Jump to content

Graphics and framerate - What must I do?


adam123453

Recommended Posts

Then Skyrim could still only use 3.1GB. Assumed you aren't running multiple RAM-intensive programs in the background while playing Skyrim then your System will use 1-2GB RAM for operating. Do the math and you will see that all this huge amount of RAM will bring you ZERO performance plus. ZERO. Not a single frame!

 

But these are facts that are widely discussed here on Nexus and that could be find if somebody just searched for it.

That's not entirely true...(quotes a movie).

 

8.2 GB RAM

15.1 GB System commit

2.9 GB VRAM

8 Threads (4 core)

RAM usage is a bit high here i forgot to close a few programs...lol..but my normal is 6 GB/10 GB....Just Skyrim (system overhead is about 2 GB.)

Me playing Skyrim the way its ment to be played. Heavily modded setup ...239 mods/359 installers. ENB is doing this for me..Enbhost process runs multiple instances. Skyrim is fast and Gorgeous.

PIC

Everyone also says you cant run MSI afterburner and PrecisionX onscreen while playing...look at PIC..simple edit is all it takes.

Link to comment
Share on other sites

Then Skyrim could still only use 3.1GB. Assumed you aren't running multiple RAM-intensive programs in the background while playing Skyrim then your System will use 1-2GB RAM for operating. Do the math and you will see that all this huge amount of RAM will bring you ZERO performance plus. ZERO. Not a single frame!

 

But these are facts that are widely discussed here on Nexus and that could be find if somebody just searched for it.

Or With 64Gb you could run Skyrim from a 50GB RAM drive and still have free ram.

Besides, who buys 64GB of ram for a gaming PC expecting a game to use that much?

I don't know anyone that would. It would be for bragging rights, simply to max out the motherboard RAM, or for non-gaming programs that could use it.

 

I still don't know (and haven't seen a definitive answer here) which would be better--a box with large amounts of system ram(16gb) and a minimal amount of vram (2gb)on a wide bus or moderate amounts of system ram (4-8gb) and a relatively large amount of vram (4Gb) on a narrow bus.

Link to comment
Share on other sites

My local computer shop--manned by hardcore gamers--told me to day that two NVidia 512 mb GeForce 9800 GT's linked in SLI would work just as well as a 4 GB at 256 or even higher. (mind you they can't come up with two 9800's simply because they're not being made anymore but...)

 

And with a slew of texture mods.

 

I don't know...not that tech savvy...just thought I would throw that out there for possible comment.

 

Oh my... There's so many misconceptions in this topic I don't even know where to start :pinch:

 

First of all that local shopkeeper is trying to sell concept he believes in and not what is best for his customer. Don't even think about 512mb vram card in this era if you plan on playing new games in high resolution on high settings. 9800 gt sli would be great for old games (~8 years), but not for this or next gen gaming. In addition Nvidia announced that legacy cards (9800 gt including) support ends this year. What it means is no more driver updates. The old ones may have been polished, but you won't get any performance boost updates for new games in the future (not that you would want to play them on this card).

 

 

Now what you should not do is mix VRAM with Memory Interface Width. VRAM is what caps your resolution and textures size. The higher it is and the more effects you use, the more VRAM your gpu eats. In some cases your game will crash when you run out of it, but most of the time your game starts to chop when you hit the limit or your fps get cut down. Now most modern games will try to adapt to your video memory size and with 1920x1080p / 1920x1200 resolution (on ultra settings) 3gb is sufficient in modern non moded games as of now (this may change in next 2 years). All video cards nowdays are gddr5 type, which means their VRAM will perform similar (note: just vram, not the card itself). I heard rumors that 3gb cards are easier to overclock than 6gb, but that is no concern for average gamer. I would choose 2gb card for gaming on high setting and 3gb for ultra. The only exception is Skyrim, the only game that you can overmod to the point where it needs more than 3gb at 1080p. However even if you won't get restricted by vram, you will still need more computing power than a single card can offer, with such set up. Hence I would stick to 3gb option.

 

I see sometimes people trying to convince other they can substitute GDDR5 VRAM with DDR3 RAM and DDR3 RAM with Hard Drive memory. No, just no.... Any conversion like that cuts down your machine performance tremendously. For instance Pagefile assigns certain hard drive space as ram just so application won't crash when you're for example Alt+Tab'ing to desktop, but the process itself will be slow. I recommend 8gb ram with 64 bit system, for those that don't use professional programs or those that don't use their ram as disk. The windows itself uses ~1.5gb and most games will use from 2 to 3gb. Always buy two identical sticks (like 2x4gb) for dual channel to avoid complication or performance lost. What can help with fps is getting any ram that is faster than 1333mghz, IF you have Hasswell processor (1600mghz CL7, 1800mghz CL8, 2133mghz CL9 or 2400mghz CL10).

