The Myths Of Graphics Card Performance: Debunked, Part 1

[Jaroslav Jandek]

@photonboy: you are correct (well said), except:

I suspect this mechanism doesn't work in WINDOWED mode.

It works on the principle of data usage: the most used data are allocated in the fastest memory available (VRAM, then RAM, then disk) - the same principle behind Windows virtual memory. In case VRAM is full, a running game will force most of "DWM data" off VRAM (depends on the applications you are running beside your game) - of course, it is still better to run demanding games in full-screen exclusive mode.

 

[Christopher Shaffer]

I'm sorry, but this is just WAY off. If you're only planning to play 5 year old games like Skyrim (which is a GREAT game, nothing against it) then, yes, 1GB is probably fine.

Ummm... Skyrim was released November 2011, so it is just barely over 2 years old... It isn't even halfway to being a "5 year old game" yet.

I may have been exaggerating but its still an older game in terms of graphics, though I think it still looks pretty damn good. Also, it runs on DX9. It was also built with a slightly updated version of the Creation Engine which was used for Fallout 3.

The point was not the age of the game but it's relevance in modern graphics. A game running in DX9 that you can max out and still barely tax a card from a year ago isn't a good comparison in the world of DX11 GPUs, DX11 games, upcoming heavy-graphic titles like Star Citizen, and current titles like BF4, COD:G and Titanfall. Certainly not a good comparison when considering VRAM usage and how you should be choosing your next GPU.

 

[spellbinder2050]

"The second system's main advantage is more overclocking headroom. Our Core i7-4770K sits at a stable 4.6 GHz on air, while the Core i7-950 can't exceed 4 GHz cooled by water."I've been running my Core I7 950 at 4.0 ghz with a Prolimatech Megahalems. My Vcore was randomly chosen and sits at 1.25v with idle temps around 35-43 degrees, and gaming temps that never exceed 56 degrees. All of this was stress tested with Prime95 and Intel burn test of course. I'm pretty sure I could push it even further. Is there something I'm missing, or am I just lucky?

 

[lol13222]

My reaction time per those links is 42ms for sound and 163 for visual ....something is wrong there or I am better the those pilots 😛

 

[smeezekitty]

My reaction time per those links is 42ms for sound and 163 for visual ....something is wrong there or I am better the those pilots 😛

That is crazy quick. I got 300ms for visual and 160 for audio

 

[Ponyface]

dayum, i got around 215 for visual & 150 for audio... no wonder i get my ass kicked in BF4...

 

[Ponyface]

just checked again, my best is 102ms for audio and 195 for visual

 

[nullifier]

Very good article! I especially enjoyed the memory portion, I have been wanting to see something like this for some time.

 

[Mousemonkey]

I would love to see a Tom's article on debunking the 2GB vs 4GB graphic card race. For instance, people spam the Tom's forum daily giving advice to buy the 4GB GTX 770 over the 2GB. Truth is, the 4 GB costs 50$ more and offers NO benefit over the 2GB. Even worse, I see people buying/suggesting the 4GB 760 over a 2GB 770 (which runs only 30$ more and is worth every penny). I am also curious about the 4GB 770 sli scenario. For everything I have seen, even in Sli the 4GB offers no real-world benefit (with the exclusion of MAYBE a few frames per second higher at 3 monitor scenarios, but the rates are unplayable regardless so the gain is negligible). The other myth is that the 4GB 770 is more "future proof". Give me a break. GPU and future proof do not belong in the same sentence. Further, if they were going to be "future proof" they would be "now proof". There are games that are plenty demanding to show the advantage of 2gb vs 4gb - and they simply don't. It's tiring seeing people giving shoddy advice all over the net. I wish a reputable website (Tom's) would settle it once and for all. In my opinion, the extra 2 GB of RAM isn't going to make a tangible difference unless the GPU architecture changes...

Games are currently going through a transitionary period due to the new consoles coming out. Those consoles have an awesome 8GB of shared RAM (about 5GB used for games), which developers are going to start using more and more. Games are already pushing the limits with settings turned up (BF4 gives off loads of VRAM problems with using a 2GB card on the Ultra preset), so as more and more demanding games come out with these consoles, 2GB just ain't gonna kick it.I'm fine with that because I upgrade my cards as soon as a newer generation comes out, but for people holding onto cards for 2, 3 or 4 years at a time, being told to invest in the extra VRAM when 1440p is slowly becoming the new standard and texture resolutions themselves are going up, isn't that bad of an idea really. But hey, I'll happily answer your questions when you come back in a couple years time asking why you're having massive juddering playing intensive games 😉

V-Sync functions by dropping FPS to 1/2 at a certain level; there's no two ways about it. Adapative Sync (Nvidia) is a little bit more "adaptive" but still has similar issues.

