To Anandtech and company

[Vesku]

Forget Atom + mid range GPU silliness, are there any solid reviews of Pentium G2020 and i3 3220 with mid range GPUs?

 

[Anubis]

Who in their right mind is going to blow a ton of money on a top of the line GPU whilst spending less than half that amount on a bottom end I5.

That isn't real world it is stupid.

most people who dont post on a tech forum all the time

 

[lakedude]

So basically he wants to strap an atom processor to a 640GT class and bench it.

That is just silly. The OP is suggesting that testing a mid-range card with a high end CPU does not produce real world results. Your comment is too far the other way.

What the OP wants to see (understandably) is a bunch of mid range parts tested together, with not any one part being ridiculously fast or ridiculously slow. He want a realistic assessment of a well balanced mid-range (or perhaps lower mid-range) setup.

 

[frozentundra123456]

That is just silly. The OP is suggesting that testing a mid-range card with a high end CPU does not produce real world results. Your comment is too far the other way.

What the OP wants to see (understandably) is a bunch of mid range parts tested together, with not any one part being ridiculously fast or ridiculously slow. He want a realistic assessment of a well balanced mid-range (or perhaps lower mid-range) setup.

I think you can get a pretty good estimate by trying to determine what the bottleneck will be in your particular system and looking at wide ranging tests of that component.

Hard to explain what I mean, but in a gpu test you can find a wide range of gpus tested with a powerful cpu. In a cpu test you can find a wide range of cpus tested with a powerful gpu. By combining these two I think you can get a fairly good idea of how a mid range system will perform. Granted it would be nice to have it all in a single test, but that is just the way it goes.

The high end gets all the attention.

 

[aigomorla]

That is just silly. The OP is suggesting that testing a mid-range card with a high end CPU does not produce real world results. Your comment is too far the other way.

What the OP wants to see (understandably) is a bunch of mid range parts tested together, with not any one part being ridiculously fast or ridiculously slow. He want a realistic assessment of a well balanced mid-range (or perhaps lower mid-range) setup.

So whats the difference between a 4.5ghz low ball I3 vs a 4.5ghz top of the line K series when both benched on single threaded programs for GPU performance?

So im lost in the test he wants....

The high end gets all the attention.

cuz the high end doesnt have bottle necks... i dont consider an overclocked processor low end either, even tho it was cheap.

Most of these testers have the top of the line processors... (i even had it when i was testing)
Because i think 80% bling to show off u got a high end processor and u look like a professional tester.... and 20% because we usually get them sponsored free, or the tester is a avid overclocker and pays for high end stuff.

The main point is to see if something is mismatched with hardware, and it just doesnt show anything to do the benchmarks on a unstable overclocked system.. or a low end which will run into said bottleknecks.

 

[lakedude]

So im lost in the test he wants....

Well I'm not completely sure either.

Before I go any further note that I understand why the tests are typically done as they are, to eliminate bottlenecks so as to highlight relative performances differences between the hardware being tested. Also testing every possible combination would be prohibitively complex and time consuming.

So the way I took it was that the OP just tossed some hardware out there without all that much thought just as an example. I believe what he is saying is that no one is going to buy a system like any of the combinations shown in this picture:

Picture taken from this review: http://www.anandtech.com/show/6837/...feat-sapphire-the-first-desktop-sea-islands/5

No one is typically going to pair up an HD 7770 with an i7-3960X!

I think what the OP is asking for is a more realistic match, like the above tested video cards with an i5-3570 or even an i3 or FX.

 

[aigomorla]

I think what the OP is asking for is a more realistic match, like the above tested video cards with an i5-3570 or even an i3 or FX.

and everyone is telling u its moot when both processors are overclocked on a single threaded benching applicaiton. :whiste:

unless we cross companies... like AMD vs INTEL.

ur gonna get near the same score on a i3 thats 4.5ghz vs a i5 @ 4.5ghz on the same motherboard with the same GPU.

 

[Vesku]

One of the few times I've seen IB Pentium included:

http://www.xbitlabs.com/articles/cpu/display/fx-8350-8320-6300-4300_6.html#sect0

 

[dave_the_nerd]

No one is typically going to pair up an HD 7770 with an i7-3960X!

I could see doing that, if you were building a home office workstation and maybe wanted to play an occasional round of X-COM or something.

Historically, although not at the moment, I've always had more money sunk into my CPU than my GPU, because the CPU also gets used for all the stuff that isn't gaming.

 

[aigomorla]

I could see doing that, if you were building a home office workstation and maybe wanted to play an occasional round of X-COM or something.

