Does fast gpu make up for a slow cpu.. ??

[trog100]

That's not a problem - that's a good thing! They should use the most powerful/fastest CPU, motherboard, RAM and drives possible.

By using the most powerful CPU available, that removes the CPU from the equation (test results) by ensuring the CPU is not creating a bottleneck and skewing the results.

As I noted above, for many graphics tasks, the results are dependent on how fast the CPU can hand off those tasks to the GPU. If the GPUs are sitting there in wait states until a slow CPU can hand off the next task, the full potentials of the GPUs will never be realized.

You can't determine how much water a 6" drain pipe can drain in one minute if you are feeding water to it from a 1/2" garden hose.

So, if the comparative review is to be considered unbiased and legitimate, they must use the most powerful CPU, fastest and gobs of RAM, a motherboard with the fastest bus speeds, and drives with the fastest access and reads times to properly, fairly, and accurately test and compare any graphics card, from high-end to entry level.

that makes sense when reviewing a graphics card.. what dosnt make sense is what they (to me at least) do when reviewing a cpu..

they always exaggerate the difference by using a very low resolution that would not be used in the real world.. this cpu is better at gaming when it reality at a more normal (higher) resolution sthe difference would not be noticeable..

i am gonna stick my neck out here and say.. the higher the resolution the less difference a cpu makes.. at normal gaming resolutions any half decent cpu will do the job well enough.. assuming the gpu power is there.. if it isnt the fastest cpu on the planet wont help..

trog

 

[Bill_Bright]

i am gonna stick my neck out here and say.. the higher the resolution the less difference a cpu makes..

Then stay away from guillotines then! Because like everything thing else, it depends on the task. It is important to remember that gaming is but one task users use their computers for. And actually, when looking at the big picture, the majority of users are not gamers, or gaming is just a minor pastime function.

 

[trog100]

a pretty dumb answer bill.. forgive me for saying so..

i do know the majority of users are not gamers but cant quite figure out where guillotines come into it.. he he

you need to gen up on the smart assed replies bill that one wasnt very bright.. he he

trog

 

[ZenZimZaliben]

My answer to your question. Is of course a GPU will make up for a slow CPU on GPU intensive games. Try running your slower clocked cpu with few cores on a game like Skylines, Latest Civ game and other CPU intensive games.

 

[Bill_Bright]

Wow Trog. So denigrating others by resorting to personal insults is how you handle it when someone says something different from you? Pretty puerile, no?

My answer was sound, BTW. You made a blanket statement,

"the higher the resolution the less difference a cpu makes"

Fact of the matter is, that was a dumb statement for the exact reason I stated.

but cant quite figure out where guillotines come into it..

Really? Wow! You cannot even follow your own lead about "sticking your neck out" and getting it chopped off with a guillotine??? And you question my brightness?

(Edit comment: fixed quote typo)

 

[trog100]

ahh.. sticking my neck out.. guillotine.. he he..

i take back my comment.. my apologies..

but here is a gaming cpu review done as i think it should be done.. one using real life settings and resolutions and not made up ones to exaggerate the difference which in the real gaming world would not be there..

the guy does take some flack and is accused of cherry picking his games.. always the case when some folks see what they dont want to see. he he

http://www.ocaholic.co.uk/modules/smartsection/item.php?itemid=3948

trog

ps.. this thread is about slower cpus and the effect on gaming bill.. your comment as regards all the users who dont play games was a tad out of context i think.. but lets move on..

the cpu comparison review i linked to is on the front page which is the only reason i read it..

 

[Bill_Bright]

i take back my comment.. my apologies

No problem. Thanks.

the guy does take some flack and is accused of cherry picking his games.. always the case when some folks see what they dont want to see. he he

Yes, when folks see what they don't want to see, but really, it is just always the case. No review can address all scenarios so there has to be some "cherry picking" going on. This is just matter of basic statistics. It is also how every browser is number 1, every new TV show is number 1, and how both AMD and Intel are number 1.

Take polls and surveys. They only take a "sampling" and therefore, always (or the professional ones do anyway) come with a ± percentage "margin of error". The more samples, the more accurate. But unless they can sample 100% of the voters or scenarios, there will always be a margin of error.

ps.. this thread is about slower cpus and the effect on gaming bill..