Yes, V-Sync halves framerate if there are no more frames in the queue on scan. The issue is with "lag spikes" where you temporarily get below 60 fps (stuttering) - not when you are constantly below 60 fps (you may just as well disable V-Sync altogether), which is solved by the methods I have described.
Adaptive V-Sync has issues, yes. But, frame halving simply does not happen with adaptive V-Sync (I get whatever framerate below 60 that the GPU can handle, not just 30).

That said, there is a TON of outdated, incorrect and just plain WRONG information about VRAM. For example, you claim VRAM size doesn't matter, yet point to the (correct) fact that the size of the textures and the number of on-screen textures are affected by VRAM... If your VRAM is not large or fast enough to handle the quality setting you've chosen, your FPS and ultimately, response time, are negatively affected. This is science and math, not opinion.

What he wrote in the article is "The memory capacity a graphics card ships with has no impact on that product's performance, so long as the settings you're using to game with don't consume all of it.", which is pretty much what you say, I don't see your issue? The reason the performance suffers with high resolution or quality settings is because memory is allocated from the system's RAM instead of the GPU's VRAM via virtual memory when you run out of VRAM.

The rest I don't have issue with.
P.S.: it is not always clear who are you responding to (why not use quotes?).

Adapative V-sync works well, but is nothing in comparison to G-Sync (seriously, I'm in love with this. It's like playing a live video instead of a game) and yes, you're correct regarding the 1/2 FPS. However, Adapative V-sync is an Nvidia -only technology, though AMD has provided some similar solutions, though in my experience, they are more laggy. G-sync is literally lag-free, tear free. (Yes, Nvidia, I would love a job promoting G-Sync because it's awesome and I truly believe in it.)

The key in your last paragraph addresses your own concern with my statement about VRAM:

The reason the performance suffers with high resolution or quality settings is because memory is allocated from the system's RAM instead of the GPU's VRAM via virtual memory when you run out of VRAM.

This is only partially true. The on-screen textures need to be in VRAM prior to actually being displayed by the GPU. Only true Unified Architecture will allow what you're describing.

That "suffering performance" will only occur when you don't have enough VRAM to handle the resolution and quality settings, so indeed your statement explains clearly why more VRAM *is* a benefit in the right situations, not why it is not a factor.

The issue I had in his article was the statement that most systems won't benefit from a GPU with more than 1GB of VRAM, and his incorrect description of how cards behave in SLI/Crossfire:

A GeForce GTX 690 with 4 GB, for instance, behaves like two 2 GB cards in SLI. Moreover, when you add a second card to your gaming configuration in CrossFire or SLI, the array's graphics memory doesn't double. Each card still has access only to its own memory.

Two 4GB cards do NOT act like 2, 2GB cards in SLI. They act like one, very fast 4GB card and have access to 4GB of VRAM. The purpose of SLI is to link the cards for communication of which card is handling which frame, etc.

If you own a 1 GB card and a 1080p display, there's probably no need to upgrade right this very moment. A 2 GB card would let you turn on more demanding AA settings in most games though, so consider that a minimum benchmark if you're planning a new purchase and want to enjoy the latest titles at 1920x1080.

As you scale up to 1440p, 1600p, 2160p or multi-monitor configurations, start thinking beyond 2 GB if you also want to use MSAA. Three gigabytes becomes a better target (or multiple 3 GB+ cards in SLI/CrossFire).

I'm sorry, but this is just WAY off. If you're only planning to play 5 year old games like Skyrim (which is a GREAT game, nothing against it) then, yes, 1GB is probably fine.

However, if you're playing modern titles like BF4, ACIV, etc. where Ultra settings and MSAA can push the VRAM usage beyond the 3GB mark (BF4 averages about 2.6GB on my system but has hit as high as 3.2 at times) then a 3 or 4GB card is a great consideration, even at 1080p. I have Afterburner records showing that 64 player games with a lot of action going on with Ultra, 1080p and MSAA x4 have hit 3.2GB, meaning a 3GB card would not suffice, FPS would suffer, and it's relatively close to the real-world available VRAM of about 3.6GB.