Historically, although not at the moment, I've always had more money sunk into my CPU than my GPU, because the CPU also gets used for all the stuff that isn't gaming.

im the reverse...

cuz 3 gpu's always cost more then 1 cpu. :whiste:

 

[SlowSpyder]

That is like saying you would take our vikings quarterback Christian Ponder at the helm of an NFL team vs Tom Brady quarterbacking a high school team. Which team do you think would win??

Does that make Ponder a better quarterback?? we wish, but that hardly makes it the case.

It is a totally unfair comparison.

I thought it was a perfectly fair comment given the spirit of this thread, what the OP is asking. And yes, your comparrison is right on. I would keep my Thuban and 7970 over a much faster CPU with a much lower end card. One system is a CPU bottleneck, one is a GPU bottleneck, generally speaking. So with what the OP is asking, seeing what system I have, my tendency is to lean towards more GPU power over CPU power. In his comparrison I'd take the low end i5 and GTX680.

All of these upper 4GHz-5GHz Sandys and Ivys are nice, and certainly are amzing pieces of engineering. But most of the time when gaming, at the settings I would play at, I'll take the slower CPU/faster GPU combo. Seeing those kinds of benches would be good, it'd show just how important a fast CPU is or isn't in different games.

 

[ElFenix]

and everyone is telling u its moot when both processors are overclocked on a single threaded benching applicaiton. :whiste:

what's single threaded anymore that people actually use?

i might understand testing only 1 intel quad because (unless you get into S and T processors) they're all within a few hundred mhz of each other and that little difference leads to pretty much 0 out in the real world. but it'd be interesting to see what the difference is between, say, a 3570 with a 670 and a 3220 with a 680

 

[BrightCandle]

Toms hardware still does system builder articles where they put mid range parts together and benchmark that. Now admittedly that doesn't give you the fully range of tests but what many sites did initially was these tests, put out one article showing it makes no practical difference in a load of games and haven't revisited it since.

Since we have quite a few games out today that struggle to get to 60 fps due to CPU limitations it would be nice to see this topic revisited even with just i3 (low end ) v i7 (high end) and 1 mid range and 1 top card. Would certainly be interesting.

But Anandtech is not the place to ask. They about 4 years behind everyone else in graphics performance testing. Try requesting pcper to look into it alongside its impact on frame times, they are quite likely to be interested in doing it.

 

[parvadomus]

I would also like to see how a high-end cpu works with very low end motherboards, if you dont want to OC why care? Like a 8350 on a 760G based mobo, or a 3470 on an intel low end.
Mobos arent tested nowdays, whay happens people??

 

[aigomorla]

I would also like to see how a high-end cpu works with very low end motherboards, if you dont want to OC why care? Like a 8350 on a 760G based mobo, or a 3470 on an intel low end.
Mobos arent tested nowdays, whay happens people??

this makes absolutely no sense.

Low end board wont be able to unlock a high end cpu.


And if ur not caring about Overclocking why are u getting a high end cpu outside enterprise series?
You should be on enterprise series if ur after stability and no overclock.

And in enterprise series, the low end is still considered high end....

 

[myocardia]

I'm more like this, too. I get why low res benches are important, they show what the CPU is capable of, which is handy when benching different CPU's. But I also like, and put more value on as far as what to expect in my real world use, is common higher resolution benches, like 1080P, since I game at 19x12.

For what I do, I'll take my Thuban and 7970 over a 5GHz Ivy with a 7750 level card.

Steve, if CPU benchmarks should be done at 2680x1600, then GPU benchmarks should quite obviously be done @ 1280x1024...

edit: If not @ 1024x768

 

[Enigmoid]

Steve, if CPU benchmarks should be done at 2680x1600, then GPU benchmarks should quite obviously be done @ 1280x1024...

edit: If not @ 1024x768

Look at it this way. If a cpu bench is done at 768x1024 and everything is over 150 fps then you can conclude that logically all the cpus are going to perform the same. There is nothing wrong with performing the test at that resolution. What is wrong is saying that the one that got 250 fps is better than the one with 150 fps because no one cares about 150+ fps. The cpu needs to be good enough to get the game playable (preferably minimums over 60) after which it does not matter. A cpu test is valid at any resolution (but the whole point is not to have the gpu bottleneck in any way) but the results become meaningless if the game is playable and fps is high enough.