Going by the your extremely generalized title, "Does fast gpu make up for a slow cpu.. ??" and the fact you just once, briefly, mention just one game as just one example, I would hardly call the subject of this thread exclusive to the effects of slower CPUs on gaming.

This is a technical forum, as you are well aware. If you want specific "contextual" answers, I suggest you pose specific questions limited to a "cherry picked" contextual scenario instead of leaving it up to wide interpretations, expecting everyone to read your mind and automatically know the context of your boredom driven inquiry.

As for your link, yes, it does conclude the performance difference between those two specific CPUs for gaming does not matter. But note it goes on to say,

Buying a Core i7-6700K over a Core i7-6600K makes sense if you do a lot of video encoding or you run other applications which require a lot of processing power from numerous threads.

but lets move on

Good idea.

 

[chuck216]

Seems I opened a can of worms. I apologize.

When I stated that they always use the fastest CPU possible it wasn't a negative point, just saying how come they don't test the same GPUs with multiple processors so the reader (consumer) knows what type of performance a system similar to his would do with a given GPU.

Most of us will never be able to afford a $1000 i7 5960X so knowing an R9 Fury X can hit 44K in Firestrike with one does me little good, I'd rather see how my FX-8320 does with the same card and see if going that extreme of a videocard is worth it for my processor.

A good reviewer would test multiple GPU/CPU combinations to gauge performance for different budgets. in most cases you aren't going to pair a $1000 GPU with a $140 CPU. Most people don't have that kind of money so why only test a $200 ~ $400 GPU with a $1000 i7 5960 when we know the consumer is going to be using at most a low end i7 or I5... or an AMD FX 8 or 9 series.

 

[NdMk2o1o]

Any modern i5/i7 will run any modern games with ease and there's little between them cause 99% of games don't take advantage of HT or more than 4 cores(threads). You seem to be veering off your original OP that review doesn't show how a less capable and slower CPU performs against a high end one, in fact you have a high end top of the range quad and a high end top of the range quad with HT. I'm still waiting for your benches with 2/3 cores disabled and clock speed of <2GHZ

And yes with regards to your comment on him cherry picking his games, none of the games he tested are particularly CPU intensive anyway.

Just look around the forums and you'll see plenty of threads with users who still have 1st gen i7/i5's/ C2Q/Phenom systems where there is a big difference between their systems and similar GPU's who have a very capable i5/i7 etc

 

[Bill_Bright]

Most of us will never be able to afford a $1000 i7 5960X so knowing an R9 Fury X can hit 44K in Firestrike with one does me little good, I'd rather see how my FX-8320 does with the same card and see if going that extreme of a videocard is worth it for my processor.

This is a critical point to remember. Most user cannot afford monster graphics cards either. So game developers code games to have good "game play" on much lessor systems.

 

[chuck216]

This is a critical point to remember. Most user cannot afford monster graphics cards either. So game developers code games to have good "game play" on much lessor systems.

Actually I'm at somewhat of a dilema right now along the same lines I currently have a fairly new R9 380 (it was basically "free" an RMA replacement for a year old R9 270X with a bad fan) but am considering using part of my tax return to upgrade to either an R9 390 or 390X. but the upgrade would only be worth it if I could get significant gains @ 2560X1440 over my current GPU.

 

[EarthDog]

The problem with measuring performance in a synthetic application is that the number of cores matter due to the Physics tests that it runs. It loves cores and clockspeed.

Synthetic benchmarks that do that should have little to no bearing on your GPU choice... look at the GAME performance... or, if you insist on running synthetics, look at the GPU score and FPS only as those should not change with the number of cores.

 

[ZenZimZaliben]

Unigine Heaven benchmark is great for just raw gpu power since cpu isn't utilized that much.

http://www.techpowerup.com/forums/threads/unigine-heaven-4-0-benchmark-scores.198888/

 

[chuck216]

The problem with measuring performance in a synthetic application is that the number of cores matter due to the Physics tests that it runs. It loves cores and clockspeed.

