Question regarding how CPU and GPU work together

zaibutzu

Junior Member
May 16, 2011
1
0
0
Will GPU be choked by a older gen CPU?

I am currently using 8800GT in SLI with
Intel Core 2 Duo E8500 @ 3.16GHz

I plan on getting ATI HD6950. Am i wasting my money because the card may be ineffective, because i am not using an i5 or i7 processor?

Or does choking of GPU only apply to really slow processors (mine seems pretty capable to me), and not simply to older-gen cards.

Thanks
 

Seero

Golden Member
Nov 4, 2009
1,456
0
0
Will GPU be choked by a older gen CPU?

I am currently using 8800GT in SLI with
Intel Core 2 Duo E8500 @ 3.16GHz

I plan on getting ATI HD6950. Am i wasting my money because the card may be ineffective, because i am not using an i5 or i7 processor?

Or does choking of GPU only apply to really slow processors (mine seems pretty capable to me), and not simply to older-gen cards.

Thanks
Yes
 

apoppin

Lifer
Mar 9, 2000
34,890
1
0
alienbabeltech.com
Will GPU be choked by a older gen CPU?

I am currently using 8800GT in SLI with
Intel Core 2 Duo E8500 @ 3.16GHz

I plan on getting ATI HD6950. Am i wasting my money because the card may be ineffective, because i am not using an i5 or i7 processor?

Or does choking of GPU only apply to really slow processors (mine seems pretty capable to me), and not simply to older-gen cards.

Thanks
No

Not generally unless you play a game that effectively utilizes more than 2 cores although it might help in some games to overclock your CPU a bit further.

The Penryn and Phenom II processors are still very capable for PC gaming and will give you a similar practical gaming experience to i5-i7
 
Last edited:

exar333

Diamond Member
Feb 7, 2004
8,518
8
91
You dualie will be fine for a lot of games, and the new 6950 would be a great GPU. Dual-cores do choke on some games like RTS titles when the number of units/objects gets very high. If you find yourself needing more CPU power, you have a lot of OC room on the E8500, and you always have the option to upgrade your CPU if you find it is really needed for your games.
 

toyota

Lifer
Apr 15, 2001
12,957
1
0
your resolution needs to be known as that hugely determines what level of card to get. a 6850 or 6870 level of card would be more appropriate with an E8500 especially below 1920x1080/1200.
 

Seero

Golden Member
Nov 4, 2009
1,456
0
0
No

Not generally unless you play a game that effectively utilizes more than 2 cores although it might help in some games to overclock your CPU a bit further.

The Penryn and Phenom II processors are still very capable for PC gaming and will give you a similar practical gaming experience to i5-i7
Your statement contradicts itself. If CPU isn't hindering the performance of video cards, then increase the number of cores doesn't help, and increase of frequency does't help. If it helps on some games and OC helps, then it is a yes. But that is actually not the reason why I said yes.

Unless you are still playing doom, the speed of CPU affects a game's visual performance. Lots of visual improvement requires CPU and GPU, not just GPU, and therefore a weak CPU will have a negative impact on visual performance. To top that off, all instructions are sent by CPU to GPU, meaning that both the speed of the CPU and the communication between CPU and video card matters when it comes to gaming. You don't need to play the latest and greatest Dx whatever games at max to see its effect. World of warcraft, Dx9 game, already shows that. Other than MMO, any RTS and FTS are also affected as it isn't about the type of games, but the type of graphics enhancement used.

A single 8800 on I7920@2.66 performs better than on Q6600@3.4. I7 920+8800 sometimes perform better than Q6600 + GTX285. The reason of that is QPI. If FSB is a public transit connected to memory, pci-e, and SB, which then connects to any i/o devices, then QPI is a phone call to pcie and any i/o devices. The data rate of QPI is about 8xFSB plus it is bidirectional, meaning that it can read and write concurrently, while FSB cannot. With the memory controller removed, the traffic of memory io on QPI does not exists, allowing even faster communication as a whole. This architecture allows better utilization of any PCIe connection, including video cards.

