The Myths Of Graphics Card Performance: Debunked (Toms Hardware)

[bystander36]

1. I didn't say it takes away the need for g-sync/freesync
2. How is this justification for pretending it doesn't exist?
3. No, it isn't a compromise, it is just imperfect. Tearing is the default state not something you introduced by the feature. Having vysnc set to enable is a compromise (eliminate tearing, lose FPS). Adaptive vysnc obsoletes the "don't use vsync" choice such that the only 2 legitimate choices are "vsync-on" (the compromise feature) and "vsync-adaptive" (the imperfect non compromise)

1. fair enough.
2. remember, this was just a teaser, there is supposed to be a followup with far more info. Who knows what will be said there, and since it is only on Nvidia cards, it is easy to forget.
3. It definitely is a compromise, but it is a good one for many. Some will not like it because it does allow tearing. Being imperfect does make it a compromise.

 

[taltamir]

Here is the thing, you want to hear all the great things about AMD, and not the faults. Tom's does not ignore the faults. For people who care about those faults, it is a great site to read.

The accusation that tom is biased is ludicrous as was shown by various examples given by people where tom recommended an AMD product.

But not being biased does not make them great, merely not awful. Overall tom's article quality tends to be of much lower quality compared to other tech sites, they lack in technical details, they don't seem to make any discoveries, and they have an excessively large amount of errors that should have been caught by someone of at least moderate competence, much less the so called experts writing their articles.

 

[bystander36]

The accusation that tom is biased is ludicrous as was shown by various examples given by people where tom recommended an AMD product.

But not being biased does not make them great, merely not awful. Overall tom's article quality tends to be of much lower quality compared to other tech sites, they lack in technical details, they don't seem to make any discoveries, and they have an excessively large amount of errors that should have been caught by someone of at least moderate competence, much less the so called experts writing their articles.

They do a number of reviews that are different than other sites, but at the same time, other sites have also found things they haven't compared. Example, the vary link at the start of this tread. How many sites compare noise levels and OCing performance. They were also the very first site to notice the CF microstuttering problem, a year or 2 ahead of PCper, though it was not as in depth, as they did not have FCAT. They were also one of the first sites to compare SLI lanes and performance, though other sites have come and done a better job since.

Please explain to me what it compromises

I already did, it allows tearing. It may be a good compromise for many, but it is a compromise.

 

[VulgarDisplay]

I still fail to see the point of the overclocking section...

They seem genuinely surprised that when you practically turn off the cooler on your GPU you get no benefit from over clocking. Reread their conclusion. They don't say because of the lack of cooling with our stupid contrived scenario these cards lose performance. They actually say its pointless to over clock those cards. Moronic.

 

[TreVader]

I c you've been around aslong as me.

I use to read toms all the way back from 2000-2005 and would post regularly. The some people left and it slowly went downhill. I once in a blue moon look at some of the articles posted but don't bother with the forums since then.

It really did get pretty ridiculous.

Toms sometimes puts out legible stuff and they used to (back then) be pretty alright as a source but lately they're graphs have been horrible and their testing methods are full of holes. They have a bit of an nvidia bias but they aren't shills like pcper.


Their forum is a circus. The only other tech forum i frequent is [H]ard.

 

[taltamir]

I already did, it allows tearing. It may be a good compromise for many, but it is a compromise.

This isn't a compromise. Tearing is the default state. It just doesn't work for 100% of situations.

Adaptive vsync = eliminate tearing under some conditions, no drawbacks (aka, no compromises)
regular vsync = eliminate tearing under all conditions, drawbacks for use (aka, compromise)

 

[taltamir]

How many sites compare noise levels and OCing performance.

I thought all sites who actually benchmark OC check noise levels.

They were also the very first site to notice the CF microstuttering problem, a year or 2 ahead of PCper, though it was not as in depth, as they did not have FCAT.

Really? ok, I am curious now... do you have a link?