 

The biggest impact on gpu performance (when not capped by VRAM) will have it's actual clocks speed, bus width and Cuda Cores amount and architecture (old Fermi cards Cuda are not same as new Kelper gtx Cuda). You could look on the internet for algorithms to see how those specs multiply. The easiest way to compare cards is to check for review sites where they post Memory Bandwith. For example standard gtx 780 ti can output 336000 MB/sec while gtx 760 will 192256 MB/sec (75% less). However it does not mean you will get 75% fps less, it means it's 75% less effective at computing and the difference ingame will depend on application itself. Here is Nvidia benchmark based on Titanfall:

 

 

As you see the difference is about 50%.

 

/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////

Memory Bandwidth: Bandwidth is the largest amount of data (measured in megabytes per second) that can be moved over the external memory interface in one second. It's worked out by multiplying the bus width by its memory clock speed. If it uses DDR type RAM, the result should be multiplied by 2 once again. If DDR5, multiply by 4 instead. The better the card's memory bandwidth, the better the card will be in general. It especially helps with AA, HDR and high resolutions.

 

Texel Rate: Texel rate is the maximum number of texture map elements (texels) that are applied per second. This figure is worked out by multiplying the total amount of texture units of the card by the core clock speed of the chip. The better this number, the better the video card will be at texture filtering (anisotropic filtering - AF). It is measured in millions of texels per second.

 

Pixel Rate: Pixel rate is the maximum number of pixels that the graphics chip could possibly record to the local memory in a second - measured in millions of pixels per second. Pixel rate is worked out by multiplying the number of Raster Operations Pipelines by the the core speed of the card. ROPs (Raster Operations Pipelines - aka Render Output Units) are responsible for filling the screen with pixels (the image). The actual pixel fill rate also depends on lots of other factors, most notably the memory bandwidth - the lower the memory bandwidth is, the lower the ability to reach the maximum fill rate.

////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////

 

The last thing I want to mention is SLI. Two cards in sli mode will duplicate information in their vram and you will not get double vram size of the single gpu. Also there is performance lost with each additional card added to the system. So 2x gpu won't always equal to 200% increase in performance and 3 cards won't output more than 280%. That of course depends on application sli performance, which is kinda bad in old tittles, but it's getting better in newer ones. In worst case scenario you won't see improvement at all if the game was not updated for sli.

 

 

Hope that helps!

Edited by BlackRoseOfThorns
Link to comment
Share on other sites

 

"minimal amount of vram (2gb)on a wide bus or moderate amounts of system ram (4-8gb) and a relatively large amount of vram (4Gb) on a narrow bus."

 

 

GTX Titan Black 6gb GDDR5 vs GTX 780ti 3gb vs Skyrim +150 mods on ENB

 

 

Both have identical bus width (384 Bit), cuda cores (2880) and almost the same clocks (18mghz difference).

Link to comment
Share on other sites

 

My local computer shop--manned by hardcore gamers--told me to day that two NVidia 512 mb GeForce 9800 GT's linked in SLI would work just as well as a 4 GB at 256 or even higher. (mind you they can't come up with two 9800's simply because they're not being made anymore but...)

 

And with a slew of texture mods.

 

I don't know...not that tech savvy...just thought I would throw that out there for possible comment.

 

Oh my... There's so many misconceptions in this topic I don't even know where to start :pinch:

 

First of all that local shopkeeper is trying to sell concept he believes in and not what is best for his customer. Don't even think about 512mb vram card in this era if you plan on playing new games in high resolution on high settings. 9800 gt sli would be great for old games (~8 years), but not for this or next gen gaming. In addition Nvidia announced that legacy cards (9800 gt including) support ends this year. What it means is no more driver updates. The old ones may have been polished, but you won't get any performance boost updates for new games in the future (not that you would want to play them on this card).

