it can but 8GB I would die. 16GB I can suffice but I can't actually use my RAM disk anymore.
 
also if you have plugins running they stay active and suck up a tab worth of memory each

I will say though, TH2 has dropped my utilization by about 15%...
 
Finally got in my 32GB of Crucial Ballistix 1866 in from a Jet order. Put it in and fired up Euro Truck Simulator 2. When I only had 8GB of Corsair DDR3 1866, the game would run in the 60fps range, but halfway through a long delivery the game would develop this annoying stutter like a bad case of VSYNC. It was never terrible, but sometimes it would be distracting enough that I would crash into other vehicles because they just "suddenly appeared out of thin air". It was to the point where I would quit after one delivery.

With the 32GB of ram I get zero stutter, and the game runs like glass from beginning to end. The new DLC areas haven't been optimized much and the frame rate still tanks if you put everything on Ultra in those areas. No magically appearing cars. My minimum and max frame rates increased across the board also. I'm well above 70 most of the time in the older areas that used to run in the mid 60's.

My old Corsair memory was 8GB DDR3 1866 @9-10-9-27
My new Crucial Ballistix is 32GB DDR3 1866 @9-9-9-24

Bought the memory for Battlefront and to build a makeshift PC for my great niece. Might keep it for now and give her my old HP in the living room until Zen or Broadwell-E :)
 
I have three PCs with three different speeds of DDR3 The 1333 Mhz has a CAS latency of 7. One set of 1600 Mhz has a CAS latency of 8. Either one is faster than the last one with a speed of 1600 Mhz and a CAS latency of 9. The 1600 Mhz/ CAS latency 8 is faster than 2133 Mhz with a CAS latency of 11. If a review doesn't list the timings of the various RAM speeds they're testing its useless.
 
Jeffredo said:
I have three PCs with three different speeds of DDR3 The 1333 Mhz has a CAS latency of 7. One set of 1600 Mhz has a CAS latency of 8. Either one is faster than the last one with a speed of 1600 Mhz and a CAS latency of 9. The 1600 Mhz/ CAS latency 8 is faster than 2133 Mhz with a CAS latency of 11. If a review doesn't list the timings of the various RAM speeds they're testing its useless.
Click to expand...

no those are exactly the same

MT 1600
MHz 800
ns 1.25
CL 8
True Latency 10 ns

MT 2133
MHz 1066.5
ns 0.937646507
CL 11
True Latency 10.31411158 ns

Well sure one is .3 ns slower but you won't notice that small of a difference.

I made this Excel DDR true latency calculator for myself so I can shop and compare DDR RAM especially DDR4. Feel free and download it and use it for yourself

https://www.dropbox.com/s/f4kn6qp93576mhe/DDR RAM True Latency Calculator.xlsx?dl=0
 
cageymaru said:
Corsair already proved during the Battlefield 4 days that ram speed affects frame rate to a great degree. If your video card doesn't have a lot of VRAM then it makes sense that it's going to texture swap more. From watching Fallout 4 streams, there seems to be an obscene amount of interactive objects with 3D models. So it makes sense that the engine would continuously flush and swap textures to keep the relevant ones in VRAM. I don't see what's wrong with that.

In short, the better your system build is overall, the better your experience is going to be while gaming. Added this chart from the Corsair article to show that running faster ram speeds is like skipping a generation of video cards. Or getting high end performance from your midgrade video card.

BF41920.png
Click to expand...

You know, this thread just might have inspired me to buy some ram to keep me occupied until Pascal's get released. My SB 2500K is a monster overclocker at 4.7ghz that keeps on chugging so I have no hurry to change it anytime soon, but my information about RAM speed comes from the days the CPU was originally released which said that going above 1600Mhz is useless. It seems like this is no longer the case plus 8GB is starting to get bit small. What do you say, would upgrade from 8GB CL9 1600Mhz DDR3 to 16gb 2133Mhz CL10 make sense? Its only 95€ for a pair of G.Skill Ares sticks. 2400mhz is also buyable but latency increases to CL11 which may be too much plus most sticks are 1.65V at that speed which is not good for the health of SB and Ivy memory controllers in the long run.

I was originally going to finally get an SSD for OS drive but honestly, its not really an upgrade for a 100% gaming machine. It may make windows start faster and make programs feel snappier, which is a huge quality of life improvement but it does not offer any real gaming gains (unless the game is installed on it) which is the stuff that matters the most for this rig.


*edit* Also regarding the timing discussion, I'm under the impression that loose timings + high speed is better for gaming than low speed but tight timings. For rendering stuff and such its other way around, there tight timing work faster than raw clock speed. Is this correct?
 
Last edited:
Buying typically more expensive RAM for one game? No thanks.
 
sonsonate said:
Buying typically more expensive RAM for one game? No thanks.
Click to expand...

If you are referring to above Battlefield 4 test thats no longer the case. Other games use the same engine too, like Dragon Age Inquisition, which is much closer to my interests BTW. Dont give two shits about CoDs and BF's...
 
