1329 replies to this topic

[Ilithi Dragon]

Wow, that's a lot of info on graphics cards... I won't be in the market for anything for a while, but I am book-marking this thread to review when I am, and I think this thread should be pinned as a general forum information source.

[Vulpesveritas]

Added some input on brands. If anyone happens to be able to add to it, please do so. I would like to make this as complete as possible.

[Thorqemada]

Club3D - usually one of the cheapest brands, cards with the CoolStream (double fan) cooler generate annoying loud sounding vibrations.

Edited by Thorqemada, 02 June 2012 - 05:17 PM.

[Lakevren]

Vulpesveritas, on 02 June 2012 - 03:37 PM, said:

They actually did; http://www.techpower...560_2_GB/1.html
In the majority of games, the 560 non-ti performs about the same as a Radeon HD 6850, and only pulls ahead in heavily tessellated titles, i.e. Crysis 2, HAWX, etc. And even when it does, it barely (see less than 5fps) outperforms. At the same time the 6870 has more aftermarket options, and as such you can get higher overclockers or quieter cards for lower prices. Also, price wise the 6870 is slightly cheaper as well.
6870; http://www.newegg.co...N82E16814102948
560; http://www.newegg.co...N82E16814162084

Not quite what I was looking for. TPU doesn't have an up to date comparison with the rest of the cards, especially with the 7000 and 600 series cards, even though it's one of the better price-performance cards.

The anandtech review also shows that it trades blows with the 6870 pretty well. But of course it shows different games.

In terms of the myth that Nvidia drivers are no better than the AMD drivers, yeah... I'm gonna have to disagree with that one, especially since the GTX 590 video was more of a hardware defect (or oversight, likely due to poor quality VRMs) than caused by a driver. That and I think people overestimated the overclocking capabilities of it, as TPU's W1zzard did. As for AMD and Nvidia drivers, each has their own faults, but AMD's is still lacking by comparison, especially in the compatibility department. Not to mention a lack of a custom profiler, and the fact that I need to cherry pick a driver to find a good one constantly. Then you have Nvidia's TDR issue, which, for me, was caused by a lower voltage on the video card. An updated vbios fixed that. Could be that manufacturers are setting the voltage too low. Cherry picking drivers on the Nvidia end still happens, but much less so than AMD's.
However, there was one point where Nvidia's drivers were no better than AMD's, or ATI's rather, but that was due to less of a need to cherry pick a driver for certain games. I think it was during the 9 and 10 catalyst versions.

As for brands, avoid PNY. It's not average customer service. It's below average. Been known for a decade. The quality of their reference video cards aren't bad, but you better hope they don't break.

As for Asus, I heard the trick is to call, not use the online support. But I haven't really used theirs as of yet.

Edited by Lakevren, 02 June 2012 - 10:18 PM.

[Shivus]

There was also that issue of certain nvidia cards back in the 200 era I believe overheating with the computer in sleep mode.

[Vulpesveritas]

Lakevren, on 02 June 2012 - 10:10 PM, said:

Not quite what I was looking for. TPU doesn't have an up to date comparison with the rest of the cards, especially with the 7000 and 600 series cards, even though it's one of the better price-performance cards.

The anandtech review also shows that it trades blows with the 6870 pretty well. But of course it shows different games.

In terms of the myth that Nvidia drivers are no better than the AMD drivers, yeah... I'm gonna have to disagree with that one, especially since the GTX 590 video was more of a hardware defect (or oversight, likely due to poor quality VRMs) than caused by a driver. That and I think people overestimated the overclocking capabilities of it, as TPU's W1zzard did. As for AMD and Nvidia drivers, each has their own faults, but AMD's is still lacking by comparison, especially in the compatibility department. Not to mention a lack of a custom profiler, and the fact that I need to cherry pick a driver to find a good one constantly. Then you have Nvidia's TDR issue, which, for me, was caused by a lower voltage on the video card. An updated vbios fixed that. Could be that manufacturers are setting the voltage too low. Cherry picking drivers on the Nvidia end still happens, but much less so than AMD's.
However, there was one point where Nvidia's drivers were no better than AMD's, or ATI's rather, but that was due to less of a need to cherry pick a driver for certain games. I think it was during the 9 and 10 catalyst versions.

As for brands, avoid PNY. It's not average customer service. It's below average. Been known for a decade. The quality of their reference video cards aren't bad, but you better hope they don't break.

As for Asus, I heard the trick is to call, not use the online support. But I haven't really used theirs as of yet.