 

 

Now what you should not do is mix VRAM with Memory Interface Width. VRAM is what caps your resolution and textures size. The higher it is and the more effects you use, the more VRAM your gpu eats. In some cases your game will crash when you run out of it, but most of the time your game starts to chop when you hit the limit or your fps get cut down. Now most modern games will try to adapt to your video memory size and with 1920x1080p / 1920x1200 resolution (on ultra settings) 3gb is sufficient in modern non moded games as of now (this may change in next 2 years). All video cards nowdays are gddr5 type, which means their VRAM will perform similar (note: just vram, not the card itself). I heard rumors that 3gb cards are easier to overclock than 6gb, but that is no concern for average gamer. I would choose 2gb card for gaming on high setting and 3gb for ultra. The only exception is Skyrim, the only game that you can overmod to the point where it needs more than 3gb at 1080p. However even if you won't get restricted by vram, you will still need more computing power than a single card can offer, with such set up. Hence I would stick to 3gb option.

 

I see sometimes people trying to convince other they can substitute GDDR5 VRAM with DDR3 RAM and DDR3 RAM with Hard Drive memory. No, just no.... Any conversion like that cuts down your machine performance tremendously. For instance Pagefile assigns certain hard drive space as ram just so application won't crash when you're for example Alt+Tab'ing to desktop, but the process itself will be slow. I recommend 8gb ram with 64 bit system, for those that don't use professional programs or those that don't use their ram as disk. The windows itself uses ~1.5gb and most games will use from 2 to 3gb. Always buy two identical sticks (like 2x4gb) for dual channel to avoid complication or performance lost. What can help with fps is getting any ram that is faster than 1333mghz, IF you have Hasswell processor (1600mghz CL7, 1800mghz CL8, 2133mghz CL9 or 2400mghz CL10).

 

The biggest impact on gpu performance (when not capped by VRAM) will have it's actual clocks speed, bus width and Cuda Cores amount and architecture (old Fermi cards Cuda are not same as new Kelper gtx Cuda). You could look on the internet for algorithms to see how those specs multiply. The easiest way to compare cards is to check for review sites where they post Memory Bandwith. For example standard gtx 780 ti can output 336000 MB/sec while gtx 760 will 192256 MB/sec (75% less). However it does not mean you will get 75% fps less, it means it's 75% less effective at computing and the difference ingame will depend on application itself. Here is Nvidia benchmark based on Titanfall:

 

 

As you see the difference is about 50%.

 

/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////

Memory Bandwidth: Bandwidth is the largest amount of data (measured in megabytes per second) that can be moved over the external memory interface in one second. It's worked out by multiplying the bus width by its memory clock speed. If it uses DDR type RAM, the result should be multiplied by 2 once again. If DDR5, multiply by 4 instead. The better the card's memory bandwidth, the better the card will be in general. It especially helps with AA, HDR and high resolutions.

 

Texel Rate: Texel rate is the maximum number of texture map elements (texels) that are applied per second. This figure is worked out by multiplying the total amount of texture units of the card by the core clock speed of the chip. The better this number, the better the video card will be at texture filtering (anisotropic filtering - AF). It is measured in millions of texels per second.

 

Pixel Rate: Pixel rate is the maximum number of pixels that the graphics chip could possibly record to the local memory in a second - measured in millions of pixels per second. Pixel rate is worked out by multiplying the number of Raster Operations Pipelines by the the core speed of the card. ROPs (Raster Operations Pipelines - aka Render Output Units) are responsible for filling the screen with pixels (the image). The actual pixel fill rate also depends on lots of other factors, most notably the memory bandwidth - the lower the memory bandwidth is, the lower the ability to reach the maximum fill rate.

////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////

 

The last thing I want to mention is SLI. Two cards in sli mode will duplicate information in their vram and you will not get double vram size of the single gpu. Also there is performance lost with each additional card added to the system. So 2x gpu won't always equal to 200% increase in performance and 3 cards won't output more than 280%. That of course depends on application sli performance, which is kinda bad in old tittles, but it's getting better in newer ones. In worst case scenario you won't see improvement at all if the game was not updated for sli.

 

 

Hope that helps!

 

Some of that went a bit over my head but I understood most of it. And I thank you for taking the time.

 

So if I may , lets get specific...I have an older motherboard--an Asus P7H55-MLF. That's the delimiter in my quest to upgrade my ssystem...so no Haswell for me.

 

That said I am looking at an i7 quad core cpu to replace the i5 dual core that is in there.

 

And of course 8gb sram (2x4) to replace the 4gb.

 

So the video card is the issue. I need to stay at right around $200.00 for the video card --that's the hard part and I'd welcome suggestions.

 

Looking at a SAPPHIRE 100364-4GL Radeon R9 270X 4GB GDDR5 (is this PCEe?) or a SAPPHIRE 100352-3L Radeon HD 7950 3GB 384-bit GDDR5 PCI Express 3.0 x16 or something like it.