SomeGuy133 said:
no those are exactly the same

MT 1600
MHz 800
ns 1.25
CL 8
True Latency 10 ns

MT 2133
MHz 1066.5
ns 0.937646507
CL 11
True Latency 10.31411158 ns

Well sure one is .3 ns slower but you won't notice that small of a difference.

I made this Excel DDR true latency calculator for myself so I can shop and compare DDR RAM especially DDR4. Feel free and download it and use it for yourself

https://www.dropbox.com/s/f4kn6qp93576mhe/DDR RAM True Latency Calculator.xlsx?dl=0
Click to expand...

MaZa said:
You know, this thread just might have inspired me to buy some ram to keep me occupied until Pascal's get released. My SB 2500K is a monster overclocker at 4.7ghz that keeps on chugging so I have no hurry to change it anytime soon, but my information about RAM speed comes from the days the CPU was originally released which said that going above 1600Mhz is useless. It seems like this is no longer the case plus 8GB is starting to get bit small. What do you say, would upgrade from 8GB CL9 1600Mhz DDR3 to 16gb 2133Mhz CL10 make sense? Its only 95€ for a pair of G.Skill Ares sticks. 2400mhz is also buyable but latency increases to CL11 which may be too much plus most sticks are 1.65V at that speed which is not good for the health of SB and Ivy memory controllers in the long run.

I was originally going to finally get an SSD for OS drive but honestly, its not really an upgrade for a 100% gaming machine. It may make windows start faster and make programs feel snappier, which is a huge quality of life improvement but it does not offer any real gaming gains (unless the game is installed on it) which is the stuff that matters the most for this rig.


*edit* Also regarding the timing discussion, I'm under the impression that loose timings + high speed is better for gaming than low speed but tight timings. For rendering stuff and such its other way around, there tight timing work faster than raw clock speed. Is this correct?
Click to expand...

download my excel file and edit the numbers. The 2400 has a lower latency then the 2133 or the 1600.

2400 11>2133 10>1600 9

> equal better
 
MaZa said:
*edit* Also regarding the timing discussion, I'm under the impression that loose timings + high speed is better for gaming than low speed but tight timings. For rendering stuff and such its other way around, there tight timing work faster than raw clock speed. Is this correct?
Click to expand...

Loose timings are somewhat irrelevant although often they allow you to push clocks higher. The clockspeed is going to be the driving factor on available bandwidth. For CPU intensive tasks you generally want lower latency, for streaming to a GPU, bandwidth.

When a GPU starts swapping it will typically be large sequential file copies from system ram to video ram. The video card in most cases will have far more bandwidth than the system, which is limited by bus speed for transferring data. So the bus would limit the card to 8GB/s or 16GB/s for newer cards. DDR3-2133 is around 17GB/s per memory channel for reference. So a newer GPU could utilize roughly half your bandwidth for a typical dual channel system if it liked swapping.
 
Anarchist4000 said:
Loose timings are somewhat irrelevant although often they allow you to push clocks higher. The clockspeed is going to be the driving factor on available bandwidth. For CPU intensive tasks you generally want lower latency, for streaming to a GPU, bandwidth.

When a GPU starts swapping it will typically be large sequential file copies from system ram to video ram. The video card in most cases will have far more bandwidth than the system, which is limited by bus speed for transferring data. So the bus would limit the card to 8GB/s or 16GB/s for newer cards. DDR3-2133 is around 17GB/s per memory channel for reference. So a newer GPU could utilize roughly half your bandwidth for a typical dual channel system if it liked swapping.
Click to expand...

timings are only relevant when put into perspective/context as i keep telling people who keep missing the major point that timings is literally only half of the equation.

2133MT/2=1066.5Mhz

(1/(1066.5mhz*10^6))*1000000000=0.937646507ns (cycle time)*10 CL=9.376465073 ns, which is the true latency of the RAM.

The red number is the only number you should care about. CL is meaningless without considering the MTs.

You can have a 4000 CL for all i care as long as the MTs is freakin gigantic.

It is simpler to just use my excel sheet.
 
Last edited:
schizrade said:
Dude, you are gonna give yourself a heart attack...
Click to expand...

why? I just hate ignorance...especially when it is something like this where people make dumb purchases for no reason. It is as bad as buying a CPU or GPU without reading a review.
 
SomeGuy133 said:
why? I just hate ignorance...especially when it is something like this where people make dumb purchases for no reason. It is as bad as buying a CPU or GPU without reading a review.
Click to expand...

You are storming around the board like a maniac. More people will listen to your point if you are not acting like a maniac
 
schizrade said:
You are storming around the board like a maniac. More people will listen to your point if you are not acting like a maniac
Click to expand...

if you say so. Not sure where u get maniac but ok.

I simply provided the excel sheet and the formula so whatever....and made it simple with color codes so people can see each step.
 
You must log in or register to reply here.
Back
Top