They were also one of the first sites to compare SLI lanes and performance, though other sites have come and done a better job since.

Alright, so they have the occasional original idea (assuming it was an original idea and not just the first place you saw doing this).
However, having the occasional good idea only addresses ONE of the issues I raised. And it merely lowers it from "no original discoveries" to "few original discoveries".

 

[bystander36]

I thought all sites who actually benchmark OC check noise levels.

I have never seen a sight take a predefined max noise level, and see the max performance you can achieve with that budget. It is an interesting way to test. I could have used such testing in the past to better choose a card based on my noise needs.

Here is the microstutter review in 2011, prior to PCper's FCAT review: http://www.tomshardware.com/reviews/radeon-geforce-stutter-crossfire,2995-4.html

 

[blackened23]

So. These websites being biased because of their so-called discoveries.

Correct me if i'm wrong, these discoveries were absolute legitimate issues that were discovered and acknowledged by AMD, with CF microstuttering being a prime example. I just don't get it. These discoveries make these sites biased, but these were real, factual, and acknoledged bugs. What gives? Should users just be left in the dark? Apparently, that's what some people here want. Everyone left in the dark. Do you want to be left in the dark with other consumer products? Like an automobile with faulty brakes? The bias review site that makes mention of faulty brakes. How dare they do that. Shills. Jerks. For mentioning a product fault. Excuse me while I roll my eyes.

Come on. I can't believe i'm reading this stuff. Real issues. Get over it. That doesn't make them bias. Funnily enough, HardOCP was actually the website that gave AMD cards the hardest time with crossfire microstutter particularly with the 79xx cards. They just did not have a means to test it with FCAT, and they don't plan to do that, but they made clear mention of microstutter problems with AMD cards for years.

Were they right for doing that ? Yes. You don't leave consumers out in the dark by omitting product faults. Product faults should be front and center. Again I can't believe i'm reading that mentioning product issues are the fault of the review site and not the fault of the product. The level of persecution syndrome to get to that level is quite ridiculous. Brand doesn't even matter here. If NV had an egregious issue that affected their GPUs on a wide level which ruined immersion or experience, I WOULD want to know. I would expect sites to do that. And they have. The 590 VRM issues being a prime example, that was publicized everywhere. GTX 480 heat and noise issues. Review sites had a field-day with it, and I didn't disagree. While I do really like what NV has these days in terms of GPUs and prefer them, I was definitely not impressed by the 480 due to heat and noise issues. And review sites were incessant in slagging the 480. Should they have slagged it? Absolutely. I agreed. I feel similarly about the 290/290X reference design, and AMD was made aware of this as numerous websites made mention of the heat and noise issues. Guess what. The successor to the 290? How much do you want to bet that AMD will improve the reference design of whatever Hawaii's successor will be as a DIRECT RESULT of reactions to their 290X reference design? I'm sure it's going to happen. Product improvement due to issues being brought to their attention FRONT AND CENTER on review sites.

People need to know about these things. It isn't a brand thing. It's a consumer protection thing. It also improves subsequent product. How can anyone not see this. Fermi GTX 480 was slagged for being hot and loud. Nvidia went to the drawing board for the much improved 580, and then the very efficient Kepler and Maxwell architectures. AMD will undoubtedly improve their 290 successor in terms of reference cooling. And even on the 290, frame pacing was fixed. WHY? These product improved are MADE BECAUSE OF DISCOVERIES just like the ones mentioned here. Yet despite this, cries of bias website. Shills. What-ever. These types of reviews directly improve the products you get. These sites doing these discoveries are helping you with subsequent products being better. Not hurting you. Just stop with the persecution syndrome stuff by blaming the website for a product issue. Place the blame where it lies. Has got nothing to do with brand. Everything to do with consumer awareness and protection.