The 590 was a driver issue as unlike normally with a GPU where the driver forces the card to shut down if it's overheating, it just kept getting hotter and bang. Beyond that, the majority of games have no real driver issues. The only thing AMD is really lacking on is crossfire drivers, and that is only in a few games on it's own. Where crossfire works it has far better scaling than SLI, however SLI works in more cases. So there is blow trading there. Beyond that, Nvidia release drivers do tend to be better than AMD release drivers on average.
Another thing is that TPU shows a larger number of games versus anandtech. Therefore TPU is likely to have a more accurate assessment due to a larger sample size.

[Psyctooth]

I shall be honest, you've enlightened me on a small number of key things. BTW people pay duteous attention to his section on Power Supplies. Also after market coolers are very important regardless if you overclock or not if you are using 2011socket intel or the latest CPUs from both AMD and Intel because simply keeping the CPUs cool always is always something you should strive for even if it is not overlocked. It keeps it operating at peak efficiency and extends its life.


But as far as communities go, I often found mechwarrior fans, although often we are frowned down on by other nerd and geek communities know computers and what and what not to do. Also keep an eye out for faulty caps in PSUs, Graphics Cards and Screens, although it is a thing of the past in motherboards since we are all very careful when we buy them now, we should also be careful about which graphics cards and PSUs we buy, check the info on the product, there are still some floating around... mainly in screens and PSUs these days.

As to other things, often and too often the case the vast majority of people who are gamers are not tech heads. As counter intuitive to the case of 10+ years ago when you pretty much needed to know your computer as much as you needed to know your software to be able to be a gamer of any creditable level. Case in point.

However, today we have so much variety mainly due to ignorance, companies pump out all sorts of wizz bang swag that it's hard to tell the difference between... crap, what you need, what you need for what you want to do and what is so crazy enthusiastically over the top that even games scatch their heads when faced with the experimental technology which companies for some reason beyound reasonable understanding decide to sell on the open market instead of restrict it to workstations and corporate.

As far as whiz bang OMG fast as bloody hell on a horse CPUs and a bucket full of RAM go, they are mostly used by Game Designers and Media Designers for two purposes, unoptimized game engines running games in the development stage(alpha) or closed beta where they are mostly reliant on the CPU and onboard RAM for processing without making use of snazzy GPU architecture and for creating game worlds where with a lot but not all world builders still operate in an environment which heavily taxes the RAM and then compiles ("bakes") the world into a compressed state (e.g.. WorldBuilder 2, Radiant, Segmental world builders which are even more commonly used which build a world from prefabricated sections in a void with limited sandbox capacity) which only renders visible content in view of the player camera or for the UnrealED Compiles it in real time after each section is included. OR for Media Designers for the purpose of, you guessed it... rendering; Be it will be rendering for 3D graphical content or Virtual Visual Effects.


As far as for gaming you simply need enough RAM and a fast enough CPU as not to bottleneck your GPU and to make sure you can multitask at a comfortable speed without affecting performance. There was a time when there were "no-nos" to what you did. Now days the only "no-nos" which you still need to follow is not to hard-restart or hard-shutdown a computer because you risk damaging your electronics, CPU and GPU (the later mainly due to the risk of having a defective driver or bios).

Edited by Psyctooth, 05 June 2012 - 07:47 AM.

[cipher]

Psyctooth, on 05 June 2012 - 07:25 AM, said:

As far as whiz bang OMG fast as bloody hell on a horse CPUs and a bucket full of RAM go, they are mostly used by Game Designers and Media Designers for two purposes, unoptimized game engines running games in the development stage(alpha) or closed beta where they are mostly reliant on the CPU and onboard RAM for processing

That's just not true. Granted, coding practices today are working with fairly high-level engines/libraries, but many things are still coded in C/C++ when you get down to the nitty gritty. But more importantly, the addition of high-rez/def videos and larger textures in this world of media overload is what accounts for the greater memory demands. As for processing, physics, advanced lighting, and post processing eat up exponentially more processing cycles than they did just just 5 years ago, let alone 10.

As for system memory (and not graphics/video memory), most games even today don't require much. The memory focus is on video memory for keeping large textures buffered in memory.

System (CPU) processing is still pretty high due to physics and game mechanics, but video (GPU) processing is also on the rise thanks to increased vertex count along with advanced tessellation in DX11.

One thing for sure: the requirements are not due to bloated code or alpha/beta versions. Usually the only difference between a beta and release build is whether debugging is enabled or not.