In short, video card didn't run faster, but can be utilized more easily on the QPI architecture which cannot be acquired through overclocking the CPU on FSB architecture.
 
Last edited:

darckhart

Senior member
Jul 6, 2004
517
2
81
@seero: good points because i'm sure it's clear to everyone that has kept up with tech that yes, the pairings of cpu and gpu matter.

But i think the op is more concerned about whether he is wasting money on the purchase of this card in his current system. And the best answer so far is, depends on the games you play, the resolution, and what eyecandy you want. There's no question there will be performance improvement. The question is if that improvement will make op feel good after dropping the cash. At the very least, if the 6950 is purchased, eyecandy settings can get cranked way up now.
 

Seero

Golden Member
Nov 4, 2009
1,456
0
0
@seero: good points because i'm sure it's clear to everyone that has kept up with tech that yes, the pairings of cpu and gpu matter.

But i think the op is more concerned about whether he is wasting money on the purchase of this card in his current system. And the best answer so far is, depends on the games you play, the resolution, and what eyecandy you want. There's no question there will be performance improvement. The question is if that improvement will make op feel good after dropping the cash. At the very least, if the 6950 is purchased, eyecandy settings can get cranked way up now.
My personal experience is this. With Q6600@3.2, the system was not able to unleash the full potential of GTX285, which is slower than 6950. If I am to unload about 250 USD on a video card, I will expect fluid gameplay, but OP's CPU may not be able to deliver that. OP will find it a waste of money because it does next to nothing to some games after the purchase, which is the core of OP's question.

In an other way of saying this is, a newer video card may be ineffective on some games with OP's CPU compare to 8800 SLI.
 

apoppin

Lifer
Mar 9, 2000
34,890
1
0
alienbabeltech.com
My personal experience is this. With Q6600@3.2, the system was not able to unleash the full potential of GTX285, which is slower than 6950. If I am to unload about 250 USD on a video card, I will expect fluid gameplay, but OP's CPU may not be able to deliver that. OP will find it a waste of money because it does next to nothing to some games after the purchase, which is the core of OP's question.

In an other way of saying this is, a newer video card may be ineffective on some games with OP's CPU compare to 8800 SLI.

You mention just one personal example - "some games". i test video cards and CPU combinations for AlienBabelTech and i have explored this for 3 years and what you are saying is not *generally* true.

The OP'ss e8500 at his speeds is fine paired with a HD 6950 for 99.99% of PC games. Only a handful of games - mostly released this year - practically benefit from a Quad-core over a dual core.

And Penryn or Phenom II is not much slower than i7 in practically playing PC games. i just proved it again using Phenom II 955 BE versus Core i7 920 with both clocked at 3.7 GHz using a GTX 590. :p
 
Last edited:

Seero

Golden Member
Nov 4, 2009
1,456
0
0
You mention just one personal example - "some games". i test video cards and CPU combinations for AlienBabelTech and i have explored this for 3 years and what you are saying is not *generally* true.

The OP'ss e8500 at his speeds is fine paired with a HD 6950 for 99.99% of PC games. Only a handful of games - mostly released this year - practically benefit from a Quad-core over a dual core.

And Penryn or Phenom II is not much slower than i7 in practically playing PC games. i just proved it again using Phenom II 955 BE versus Core i7 920 with both clocked at 3.7 GHz using a GTX 590. :p
To start off, I have no idea what is with the hostile attitude. I don't recall doing/typing something to you. I have been PLAYING computer games for about 30 years since Apple, thank you very much. I say CPU bottlenecking is "generally" true. We can have our difference in opinions, but that is not the point. The point should be, how we back it up.

You said 99.99%, so you must have tested at least 1000 different games on the same systems with different CPUs. Name 100 of the games you have tested which shows a CPU/architecture upgrade doesn't change performance. To be specific, C2D E8500 vs SB 2600k with HD6950. So far, I showed 1, you showed 0.