(This is what I see a lot with people comparing the 8350 and the 3570. The only time it matters is if we are cpu bottlenecked BELOW smooth. The intel is not 'better' if its over 180 fps while the 8350 is at 140 because even people with a 120 hz monitor aren't going to notice any difference. With a 60 fps monitor this drops down to 60.).

If the game is cpu limited at 768p below playable settings then the 768p bench becomes perfectly valid as no matter how much gpu load you put on it its still going to be cpu limited below nicely playable settings. If benchmarking at 1080p results in a gpu bottleneck that is below playable (ie something like 20 fps average for the highest card) and shows all the cpus performing similarily then we need to throw this benchmark out as no one is going to set the graphics settings up so much that they only get 20 fps. In this case we would aim for the best PLAYABLE settings and check to see if there is a cpu bottleneck below play-ability.

 

[myocardia]

Look at it this way. If a cpu bench is done at 768x1024 and everything is over 150 fps then you can conclude that logically all the cpus are going to perform the same. There is nothing wrong with performing the test at that resolution. What is wrong is saying that the one that got 250 fps is better than the one with 150 fps because no one cares about 150+ fps. The cpu needs to be good enough to get the game playable (preferably minimums over 60) after which it does not matter. A cpu test is valid at any resolution (but the whole point is not to have the gpu bottleneck in any way) but the results become meaningless if the game is playable and fps is high enough.

Which is exactly the reason both you and Steve should stay as far away from any and all CPU benchmarks as you can possibly get, if either of you believe that benchmarking CPU's at any resolution that stresses the GPU in any way is a valid test. You two should obviously be looking at nothing other than GPU benchmarks, at no resolution lower than 7680x1600, with (obviously) 16x MSAA.

 

[Enigmoid]

Which is exactly the reason both you and Steve should stay as far away from any and all CPU benchmarks as you can possibly get, if either of you believe that benchmarking CPU's at any resolution that stresses the GPU in any way is a valid test. You two should obviously be looking at nothing other than GPU benchmarks, at no resolution lower than 7680x1600, with (obviously) 16x MSAA.

???

I'm looking at it from the end use perspective. As long as the cpu is good enough to play the game without problems you are not going to care about how much faster another cpu would be. Unlike a gpu were I can turn up the resolution again and again (in essence with a more powerful gpu I can play at higher resolutions) there is no difference in straight cpu performance.

If you actually read what I said, I said was that the testing is sound but no meaningful results can be obtained if we are looking at really high averages (game is easily playable on both cpus, doesn't matter to the end user if its 500 fps vs 180 fps because they are gaming on a 60 hz or 120 hz screen). If we see a problem at low resolutions that problem will not go away at higher resolutions. I'm simply saying that fps over what the monitor can display is irrelevant. I never said that low resolutions were bad or that high resolutions were required.

The best example would be crysis 3 (strong cpu and gpu bottleneck). Benchmarking at 768p max settings shows that certain cpus fare poorly. If we do highest playable settings using a 680 at 1080p then we will hit a gpu bottleneck at the high end. However the problem then is that certain cpus are obscured. What happens when we use titan SLI? Looking at the 1080p bench we would not know at what point we are hitting a cpu bottleneck, while the 768p probably would (unless the game is extremely gpu heavy and we are still bottlenecked at 768p which is extremely unlikely).

I'm basically saying the exact opposite of SlowSpyder.

 

[disappoint]

Hi guys, I like the website and have been a reader for over 10 years, but I've always wanted to see some real life tests done, except for the best case scenario and no website has still done it.

What I mean is I'd love to see tests done with lower end parts and configurations that aren't optimal, you know like 90% of the people have.

Like I would really love to see a Intel I5 2300 with GTX 680 and 2GB ram tested, or AMD FX6300 with AMD 5570 and 8GB ram.

In other news, Slick here would also like to see how fast a Ferrari 458 Italia can accelerate from 0-60mph with no tires. Just riding on the rims on bare asphalt.

Just because.

In other words, or as some might put it: anotherwords,

Because reasons.

 

[myocardia]

???

I'm looking at it from the end use perspective. As long as the cpu is good enough to play the game without problems you are not going to care about how much faster another cpu would be.

Making you wasting your time reading a CPU benchmark a complete waste of everyone's time. If you don't care which CPU is faster, please, tell me why you'd be looking for CPU benchmarks!

If you actually read what I said, I said was that the testing is sound but no meaningful results can be obtained if we are looking at really high averages (game is easily playable on both cpus, doesn't matter to the end user if its 500 fps vs 180 fps because they are gaming on a 60 hz or 120 hz screen). If we see a problem at low resolutions that problem will not go away at higher resolutions. I'm simply saying that fps over what the monitor can display is irrelevant. I never said that low resolutions were bad or that high resolutions were required.