His article is simply misleading. Can you play the games with a 1GB card? Yes. Will you be maxing out graphics settings or even have some on Ultra? Maybe, probably not if you want decent FPS.

You can get by with less VRAM but you have to consider this: on a 2GB card you really only have about 1.8GB available at any given time. If BF4 on even near-Ultra settings uses 2GB or more, then having a 3GB or more card makes perfect sense.

Just think about system memory: if your minimum available *system* RAM for a given game was 2GB, you wouldn't just install 2GB, right? Because at least ~1GB would be used by Windows.

This is the misleading thing in system requirements on games: when it says "2GB DDR3" it doesn't mean that's what you need in your system. It means that's how much the GAME needs to run.

2x 690's will behave the same as 4 2GB cards not one very fast 4GB card as your post implies.

 

[Christopher Shaffer]

I would love to see a Tom's article on debunking the 2GB vs 4GB graphic card race. For instance, people spam the Tom's forum daily giving advice to buy the 4GB GTX 770 over the 2GB. Truth is, the 4 GB costs 50$ more and offers NO benefit over the 2GB. Even worse, I see people buying/suggesting the 4GB 760 over a 2GB 770 (which runs only 30$ more and is worth every penny). I am also curious about the 4GB 770 sli scenario. For everything I have seen, even in Sli the 4GB offers no real-world benefit (with the exclusion of MAYBE a few frames per second higher at 3 monitor scenarios, but the rates are unplayable regardless so the gain is negligible). The other myth is that the 4GB 770 is more "future proof". Give me a break. GPU and future proof do not belong in the same sentence. Further, if they were going to be "future proof" they would be "now proof". There are games that are plenty demanding to show the advantage of 2gb vs 4gb - and they simply don't. It's tiring seeing people giving shoddy advice all over the net. I wish a reputable website (Tom's) would settle it once and for all. In my opinion, the extra 2 GB of RAM isn't going to make a tangible difference unless the GPU architecture changes...

Games are currently going through a transitionary period due to the new consoles coming out. Those consoles have an awesome 8GB of shared RAM (about 5GB used for games), which developers are going to start using more and more. Games are already pushing the limits with settings turned up (BF4 gives off loads of VRAM problems with using a 2GB card on the Ultra preset), so as more and more demanding games come out with these consoles, 2GB just ain't gonna kick it.I'm fine with that because I upgrade my cards as soon as a newer generation comes out, but for people holding onto cards for 2, 3 or 4 years at a time, being told to invest in the extra VRAM when 1440p is slowly becoming the new standard and texture resolutions themselves are going up, isn't that bad of an idea really. But hey, I'll happily answer your questions when you come back in a couple years time asking why you're having massive juddering playing intensive games 😉

V-Sync functions by dropping FPS to 1/2 at a certain level; there's no two ways about it. Adapative Sync (Nvidia) is a little bit more "adaptive" but still has similar issues.

Yes, V-Sync halves framerate if there are no more frames in the queue on scan. The issue is with "lag spikes" where you temporarily get below 60 fps (stuttering) - not when you are constantly below 60 fps (you may just as well disable V-Sync altogether), which is solved by the methods I have described.
Adaptive V-Sync has issues, yes. But, frame halving simply does not happen with adaptive V-Sync (I get whatever framerate below 60 that the GPU can handle, not just 30).

That said, there is a TON of outdated, incorrect and just plain WRONG information about VRAM. For example, you claim VRAM size doesn't matter, yet point to the (correct) fact that the size of the textures and the number of on-screen textures are affected by VRAM... If your VRAM is not large or fast enough to handle the quality setting you've chosen, your FPS and ultimately, response time, are negatively affected. This is science and math, not opinion.

What he wrote in the article is "The memory capacity a graphics card ships with has no impact on that product's performance, so long as the settings you're using to game with don't consume all of it.", which is pretty much what you say, I don't see your issue? The reason the performance suffers with high resolution or quality settings is because memory is allocated from the system's RAM instead of the GPU's VRAM via virtual memory when you run out of VRAM.

The rest I don't have issue with.
P.S.: it is not always clear who are you responding to (why not use quotes?).

Adapative V-sync works well, but is nothing in comparison to G-Sync (seriously, I'm in love with this. It's like playing a live video instead of a game) and yes, you're correct regarding the 1/2 FPS. However, Adapative V-sync is an Nvidia -only technology, though AMD has provided some similar solutions, though in my experience, they are more laggy. G-sync is literally lag-free, tear free. (Yes, Nvidia, I would love a job promoting G-Sync because it's awesome and I truly believe in it.)