BTW, in the gaming world, average forum QQers believe that a game is seriously poor in terms of optimization if it is still single threaded, and FPS really doesn't mean smoother visual experience. Benchmarking with games doesn't always reflect the actual gaming experience any particular setup can bring. Look up at the sky or look at a wall gives you good FPS. I play games, not benchmark video cards with games. I just ditched my Q6600 setup for a new SB setup while keeping the video cards. Guess if I am seeing differences.

Also noted that QPI is more or less an replicate (or improved version) of HyperTransport from AMD, and AMD has been using HyperTransport since 2003. C2D doesn't uses that. Saying that because Phenom II 955 BE matches 920 and therefore E8500 is fine seems to be 2 unrelated statements.

Other than architecture, CoreI7/5 introduced TurboBoost, and SB, which is the 2nd gen of CoreI series, enhanced TurboBoost, allowing better on demand OC. It was designed specifically for programs that doesn't utilizes all cores. E8500 doesn't have that.

I guess, you agree that it is indeed a "yes", but the part where that can be a debate is "by how much". Start a thread in the CPU forum stating that E8500 = SB on gaming or SB doesn't give better gaming experience than the old E8500 and see your feedbacks.
 
Last edited:

apoppin

Lifer
Mar 9, 2000
34,890
1
0
alienbabeltech.com
Hostile attitude? i just don't see it :p

As to being certain, i am.

There are 5 million PC games. i guarantee you that 500 titles do not take advantage of Quad-core practically over dual core. You will have a hard time naming ten titles, nevermind fifty.

http://wiki.answers.com/Q/How_many_video_games_are_there_in_the_world


As i said, i have tested this exhaustively using these CPUs:

  • Phenom II 550-X2
  • Phenom II 720-X3
  • Phenom II 955-X4
  • Phenom II 980-X4

  • Penryn E8600
  • Penryn Q9550S
  • Core i7-920
And i have used a multitude of graphics cards, from GTX 8800 to GTX 590 and from HD 2900 series with E4400 CPU to HD 6990+6790 TriFire-X3

As long as you have a capable CPU - in the above list - you will get a similar playing experience between them. There is practically little difference between actually playing the games as long as you are playing at high details and resolution as gamers do.

i have been charting this for 3 years. Forum rules do not allow me to link to AlienBabelTech reviews myself. i bench over 30 games currently in my regular suite, and only recently - a few of the new games really do seem to benefit from a Quad- over a Dual-core, one of them being Civilization V.

OtOH, you will have a very short list of games for our OP that will play well on a quad and not give good play on his dual


Start a thread in the CPU forum stating that E8500 = SB on gaming or SB doesn't give better gaming experience than the old E8500 and see your feedbacks.
Most of them don't really know better when most of the comparisons are between CPUs at low resolution/low details/no AA
Here's a current thread over there:
http://forums.anandtech.com/showthread.php?t=2163037




To start off, I have no idea what is with the hostile attitude. I don't recall doing/typing something to you. I have been PLAYING computer games for about 30 years since Apple, thank you very much. I say CPU bottlenecking is "generally" true. We can have our difference in opinions, but that is not the point. The point should be, how we back it up.

You said 99.99%, so you must have tested at least 1000 different games on the same systems with different CPUs. Name 100 of the games you have tested which shows a CPU/architecture upgrade doesn't change performance. To be specific, C2D E8500 vs SB 2600k with HD6950. So far, I showed 1, you showed 0.

BTW, in the gaming world, average forum QQers believe that a game is seriously poor in terms of optimization if it is still single threaded, and FPS really doesn't mean smoother visual experience. Benchmarking with games doesn't always reflect the actual gaming experience any particular setup can bring. Look up at the sky or look at a wall gives you good FPS. I play games, not benchmark video cards with games. I just ditched my Q6600 setup for a new SB setup while keeping the video cards. Guess if I am seeing differences.

Also noted that QPI is more or less an replicate (or improved version) of HyperTransport from AMD, and AMD has been using HyperTransport since 2003. C2D doesn't uses that. Saying that because Phenom II 955 BE matches 920 and therefore E8500 is fine seems to be 2 unrelated statements.