 

However, I'm not sure about the 3gb why it works or doesn't work, and I've heard iffy things about Radeon video cards.

 

Again I would welcome any help here.

Link to comment
Share on other sites

Before upgrading processor I would check prices at your local stores. For example here I can get i7 4770k for almost half of the price of i7 875k (I'm assuming you wanted to buy this one). So for this price I could get newest z97 motherboard and very good processor, with possibility for future upgrades. z97 are very good air clockers (generally go for Z series, if you're going to overclock everything to the max). Check you PSU for Hasswell compability. You may need to restrict yourself to z77 and ivy bridge i7 3770k 1155 LGA socket (almost as good as 4770k).

http://www.cpu-world.com/Compare/545/Intel_Core_i7_i7-4770K_vs_Intel_Core_i7_i7-875K.html

 

For gaming i5 with very high clocks (overclockable series K) will be good alternative where you can cut some costs (i5 4670k / i5 3570k), BUT Enb should take advantage of 4 core i7. If you don't plan on using heavy Enb presets, you will get by with 2 core. Always get K series to get the best performance to money ratio.

 

As for Radeon cards, I'm not a huge fan. They get hot, run loud, but are cheap, so they have this going for them together with couple small features that will not tip the scales over Nvidia drivers and comparability as of now (maybe in the future). Your motherboard is 16x PCIe 2.0 and Radoen is 16x PCIe 3.0. Don't worry 3.0 is backward compatible with 2.0 and there shouldn't be a problem with bandwidth, if you're using one card. The performance should not differ in more then single digit fps.

 

Here is some raw power comparision:

http://www.hwcompare.com/15682/radeon-hd-7950-3gb-vs-radeon-r9-270x/

You might want to check for driver issues, compatibility and benchmarks on dedicated tech sites. I'm not sure which Radeon models will be Mantle api (alternative to DirectX) compatible in future.

 

If you're very tight on money but want 1150 or 1155 LGA socket processor together with new motherboard wait a month or two. There is going to be released new Devils Canyon processor for z97 in June. The prices of old architecture may drop slightly.

 

As far as it goes for 3gb vram vs 4 gb vram it's very simple. You just need to stay under their vram for 100% performance. So if the game has problems with 3gb card, because you play on 2560x1440p resolution and hit minimum 3.6gb then you need to lower your texture resolution/quality, distant detail (LOD), AA, object amount (less different high resolution grass textures for example) to stay within 3gb. Or just run the game on ultra settings at 1080p/720p instead of 1440p and deal with monitor scaling problems (tip: you should run game at your native monitor resolution for best quality).

Edited by BlackRoseOfThorns
Link to comment
Share on other sites

 

And of course 8gb sram (2x4) to replace the 4gb.

 

 

I would forget, sell your 4gb and buy 2x4gb packed together / advertised as dual channel kit. If you mix ram and try to overclock your system, you can get nasty crashes or stability issues. And even if you don't overclock and just have bad luck with mixed ram, your system might not boot >_>

Edited by BlackRoseOfThorns
Link to comment
Share on other sites

GTX650Ti is not even remotely close to being a "bloody decent gear".

You would atleast need a GTX 660Ti. Preferably a GTX 680.

And next time take a look at benchmarks, and make sure your card is getting around 100FPS.

 

Oh my... I just read the other pages, so many false info here... People really belive what sellers say? It's their job to sell... They tell anything you wanna hear, they just want you to buy the damn thing that's all...

And seriously let's get this straight, SLI doesn't get you more VRAM.

512+512 doesn't make 1024.

Edited by Rockmassif
Link to comment
Share on other sites

GTX650Ti is not even remotely close to being a "bloody decent gear".

You would atleast need a GTX 660Ti. Preferably a GTX 680.

And next time take a look at benchmarks, and make sure your card is getting around 100FPS.

 

Oh my... I just read the other pages, so many false info here... People really belive what sellers say? It's their job to sell... They tell anything you wanna hear, they just want you to buy the damn thing that's all...

And seriously let's get this straight, SLI doesn't get you more VRAM.

512+512 doesn't make 1024.

 

When I go shopping all I want to hear from shop staff is if they have what I need in stock. 70% of the time they don't and when they suggest me some other "recommended" model I make "specific" kind of face and they just go away. They have no bloody idea what they are selling, just that this particular product is best they have. You will have to check for every little detail yourself anyway, since they don't even know what cables come with the stuff they sell.

Link to comment
Share on other sites

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...