The key in your last paragraph addresses your own concern with my statement about VRAM:

The reason the performance suffers with high resolution or quality settings is because memory is allocated from the system's RAM instead of the GPU's VRAM via virtual memory when you run out of VRAM.

This is only partially true. The on-screen textures need to be in VRAM prior to actually being displayed by the GPU. Only true Unified Architecture will allow what you're describing.

That "suffering performance" will only occur when you don't have enough VRAM to handle the resolution and quality settings, so indeed your statement explains clearly why more VRAM *is* a benefit in the right situations, not why it is not a factor.

The issue I had in his article was the statement that most systems won't benefit from a GPU with more than 1GB of VRAM, and his incorrect description of how cards behave in SLI/Crossfire:

A GeForce GTX 690 with 4 GB, for instance, behaves like two 2 GB cards in SLI. Moreover, when you add a second card to your gaming configuration in CrossFire or SLI, the array's graphics memory doesn't double. Each card still has access only to its own memory.

Two 4GB cards do NOT act like 2, 2GB cards in SLI. They act like one, very fast 4GB card and have access to 4GB of VRAM. The purpose of SLI is to link the cards for communication of which card is handling which frame, etc.

If you own a 1 GB card and a 1080p display, there's probably no need to upgrade right this very moment. A 2 GB card would let you turn on more demanding AA settings in most games though, so consider that a minimum benchmark if you're planning a new purchase and want to enjoy the latest titles at 1920x1080.

As you scale up to 1440p, 1600p, 2160p or multi-monitor configurations, start thinking beyond 2 GB if you also want to use MSAA. Three gigabytes becomes a better target (or multiple 3 GB+ cards in SLI/CrossFire).

I'm sorry, but this is just WAY off. If you're only planning to play 5 year old games like Skyrim (which is a GREAT game, nothing against it) then, yes, 1GB is probably fine.

However, if you're playing modern titles like BF4, ACIV, etc. where Ultra settings and MSAA can push the VRAM usage beyond the 3GB mark (BF4 averages about 2.6GB on my system but has hit as high as 3.2 at times) then a 3 or 4GB card is a great consideration, even at 1080p. I have Afterburner records showing that 64 player games with a lot of action going on with Ultra, 1080p and MSAA x4 have hit 3.2GB, meaning a 3GB card would not suffice, FPS would suffer, and it's relatively close to the real-world available VRAM of about 3.6GB.

His article is simply misleading. Can you play the games with a 1GB card? Yes. Will you be maxing out graphics settings or even have some on Ultra? Maybe, probably not if you want decent FPS.

You can get by with less VRAM but you have to consider this: on a 2GB card you really only have about 1.8GB available at any given time. If BF4 on even near-Ultra settings uses 2GB or more, then having a 3GB or more card makes perfect sense.

Just think about system memory: if your minimum available *system* RAM for a given game was 2GB, you wouldn't just install 2GB, right? Because at least ~1GB would be used by Windows.

This is the misleading thing in system requirements on games: when it says "2GB DDR3" it doesn't mean that's what you need in your system. It means that's how much the GAME needs to run.

2x 690's will behave the same as 4 2GB cards not one very fast 4GB card as your post implies.

The point wasn't about the number of GPUs, and my post doesn't imply that 2x 2GB cards will = a 4GB card. It states, quite clearly, that 2x 4GB cards will act like 1 4GB card. In the unique case of a 690, which is dual-gpu, it will actually act like 1, very fast 2GPU card with however much ram is found on one card, in this case it's two cards in one, so 2GB.

The point is that his statement in the article is very convoluted. If you have 2 cards with X amount of VRAM each, assuming that they both have equal amount of VRAM, the result will not be 1/2X or 2X, it will just be X.

In the unique situation of a single 690 with 4GB VRAM, each GPU has 2GB dedicated to it's processes, resulting in a 2GB SLI scenario occurring on a single card.

In the case of, say MY setup where I have 2x, single-GPU 770s w/4GB of VRAM, my cards behave like either 1) An 80% faster 770 w/4GB of VRAM (or you could say) 2) Like 2x SLI 770 4GB w/only 4GB of available memory.

Either way, the article is misleading in terms of how VRAM operates in SLI and using a dual-GPU card as an example makes it confusing.

 

[Achoo22]

The image he shows does not represent an integral multiple of screen refresh rate, nor does a single dropped frame logically translate into a halving of frame rate.