Other than architecture, CoreI7/5 introduced TurboBoost, and SB, which is the 2nd gen of CoreI series, enhanced TurboBoost, allowing better on demand OC. It was designed specifically for programs that doesn't utilizes all cores. E8500 doesn't have that.

I guess, you agree that it is indeed a "yes", but the part where that can be a debate is "by how much". Start a thread in the CPU forum stating that E8500 = SB on gaming or SB doesn't give better gaming experience than the old E8500 and see your feedbacks.
 
Last edited:

Seero

Golden Member
Nov 4, 2009
1,456
0
0
Hostile attitude? i just don't see it :p

As to being certain, i am.

There are 5 million PC games. i guarantee you that 500 titles do not take advantage of Quad-core practically over dual core. You will have a hard time naming ten titles, nevermind fifty.

http://wiki.answers.com/Q/How_many_video_games_are_there_in_the_world


As i said, i have tested this exhaustively using these CPUs:

  • Phenom II 550-X2
  • Phenom II 720-X3
  • Phenom II 955-X4
  • Phenom II 980-X4

  • Penryn E8600
  • Penryn Q9550S
  • Core i7-920
And i have used a multitude of graphics cards, from GTX 8800 to GTX 590 and from HD 2900 series with E4400 CPU to HD 6990+6790 TriFire-X3

As long as you have a capable CPU - in the above list - you will get a similar playing experience between them. There is practically little difference between actually playing the games as long as you are playing at high details and resolution as gamers do.

i have been charting this for 3 years. Forum rules do not allow me to link to AlienBabelTech reviews myself. i bench over 30 games currently in my regular suite, and only recently - a few of the new games really do seem to benefit from a Quad- over a Dual-core, one of them being Civilization V.

OtOH, you will have a very short list of games for our OP that will play well on a quad and not give good play on his dual


Most of them don't really know better when most of the comparisons are between CPUs at low resolution/low details/no AA
Here's a current thread over there:
http://forums.anandtech.com/showthread.php?t=2163037
It seems you have omitted all of the points I have made. I never say the number of cores matters, you did. Let me list the things that will make a difference other than the number of cores.

Dynamic OC.
Faster memory accesses due to ondie memory controller.
Faster communication to PCIe due to both QPI and the removal of memory i/o.
More CPU cache.
better clock for clock, SB > Lynnfield, meaning SB > Penryn, by around 10%.

Are you saying that all these won't make any difference in gaming?

I know there are lots of games, I am not challenge you on that. I am challenging you by asking you to name 100 games that you have tested, indicating that choking does not occur on 6950+e8500 by using a Sandy Bridge setup with 6950 showing that there is indeed no improvement. Fail to name 100 games means you simply pick the number 99.99% out of thin air.

My findings can be found by using Anandtech's benching system on WoW if you want quantitative results.
 

apoppin

Lifer
Mar 9, 2000
34,890
1
0
alienbabeltech.com
It seems you have omitted all of the points I have made. I never say the number of cores matters, you did. Let me list the things that will make a difference other than the number of cores.

Dynamic OC.
Faster memory accesses due to ondie memory controller.
Faster communication to PCIe due to both QPI and the removal of memory i/o.
More CPU cache.
better clock for clock, SB > Lynnfield, meaning SB > Penryn, by around 10%.

Are you saying that all these won't make any difference in gaming?

I know there are lots of games, I am not challenge you on that. I am challenging you by asking you to name 100 games that you have tested, indicating that choking does not occur on 6950+e8500 by using a Sandy Bridge setup with 6950 showing that there is indeed no improvement. Fail to name 100 games means you simply pick the number 99.99% out of thin air.

My findings can be found by using Anandtech's benching system on WoW if you want quantitative results.
i did not say*any* difference :p

Look carefully at what i wrote; i am talking about PRACTICAL difference in gaming. e.g. if Core i7-920 gets 150FPS average and Phenom II gets 120FPS with no issues with the minimums - all other things being the same - there is no practical difference playing the game.

i challenge you to find *twenty* games - at gamer's resolutions and settings - where architectural differences in the modern CPU
(SB/i7/Ph II/Penryn) makes a practical difference in gaming. Just find ten.