Psyctooth, on 05 June 2012 - 07:25 AM, said:

As far as for gaming you simply need enough RAM and a fast enough CPU as not to bottleneck your GPU and to make sure you can multitask at a comfortable speed without affecting performance. There was a time when there were "no-nos" to what you did. Now days the only "no-nos" which you still need to follow is not to hard-restart or hard-shutdown a computer because you risk damaging your electronics, CPU and GPU (the later mainly due to the risk of having a defective driver or bios).

You'll almost always have a bottleneck in your computer. What builders should be doing is minimizing the more painful bottlenecks likes the GPU (video cards) and storage (drives).

As for damaging hardware, that doesn't really happen with hard reboots or defective drivers. I really don't think you know what you're talking about in some parts as you seem to fill in the gaps between what you do know with guess-work.

Psyctooth, on 05 June 2012 - 07:25 AM, said:

Also after market coolers are very important regardless if you overclock or not if you are using 2011socket intel or the latest CPUs from both AMD and Intel because simply keeping the CPUs cool always is always something you should strive for even if it is not overlocked. It keeps it operating at peak efficiency and extends its life.

Although I encourage the use of third-party CPU coolers, and all my builds have them, your statement is false. It's not "very important" for those running at stock speeds, nor does the temp difference extend its life.

Did you know you can run most modern-day CPUs without a heat sink? Sounds crazy, I know, but with built-in overheat protection, a CPU will throttle itself depending upon the temps. And it will completely shutdown at its thermal limit to protect itself. But in the end, a stock cooler verses a third-party cooler isn't going to make any difference in a CPU's lifespan. And you'll be upgrading in 2-5 years before that lifespan ever comes into play. There are 486 and Pentium CPUs still running today. Usually power supplies, drives, and motherboards fail long before a CPU does.

[Psyctooth]

Quote

That's just not true. Granted, coding practices today are working with fairly high-level engines/libraries, but many things are still coded in C/C++ when you get down to the nitty gritty. But more importantly, the addition of high-rez/def videos and larger textures in this world of media overload is what accounts for the greater memory demands. As for processing, physics, advanced lighting, and post processing eat up exponentially more processing cycles than they did just just 5 years ago, let alone 10.

As for system memory (and not graphics/video memory), most games even today don't require much. The memory focus is on video memory for keeping large textures buffered in memory.

System (CPU) processing is still pretty high due to physics and game mechanics, but video (GPU) processing is also on the rise thanks to increased vertex count along with advanced tessellation in DX11.

One thing for sure: the requirements are not due to bloated code or alpha/beta versions. Usually the only difference between a beta and release build is whether debugging is enabled or not.

Some game engines in the development stages make use of the CPU more than any GPU until optimized, this is often the case during development and testing stages while the game engine isn't optimized as of yet. This doesn't include middleware game development packages as most are optimized for most commercial hardware by the time they enter commercial release. Also granted yes in the case of storage it is to do with high resolution textures, though it would also do with unbaked lightmaps if it is a world builder which displays a lightmap before it is rendered, I just didn't go into detail or to why the use is needed. I often never use more than 6GBs of RAM myself, though I am not working on complex maps like you'd see in Crysis 2; also few games make use of the features of DirectX11, given that until implemented have little to no real impact on world building; unless if the world builder is optimized for use with DirectX11 and/or 64-bit operating systems.

I don't go into detail about anything, reason for this is because I don't want to confuse people. Granted some developer tools recommend high amounts of ram in case you use high resolution textures but rarely make use of such amounts as often texture resolution differs between games regardless of which engine you run it on anyway despite size standards which are often followed.

As to cooling a CPU you'd get better performance if you keep it below 40 degrees C at all times. If it overheats it can reduce it's life even with it's emergancy cut-offs which depending on the chipset I believe is usually around 100 degrees C, but like you said, should upgrade every few years anyway. Don't even get me started on that beautiful little chipset the 486, that was one hardy chip anyway.

You may say I am wrong, though I am a person who sticks to the convictions of convention, may be backwards yes. However to me safety is more important than simply stating something is untrue because some wiki says so. And just because you can run a chip without cooling doesn't mean you should...

Edited by Psyctooth, 06 June 2012 - 07:17 AM.

[cipher]

Psyctooth, on 06 June 2012 - 07:01 AM, said:

Some game engines in the development stages make use of the CPU more than any GPU until optimized

NONE of the engines I've seen or even personally toyed with as a developer ever have that problem. I'm not sure what engine you're talking about, but that is not typical.

Psyctooth, on 06 June 2012 - 07:01 AM, said:

I often never use more than 6GBs of RAM myself, though I am not working on complex maps like you'd see in Crysis 2; also few games make use of the features of DirectX11, given that until implemented have little to no real impact on world building; unless if the world builder is optimized for use with DirectX11 and/or 64-bit operating systems.