Synthetic benchmarks that do that should have little to no bearing on your GPU choice... look at the GAME performance... or, if you insist on running synthetics, look at the GPU score and FPS only as those should not change with the number of cores.

Number of cores matters in modern gaming too at least if developers are smart. it makes little sense to code a came for a single processor when most CPUs these days have 4 or more cores and/or hyperthreading.

 

[EarthDog]

Not really... have a look at games and see if they scale past 4 cores... while synthetic benchmarks from 3dmark Vantage forward have Physics tests in them, you will unequivocally NOT see the same % gains found in those benchmarks translate to games with a CPU switch. SOME games will do so, but not many at all will gain or come close to the gains you see in those synthetics. You CAN use those if you have the same CPU and its clocked pretty close. It will give you a GENERAL idea. The best method, IMO, is to use games to show differences.

Again, synthetics like that when tested with a hex/octo core from Intel will show larger gains than you will see in gaming and would not be a valid comparison.

 

[trog100]

The problem with measuring performance in a synthetic application is that the number of cores matter due to the Physics tests that it runs. It loves cores and clockspeed.

Synthetic benchmarks that do that should have little to no bearing on your GPU choice... look at the GAME performance... or, if you insist on running synthetics, look at the GPU score and FPS only as those should not change with the number of cores.

purely personal opinion this but i think 3Dmark does a good job showing general gaming performance.. it includes the cpu buts its gpu biased as are most games..

the problem with looking at games is they can and do massively vary.. one needs to look at too many of them.. 3Dmark does pretty much the same thing all in one go.. which is what its intended to do..

plus it give many comparisons and is a universally used tool.. but as i say its purely my own opinion.. like many things its all open to debate..

purely me but the first thing i look at is a systems 3Dmark score.. if that comes up crap i go no further..

trog

ps.. its easy to nip into my bios.. and make any changes i want.. switch off hyper threading.. slow the cpu down.. switch of some of the cores.. my advice is simple.. run your own tests.. i do..

this thread in a way does relate to another thread i started.. how many frames per second are enough for a good gaming experience.. that never got anywhere.. this thread wont ether.. he he..

i have reached my own conclusions.. once you answer the how "many frames per second are enough" question.. anything over that figure is just a waste of time and energy..

 

[Bill_Bright]

but am considering using part of my tax return to upgrade to either an R9 390 or 390X. but the upgrade would only be worth it if I could get significant gains @ 2560X1440 over my current GPU.

Yeah, I would have to assume you would see some gains - whether "significant" or not is another matter. I just don't know. I guess much would depend on what you expect for your money. You certainly don't get twice the performance by adding a second card.

PS - don't forget to make sure your PSU can handle a 2nd card or you may be using up more of that tax return!

 

[EarthDog]

If you have the same CPU as the review, sure you can compare and it would be valid.

But if you compare say, 3DM Fire Strike, with a quad no HT versus an octo core Intel, that result will be woefully off because of those tests using all cores of the CPU.

 

[chuck216]

Yeah, I would have to assume you would see some gains - whether "significant" or not is another matter. I just don't know. I guess much would depend on what you expect for your money. You certainly don't get twice the performance by adding a second card.

PS - don't forget to make sure your PSU can handle a 2nd card or you may be using up more of that tax return!

Actually I was going to replace the 380 with a 390 or 390X and give the 380 to one of my roommates. Not going Crossfire just replacing if I were to go crossfire I could save money by getting a second 380 there is evidence my favorite game (an MMO) doesn't get significant gains from X-fire/SLI setups.

As far as PSU... no problem I've got a platinum rated Corsair 850W

 

[trog100]

If you have the same CPU as the review, sure you can compare and it would be valid.