The gaming experience mostly depends on the graphics card - not the CPU (as long as it is capable). And i have tested my CPU clocks from 2.6 to 4.3GHz.

Here is most of my *current* benchmark suite; over the past three years i have tested many other games to find CPU architectural difference in gaming. The only older game that stands out as favoring Intel is World in Conflict and my newest game, Civ 5.

  • Vantage
  • F.E.A.R.
  • X3:Terran Conflict
  • Enemy Territory: Quake Wars
  • Wolfenstein
  • Unreal Tournament 3
  • Batman: Arkham Asylum
  • Borderlands
  • Call of Duty 4
  • Left4Dead
  • Grand Theft Auto IV
  • Serious Sam, Second Encounter HD (2010)
  • Mafia II
  • Call of Juarez
  • Lost Planet
  • Crysis
  • Far Cry 2
  • Just Cause 2
  • World-in-Conflict
  • H.A.W.X.
  • Resident Evil 5
  • BattleForge
  • Alien vs. Predator
  • Dirt 2
  • STALKER, Call of Pripyat
  • Metro 2033
  • F1 2010
  • H.A.W.X. 2
  • Lost Planet 2
  • Civilization V
  • Heaven 2.5
 
Last edited:

Seero

Golden Member
Nov 4, 2009
1,456
0
0
i did not say*any* difference :p

Look carefully at what i wrote; i am talking about PRACTICAL difference in gaming. e.g. if Core i7-920 gets 150FPS average and Phenom II gets 120FPS with no issues with the minimums - all other things being the same - there is no practical difference playing the game.

i challenge you to find *twenty* games - at gamer's resolutions and settings - where architectural differences in the modern CPU
(SB/i7/Ph II/Penryn) makes a practical difference in gaming. Just find ten.

The gaming experience mostly depends on the graphics card - not the CPU (as long as it is capable). And i have tested my CPU clocks from 2.6 to 4.3GHz.

Here is most of my *current* benchmark suite; over the past three years i have tested many other games to find CPU architectural difference in gaming. The only older game that stands out as favoring Intel is World in Conflict and my newest game, Civ 5.

  • Vantage
  • F.E.A.R.
  • X3:Terran Conflict
  • Enemy Territory: Quake Wars
  • Wolfenstein
  • Unreal Tournament 3
  • Batman: Arkham Asylum
  • Borderlands
  • Call of Duty 4
  • Left4Dead
  • Grand Theft Auto IV
  • Serious Sam, Second Encounter HD (2010)
  • Mafia II
  • Call of Juarez
  • Lost Planet
  • Crysis
  • Far Cry 2
  • Just Cause 2
  • World-in-Conflict
  • H.A.W.X.
  • Resident Evil 5
  • BattleForge
  • Alien vs. Predator
  • Dirt 2
  • STALKER, Call of Pripyat
  • Metro 2033
  • F1 2010
  • H.A.W.X. 2
  • Lost Planet 2
  • Civilization V
  • Heaven 2.5

You began to play with words. The question is simple:
Am i wasting my money because the card may be ineffective, because i am not using an i5 or i7 processor? Going from 8800GTX to HD6950 on E8500. The word "may" consist of any possibility. If I can find 1 example, then I can say yes. If you can test all 5 million games and none showed signs of CPU bottlenecking, then it is a no.

I go on and ask you for 100 example instead of the 99.99% you have claimed, and you came back with a list some games that clearly showed CPU makes a difference. You even stated yourself that World in Conflict and Civ5 are on of those, so what exactly are you saying?

While console ports like Batman was optimized for consoles and C2D alone may be an over kill for the game, replacing video cards won't make a difference as it should be over 60 FPS as it is. Do we really have to dig on games that are over 60FPS on OP's current setup?