The best example would be crysis 3 (strong cpu and gpu bottleneck). Benchmarking at 768p max settings shows that certain cpus fare poorly. If we do highest playable settings using a 680 at 1080p then we will hit a gpu bottleneck at the high end. However the problem then is that certain cpus are obscured. What happens when we use titan SLI? Looking at the 1080p bench we would not know at what point we are hitting a cpu bottleneck, while the 768p probably would (unless the game is extremely gpu heavy and we are still bottlenecked at 768p which is extremely unlikely).

I'm basically saying the exact opposite of SlowSpyder.

What you said in your post to which I responded didn't read that way, to me. Since you're the one who wrote it, I'll take your word for it. So, you agree that CPU benchmarks should take the GPU's out of the equation, and GPU benchmarks should do the same for the CPU?

---------------------------------------------------------------------------------------------
To get this back on topic, the answer to the OP's question is as follows. Websites don't write reviews like he's asking to be written for two very important reasons: 1) The people who buy low to mid end CPU's and GPU's are predominantly the same people who would never look on a website for buying information to start with. They mostly rely on the pimple-faced Best Buy employee, for their computer buying advice. 2) Those of us who do in fact know how to use Google, and do so before buying a computer component, would completely ignore a website that tested such low end gear.

 

[Enigmoid]

Making you wasting your time reading a CPU benchmark a complete waste of everyone's time. If you don't care which CPU is faster, please, tell me why you'd be looking for CPU benchmarks!



What you said in your post to which I responded didn't read that way, to me. Since you're the one who wrote it, I'll take your word for it. So, you agree that CPU benchmarks should take the GPU's out of the equation, and GPU benchmarks should do the same for the CPU?

---------------------------------------------------------------------------------------------
To get this back on topic, the answer to the OP's question is as follows. Websites don't write reviews like he's asking to be written for two very important reasons: 1) The people who buy low to mid end CPU's and GPU's are predominantly the same people who would never look on a website for buying information to start with. They mostly rely on the pimple-faced Best Buy employee, for their computer buying advice. 2) Those of us who do in fact know how to use Google, and do so before buying a computer component, would completely ignore a website that tested such low end gear.

No, I never said that either. You are reading too much into this. If I'm looking at a review/benchmark of a new game and I am looking at the cpu section and I see that anything over an old dual core cpu can pull frame frates in excess of 60+ fps minimum, I can conclude that the game is very light on the cpu and that any modern higher clocked quad will perform identically to me (having a 60 hz monitor 60+ fps is useless and impossible to detect). I don't care which cpu is more powerful if either can play the game with no problems. I'm essentially making an argument against those people who say that 768 x 1024 benchmarks are pointless by saying that as long as the framerate is sufficiently low (ie non-optimal) the benchmarks are fine. What is wrong is to claim that the cpu getting 180 fps is somehow better for the person gaming than the cpu getting 130 fps.

What I mean is that this benchmark, showing skyrim shows that both cpus perform the same as seen by the gamer.

On a 60 hz monitor, neither cpu will cause a problem. In essense cpu performance is sufficient and even though the intel cpus perform better that difference is imperceptible without a 120 hz screen.

In contrast to this benchmark

in which the intel cpu clearly performs better. You can notice a difference with 40 fps vs 60 fps.

I agree that cpu benchmarks should take the gpu out of the equation as much as possible and cpu benchmarks should take the gpu out, yes. Benchmarking at a higher resolution is fine though as long as the results are still indicative of performance (should still be 60 fps or higher). To however use a 1080p benchmark that limits the performance to 40 fps because of the gpu limitations should not be done.


The OP has a little bit of a point. Reviews are highly biased toward the high end. I think what he means is how are mid range systems going to perform?

Take something like gw2 for example on a system with a 7770 and an core 2 duo. Someone can look at the cpu benchmarks and conclude that gw2 is really going to die on that cpu. However, gw2 cpu dependance really only occurs when you turn the details up. You can get really high framerates if you turn the settings down (run minimal). However, benchmark websites tend to run the cpu benchmarks at max settings which is good for the top end but ignores the fact that anyone getting 25 fps on their core 2 duo at those settings is probably going to turn something down to get a playable frame rate. Unfortunately these things are not tested at the settings anyone is going to actually run them at.

This is basically a minor argument and given factors such as costs and the relative (much less so) relevance of doing such benchmarks unfeasible.