Very few games, let alone game engines work in 64-bit. And DirectX 11 only makes use of system memory if you don't have enough video memory. Video cards with 1GB or higher at HD resolutions usually don't need this. Some integrated graphics share system memory, but that's on the low-end.

Psyctooth, on 06 June 2012 - 07:01 AM, said:

I don't go into detail about anything, reason for this is because I don't want to confuse people.

Or rather because you do not know or didn't want to derail the thread?

Psyctooth, on 06 June 2012 - 07:01 AM, said:

As to cooling a CPU you'd get better performance if you keep it below 40 degrees C at all times.

Absolutely, flat-out wrong.

Psyctooth, on 06 June 2012 - 07:01 AM, said:

You may say I am wrong, though I am a person who sticks to the convictions of convention, may be backwards yes. However to me safety is more important than simply stating something is untrue because some wiki says so.

Yes, you are wrong in many points. Sorry, no real way to sugar-coat it. But please keep in mind my responses are not meant as personal attacks. I just happen to know many of the things you mentioned are not a matter of opinion, but a matter of falsehoods since facts state otherwise.

[Psyctooth]

Quote

NONE of the engines I've seen or even personally toyed with as a developer ever have that problem. I'm not sure what engine you're talking about, but that is not typical.

One example, Guild Wars 2.

Quote

Very few games, let alone game engines work in 64-bit. And DirectX 11 only makes use of system memory if you don't have enough video memory. Video cards with 1GB or higher at HD resolutions usually don't need this. Some integrated graphics share system memory, but that's on the low-end.

My point exactly, least we agree with one thing. An example of game engine tools which work in 64-bit, the tools used by Creative Assembly for Shogun 2 Total War.

Quote

Or rather because you do not know or didn't want to derail the thread?

I do know and yes I do not want to derail the thread, last thing I want is people flaming each other.

Quote

Absolutely, flat-out wrong.

I still beg to differ, until shown physical proof, please provide then I'll change my opinion.

Edited by Psyctooth, 06 June 2012 - 07:34 AM.

[cipher]

Psyctooth, on 06 June 2012 - 07:01 AM, said:

As to cooling a CPU you'd get better performance if you keep it below 40 degrees C at all times.

Psyctooth, on 06 June 2012 - 07:32 AM, said:

I still beg to differ, until shown physical proof, please provide then I'll change my opinion.

"Physical proof"? What the heck is that supposed to be?

Your the one making outrageous claims by sharing outrageous advice, so he burden of proof is on you.

[Catamount]

Psyctooth, on 06 June 2012 - 07:32 AM, said:

I still beg to differ, until shown physical proof, please provide then I'll change my opinion.

No, Psyctooth; that's not how it works.

As the claimant, the burden of evidence is on you, not Cipher. It is you who must provide evidence (evidence, not "physical proof", since there is no such thing in empiricism as proof).


There is no evidence, of any kind, of which I'm aware, that CPU performance improves when a CPU is running at below 40C as opposed to above. If it's your opinion, then you should say "In my opinion, you'd get better performance if you keep [your CPU] below 40 degrees C at all times", and even then, not citing your opinion as fact, what you'd be saying would still be something to be regarded as factually wrong until you can provide evidence that it isn't.

Edited by Catamount, 06 June 2012 - 02:12 PM.

[Vulpesveritas]

cipher, on 06 June 2012 - 08:12 AM, said:

"Physical proof"? What the heck is that supposed to be?

Your the one making outrageous claims by sharing outrageous advice, so he burden of proof is on you.

The following would say that lower temperatures do increase CPU life...
http://www.overclock...ct-on-cpu-life/

But as it says, who keeps their CPU for 20 years or more?

[cipher]

Vulpesveritas, on 06 June 2012 - 02:13 PM, said:

The following would say that lower temperatures do increase CPU life...
http://www.overclock...ct-on-cpu-life/

But as it says, who keeps their CPU for 20 years or more?

My point exactly.

And he did say keeping a CPU below 40C "improves performance", which is the outrageous claim I was directly confronting. As you mentioned in a previous post.

Edited by cipher, 06 June 2012 - 05:03 PM.

[Skrapheap]

cipher, on 06 June 2012 - 05:02 PM, said:

My point exactly.

And he did say keeping a CPU below 40C "improves performance", which is the outrageous claim I was directly confronting. As you mentioned in a previous post.