Games like GTA4, Far Cry 2 and Mass Effect 2 are clearly CPU intensive, so I have no idea why you even bring them up. WoW, Rift, and EQ2 are MMOs that clearly CPU bottlenecked. Players know killing AF and Shadow will double FPS with ease. Again, while the number of cores may not seems to be the issue, newer architecture does rock, especially in raids and large scale combats.

To take one step in details, things like terrain distance, object distance, particle densities, ground clutter density/radius, and dynamic shadow all requires high degree of CPU power. Clearly, these things are some of the core video settings in many MMOs. Some say MMOs are not graphic intensive and requires lots of CPU power. Well, if the architecture is good, the CPU itself doesn't have to run at 5Ghz to max out graphic cards, although that depends on the number of video cards.

RTS like SC2 is another example where the speed of CPU doesn't seems to matter at first, but becomes one of the most important thing in huge battle. There is a custom map where 3v3 players spawn units sieging each and other, and if it is still a tie after N rounds, each and every single player will spawn 200 units or so, around 2000 units coming out, which is not the lag part. The lag part is once those units start to shot. You need a great PC to have 5 FPS, while most are ... you really need to play it to experience it.

That is an extreme, but CPU bring down FPS is nothing new even on Quad core. Remember, at stock, e8500 is only 3.16Ghz, while SB 2600k starts off at 3.4Ghz and can easily hit 4.4Ghz without even trying. Can a e8500 do that? Those extra hertz become FPS when it matters. Again, looking up at the sky will get you really good FPS, but that is pointless. The FPS during an intensive battles are what counts. Once upon time, on DAoC realm battle, I was walking around the PvP area and out of the sudden my PC was frozen. Well not frozen as there was scenes showing on my screens about 100 or so toons. When the lag goes away, there were nothing but corpses. Almost all of those corpses are due to the lag I experienced. Did I mention Diablo 2 where you die as soon as going into the port while playing hardcore?

10 games yet?
 
Last edited:

apoppin

Lifer
Mar 9, 2000
34,890
1
0
alienbabeltech.com
You began to play with words. The question is simple:
Am i wasting my money because the card may be ineffective, because i am not using an i5 or i7 processor? Going from 8800GTX to HD6950 on E8500. The word "may" consist of any possibility. If I can find 1 example, then I can say yes. If you can test all 5 million games and none showed signs of CPU bottlenecking, then it is a no. ...

i'm sorry you live in a different world than i do and i am done playing word games with you. You can keep your one example - or even ten RTSes - and generalize on the other 5 million games all you want.

Aloha.
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81

Fail. I'd rather have a E8400 stock + GTX 580 than 2500K@4.5 with a 6850. GPU is far more relevant and bottle necking is vastly over rated. Any better GPU will perform significantly better in game than a CPU upgrade.
 

toyota

Lifer
Apr 15, 2001
12,957
1
0
Fail. I'd rather have a E8400 stock + GTX 580 than 2500K@4.5 with a 6850. GPU is far more relevant and bottle necking is vastly over rated. Any better GPU will perform significantly better in game than a CPU upgrade.
for most games yes that will be true. a 2500k, or other fast i5/i7 quad along with that 580 will deliver a much better overall experience when talking about many newer games. but really a smart person would probably go with a gtx570 over the gtx580 with an E8400 unless they dont care about wasting 150 bucks to get the same playable experience.

just something as simple as using vsync is much better when your cpu does not bogging you down and/or causing low minimums.
 
Last edited:

Seero

Golden Member
Nov 4, 2009
1,456
0
0
i'm sorry you live in a different world than i do and i am done playing word games with you. You can keep your one example - or even ten RTSes - and generalize on the other 5 million games all you want.

Aloha.
You asked for 10 games, i gave you 10 games, but you still haven't take my challenge. :biggrin:

Fail. I'd rather have a E8400 stock + GTX 580 than 2500K@4.5 with a 6850. GPU is far more relevant and bottle necking is vastly over rated.
Rather GPU or CPU impacts more is one question. Whether the power of GPU can be hindered by CPU is another. With your example, you will experience far better game play with 2500k. You can choose E8400 + GTX 580, but I will choose 2500K+6850 at a blink of my eye.
 