But if you compare say, 3DM Fire Strike, with a quad no HT versus an octo core Intel, that result will be woefully off because of those tests using all cores of the CPU.

no octo core cpu will make an unplayable game playable but 3Dmark does have to be used with a modicum of common sense.. i have a general idea of what a 3dmark score needs to be for good game play.. to me its not what has the higher score its more about what scores high enough.. my own system scores way more than high enough and to be honest its set up to run in cruising mode..

a bit like the 300 watt rms sound system i have connected to it.. i use the PC equivalent of a volume knob.. some folks on here do seem to have a problem with volume knobs.. maybe its an age thing.. any "youth" i once had is long gone..

i am too old to be playing a game like Just Cause 3 but tis my kind of game.. i recon its brilliant.. he he..

trog

 

[chuck216]

no octo core cpu will make an unplayable game playable but 3Dmark does have to be used with a modicum of common sense.. i have a general idea of what a 3dmark score needs to be for good game play.. to me its not what has the higher score its more about what scores high enough.. my own system scores way more than high enough and to be honest its set up to run in cruising mode..

a bit like the 300 watt rms sound system i have connected to it.. i use the PC equivalent of a volume knob.. some folks on here do seem to have a problem with volume knobs.. maybe its an age thing.. any "youth" i once had is long gone..

trog

Actually more cores can make a difference even in single threaded applications (at least with my AMD it does) because when testing CPU benchmark that are supposed to be "single core/threaded" such as CPU-Z's built in one and watching the resource monitor you can see it's not one core that uses 100% of it's power but rather 3 or more each using between 20~50% with the load varying amongst the cores. It's only when using the multi-thread test that all core hit 100% use.

Also about the "volume knob" most modern CPUs have an "automatic volume level" control that depending on demand can slow or speed up the CPU take for example my 8320 it goes from 1.7 Ghz to 4.0 Ghz depending on load on it's own.

 

[arbiter]

The answer to the question is yes and no, cause it depends on game/app you are using. Some games and app's can use hardware differently. Like how a game like be more cpu dependent and where as others is more about gpu.

 

[Beastie]

My CPU is not the fastest but it is overkill for gaming. It easily keeps up with my GPU in every game (except Supercom, which did use 100% of one core even with my old G210 ).

Any half decent CPU will do an adequate job even with the fastest GPU in most games. CPUs have pretty much overtaken gaming ATM IMO.

 

[Flow]

Well, in reality it wont make a cpu intensive game smoother if the cpu is too weak. You simply need the clocks and/or architecture in that scenario.
If on the other hand such a game would run on a single core 2Ghz cpu, then you could crank up grapical settings with a faster gpu and even gain some fps.

In any case, lowering your current cpu wont paint a true picture, or you would have to really slow it down to a snails pace. We can already see that with overclocked and non overclocked i5's and i7's which in general wont show much difference in fps.
But, if you do manage to slow your cpu down, to a point where your game is loosing fps and starts to stutter, then you should run it with the faster gpu, and see if there really is a difference.
Using an older game for this will help also, many modern games rely more on the gpu and not so much on the cpu. Something the industry already wanted several years ago.

But imho, if your game is struggling to get 40 fps, then a faster gpu wont improve on that. I experienced this already with BF4 where my fps stayed the same, whether on low or high settings.
Mind you, fps was not bad, but a faster video card did not bring extra fps, just extra eye candy.

 

[EarthDog]

no octo core cpu will make an unplayable game playable but 3Dmark does have to be used with a modicum of common sense.. i have a general idea of what a 3dmark score needs to be for good game play.. to me its not what has the higher score its more about what scores high enough.. my own system scores way more than high enough and to be honest its set up to run in cruising mode..

a bit like the 300 watt rms sound system i have connected to it.. i use the PC equivalent of a volume knob.. some folks on here do seem to have a problem with volume knobs.. maybe its an age thing.. any "youth" i once had is long gone..

i am too old to be playing a game like Just Cause 3 but tis my kind of game.. i recon its brilliant.. he he..

trog

sort of...

Here is the thing... it can raise some scores 10s of percentage points above what it 'should score'. So you may have a situation where you think xxxxx score gives you playable or over a certain amount of FPS, but it was due to the overclocked octo instead of 'your' stock quad. The difference between AMD cpus and intel is pretty big as well. There is a reason those who benchmark at hwbot don't use amd for 3d...because it blows in the cpu test killing your overall score.

Again, don't use 3dmarks to 'get an idea' unless it's the same cpu you use.