The only thing I can think of that he is referencing is that if you keep a CPU below a certain temp (somewhere around 70ish C) it won't throttle itself, which is pretty well a given. However I don't think there is a significant difference in performance between 40 c and around 65 c, at which point most modern CPUs start throttling themselves as a "safety" feature. I could be wrong about this, and I am not going to volunteer my system as a testbed. Sorry, but I have "enjoyed" the aroma of burnt CPUs enough for 2 lifetimes, and my upgrade components will not arrive till next week....Upgrade my *** I am pretty well rebuilding.

[cipher]

Wulffemein, on 06 June 2012 - 10:31 PM, said:

The only thing I can think of that he is referencing is that if you keep a CPU below a certain temp (somewhere around 70ish C) it won't throttle itself, which is pretty well a given. However I don't think there is a significant difference in performance between 40 c and around 65 c, at which point most modern CPUs start throttling themselves as a "safety" feature. I could be wrong about this, and I am not going to volunteer my system as a testbed. Sorry, but I have "enjoyed" the aroma of burnt CPUs enough for 2 lifetimes, and my upgrade components will not arrive till next week....Upgrade my *** I am pretty well rebuilding.

Yes, if he was thinking about thermal throttling, then his temperature point is completely off. You'd have to intentionally have poor cooling to have it throttle. And your temp example is way too low as well unless we're talking about mobile processors. Worse, every CPU generation is different, and not all use thermal throttling.

IOW, of the CPUs that have this feature, with a stock cooler at stock speeds, a non-mobile CPU isn't going to ever thermally throttle itself unless the PC in question has an air flow issue.

[Elkarlo]

About Myth4 you might want to Add:
Aftermarket Cooling is very intresting for Modern CPU because it gives them more Performance !.
Most of the Icore5 and some AMD3 ( the ones with 1xxx) and all FX AMD Cpu's gain Performance by better Coolers.
Reason: They auto overclock.
Turbo Boost and Turbo Core Prozessors use the "free" thermal capacity to Overclock one CPU Core.
In basic words, as long as your CPU Cooler keeps Your CPU below 60 degree your Prozessor gets Overclock and you free Performance.
A good Cooler for 25 bugs dumps about 135 Watt of heat, which keeps your CPU cool and fast even in Turbo mode.

Stock coolers are only for the Unoverclocked TDP not for a Turbo Clocked CPU. So the CPu may throttle up but then very soon throttle Down with a stock Cooler.

About the 40°C and performance.
It is a physical fakt that Semiconductors works "hotter" on high Temprature. Because the Electric Resistance climbs by the Temprature in Semiconductors extrem.

This means a Prozessor which runs Coolers produces less Heat and is more frequent stable.
And because the resistance is lower it will live longer. And we are talking about thinks which are meassurable.
My Fx-6100 needs with his new cooler with all Cores maxed 10 Watts less meassured at the Powersupply Inlet.
(It's a 135 TDP Prozessor cooled from 70°C to 45°C under maximum load, but not a 135 watt Cooler but a 150 Watt cooler)

Edited by Elkarlo, 07 June 2012 - 06:30 AM.

[cipher]

Elkarlo, on 07 June 2012 - 06:17 AM, said:

Stock coolers are only for the Unoverclocked TDP not for a Turbo Clocked CPU. So the CPu may throttle up but then very soon throttle Down with a stock Cooler.

About the 40°C and performance.
It is a physical fakt that Semiconductors works "hotter" on high Temprature. Because the Electric Resistance climbs by the Temprature in Semiconductors extrem.

LOL! I assume that was very dry sarcasm? lol.

Or are you really that naive? I hope not.

Edited by cipher, 07 June 2012 - 07:24 AM.

[Skrapheap]

cipher, on 07 June 2012 - 05:55 AM, said:

Yes, if he was thinking about thermal throttling, then his temperature point is completely off. You'd have to intentionally have poor cooling to have it throttle. And your temp example is way too low as well unless we're talking about mobile processors. Worse, every CPU generation is different, and not all use thermal throttling.

IOW, of the CPUs that have this feature, with a stock cooler at stock speeds, a non-mobile CPU isn't going to ever thermally throttle itself unless the PC in question has an air flow issue.

O.k., I Just looked it up and you are right it's around 100 C, and I have never seen a CPU get up to that level unless I royally ****** up when attaching the heat-sink. Still I don't think there is any real performance difference between 40 c and 100 c, though obviously keeping the CPU cool will extend the life of the CPU. However that was already stated that no one use a processor for more than 10 years old for gaming, at least not if you intend on playing new games.