DirkGently1

Senior member
Mar 31, 2011
904
0
0
My personal experience is this. With Q6600@3.2, the system was not able to unleash the full potential of GTX285, which is slower than 6950. If I am to unload about 250 USD on a video card, I will expect fluid gameplay, but OP's CPU may not be able to deliver that. OP will find it a waste of money because it does next to nothing to some games after the purchase, which is the core of OP's question.

In an other way of saying this is, a newer video card may be ineffective on some games with OP's CPU compare to 8800 SLI.


apoppin is correct with what he is saying. You mention a GTX285 so i'll raise you a GTX580. Name some games, that when run at 1920x1200 with enough eye-candy enabled so that the target FPS is ~60, that would benefit from anything faster than a Q6600@Stock.

Most gamers will induce a GPU limitation by utilising their Graphics cards to, (shock/horror), make the graphics look nice. It's only a fool who buys a faster CPU in the hope of adding 20FPS on top of the 150 that they already achieve.
 

BFG10K

Lifer
Aug 14, 2000
22,709
2,997
126
My personal experience is this. With Q6600@3.2, the system was not able to unleash the full potential of GTX285, which is slower than 6950.
I saw the exact opposite; my GTX285 was easily saturated by an E6850 even when underclocked to 2 GHz at the settings I played my games at. I even wrote an article to prove it.

I have a library of about 130 games under active and heavy play rotation, a mixture of old and new. In practically every game the GPU makes the biggest performance difference at the settings I use. Even Quake 3 shows meaningful performance gains on a GTX580 over a GTX480 (2560x1600 with 32xS) and that game is 12 years old

You also mention QPI/FSB/cache but these are irrelevant if the GPU is the bottleneck. In the vast majority of games at decent settings, the GPU will be the primary bottleneck so you’ll gain the most by upgrading it.

I too would rather have an E8400 + GTX580 vs the fastest CPU + 6850. A 6850 would absolutely cripple the performance levels I expect at the settings I use. I’d have to massively dial back the settings in almost every game I own. Meanwhile an E8400 will show little to no performance difference over my i7 870 in the same situations.
 

Seero

Golden Member
Nov 4, 2009
1,456
0
0
I saw the exact opposite; my GTX285 was easily saturated by an E6850 even when underclocked to 2 GHz at the settings I played my games at. I even wrote an article to prove it.

I have a library of about 130 games under active and heavy play rotation, a mixture of old and new. In practically every game the GPU makes the biggest performance difference at the settings I use. Even Quake 3 shows meaningful performance gains on a GTX580 over a GTX480 (2560x1600 with 32xS) and that game is 12 years old

You also mention QPI/FSB/cache but these are irrelevant if the GPU is the bottleneck. In the vast majority of games at decent settings, the GPU will be the primary bottleneck so you’ll gain the most by upgrading it.

I too would rather have an E8400 + GTX580 vs the fastest CPU + 6850. A 6850 would absolutely cripple the performance levels I expect at the settings I use. I’d have to massively dial back the settings in almost every game I own. Meanwhile an E8400 will show little to no performance difference over my i7 870 in the same situations.
Again, I never said video card upgrade won't help. The question wasn't "Will i see improvement from 8800SLI to 6950 on E8500." The question was "Will GPU be choked by a older gen CPU?"

FPS on benchmarks are nothing but a test to video cards, which can not be used to represent the actual fluidness during gameplay. Yes, a video card upgrade increases average FPS as well as maximum FPS, but rarely help on minimum FPS. By minimum FPS, I mean user tries to get better FPS over better visual qualities because the game is too laggy to play.

Often enough I need to reduce graphic settings even though I got 120 FPS most of the time. I do it because there are occasions where FPS drops to < 20 FPS during battle, and that is what counts. Me and my wife plays in the same room, playing the same game, in the same raid, doing the same encounter. My wife's PC reminds fluid thoughout the encounter with 8800GTS, while mine stutters with 285GTX. Yes, my average FPS is higher, but mine stutters and her pc doesn't. WTF?

Beg my wife to switch pc, got yelled at, but eventually switched. Started the encounter and her is actually running similar FPS as I do, but I saw stutterness, she doesn't. WTF?

Got a friend over, play the game, start a raid, have them play while I stand behind and watch. Looking at 2 pc running on the same game, same encounter, same group and location. Because I am not playing, I finally saw what bothers me.

On her pc, everything moves independently, as it should be. On mine, things appears to move at the same inteval. I saw the pause in between 2 frames. When the character mouse turn, I can clearly see spells that are flying around appears to be teleporting from one spot to another.

To ensure that I am not having the "I see dead people" moment, I asked my friend to see what I see. He didn't see what I see, because we are talking and the encounter haven't start. I took his seat and ask him to simple observe the difference between the 2 screens, he finally sees what I saw. His comment was "her screen is far fiuld than yours." Remember, the 2 pcs is running at similar FPS during encounters, and mine has better FPS when not in combat.

Once I found what I am looking for, I realized that I don't have to be in raids to see that. Unfortunately, my brain is now picking that up every single time it happens and I start to see it at 50 FPS!

After days of researches, I found that the cause of lag isn't due to either gpu or cpu, but network i/o, memory i/o and storage i/o.

I finally understand why some people claim Xeno Nic helps while it doesn't have an impact on FPS. I tried to OC the RAM (around 20&#37;) and found small improvements. I further OC the CPU and found minimal improvement. I brought SSD and found improvement, yes during game, not loading the game. Went out and brought 2 460 and no improvement. Don't get me wrong, FPS does increase, but the lag is still there.

These findings leads to new understanding. I know that Benchmarks are good when it comes to compare video cards, but it isn't the answer to smoother game play. Benchmarks on WoW shows 60+fps on most video card, while many believed that WoW graphic sucks, actual players know that 20FPS is what you get in major towns regardless of the processing power of the video card. Some say that SLI/crossfire doesn't work at all, without knowing that the cockpit is the architectures and i/o bottlenecks.

Knowing exactly what I am looking for, I lowered all OC as I realized that OC has minimal affects.

And now, with the new CPU, which FPS does not really change, as it was 120FPS before, game experience changed, without changing the video cards.

Now do you need SB@5Ghz+arrays of 580s to play console ports? No, although some console ports do implement better graphics for high-end video cards. However, the CPU load have not increases as changing that means a game. Lots of arguments about PhysX and how it doesn't change the game, which are true. In short, better graphics and more eye candy won't change the game.

Yes, there are visual differences by upgrading video cards, but no matter how you upgrade your video card, it doesn't solve the problem cause people to upgrade, laggy gameplay.

People tries to quantitatively define laggy gameplay via FPS count, and people believe that high FPS count means not laggy gameplay. Since video card contribute alot into FPS count, the illusion of high end video cards fix the lagyy gameplay.

Guess what, 120FPS doesn't mean no lag. I will say this again, looking at the sky will give you max FPS. The FPS you get from standing somewhere without action doesn't have anything to do with the fiuldness of the gameplay. In fact, if FPS changing dynamically once battle begins or when the number of players increases, then upgrading video card doesn't help much.

Again, if you are talking about console ports, then there are no need to upgrade as E8500 + a single 8800GTS > any consoles in the market. Yes, there are still lots of games one can play with E8500 + whatever video cards. That doesn't change the fact that GPU like 6950 will be choked by older gen CPU like E8500.

As to all those personal attacks. Hear this, there are people who tries to find games to utilize their investment, and there are people who tries to invest on hardwares for gaming. Some go 3xmonitors to max out their video card, some go 3xmonitors to max out the viewport. There are no right or wrong, but preference. Sometimes people get better hardwares for more eye candies, sometimes people get better hardwares so that they can headshot others instead of being headshotted.