And the good news is the 290 series of GPU fixes these problems with microstutter. Would that have happened had these revelations not been discovered? Who knows. Maybe not. But no one is complaining now are they? They have a better product with smoother frametimes, that's the end result of these discoveries. Yet these sites are biased for the end result of an improved product, by making the powers to be at AMD aware of the issues.

 

[antihelten]

I have never seen a sight take a predefined max noise level, and see the max performance you can achieve with that budget. It is an interesting way to test. I could have used such testing in the past to better choose a card based on my noise needs.

Here is the microstutter review in 2011, prior to PCper's FCAT review: http://www.tomshardware.com/reviews/radeon-geforce-stutter-crossfire,2995-4.html

THG most certainly weren't the first site to pick up on microstuttering. To my knowledge it was pcgameshardware.de back in february 2008 (3.5 years prior to your link):
http://www.pcgameshardware.de/Grafi...ernden-Effekt-von-Multi-GPU-Loesungen-631607/

And that article came about due to a forum thread on 3dcenter.org that started in 2007:
http://www.forum-3dcenter.org/vbulletin/showthread.php?t=371844

So THG was about 4 years late to the party.

 

[bystander36]

THG most certainly weren't the first site to pick up on microstuttering. To my knowledge it was pcgameshardware.de back in february 2008 (3.5 years prior to your link):
http://www.pcgameshardware.de/Grafi...ernden-Effekt-von-Multi-GPU-Loesungen-631607/

And that article came about due to a forum thread on 3dcenter.org that started in 2007:
http://www.forum-3dcenter.org/vbulletin/showthread.php?t=371844

So THG was about 4 years late to the party.

Talking about it, and testing it with tools that can measure it are two different things. We all knew about microstutter, but that was the first one that tested it with FRAPs or any other measure tool, that I was aware of. Besides, those other two were in another language.

 

[antihelten]

Talking about it, and testing it with tools that can measure it are two different things. We all knew about microstutter, but that was the first one that tested it with FRAPs or any other measure tool, that I was aware of. Besides, those other two were in another language.

Yes talking about it and testing are indeed two different things, but I don't know why you mention this, partly since you claimed that THG were the first to notice it not that they were the first to test it, and besides pcgameshardware did in fact test it and with FRAPS (so did the very first poster in the 3dcenter thread), so again you're wrong.

And if the problem is with the links being in german here's one in english from this very forum from 2008:
http://forums.anandtech.com/showthread.php?t=168482

 

[bystander36]

Yes talking about it and testing are indeed two different things, but I don't know why you mention this, partly since you claimed that THG were the first to notice it not that they were the first to test it, and besides pcgameshardware did in fact test it and with FRAPS (so did the very first poster in the 3dcenter thread), so again you're wrong.

And if the problem is with the links being in german here's one in english from this very forum from 2008:
http://forums.anandtech.com/showthread.php?t=168482

Sorry for saying notice, rather than test and a forum post is hardly makes a review.

 

[wand3r3r]

^Blackened23

Sure flaws should be pointed out. Going out of your way to demonstrate overclocking doesn't always produce benefits on a thermally limited (POS cooler) is still pointless. We already know it's on the verge of if not throttling at stock. What good will it do to slide the slider up to 5GHz.

Why didn't they have an overvolting OC competition between 290x and 690/titan/780 ti and show that the NV cards didn't gain with voltage since NV has neutered the voltage? It would be just as pointless.

Pointless test is pointless. Making conclusions from it is even stranger.

One more thought, they conveniently picked a "noise" level which still had some headroom in an NV card. Why didn't they go quieter where it's already incapable of increasing clocks? Or louder where they both can go up? They picked the number they tested for a reason.

/done caring

 

[antihelten]

Sorry for saying notice, rather than test and a forum post is hardly makes a review.

I could find plenty of articles testing microstutter from before the THG one, including english ones, by a simple google search if you wanted, but either way this is pretty off topic.

 

[bystander36]

One more thought, they conveniently picked a "noise" level which still had some headroom in an NV card. Why didn't they go quieter where it's already incapable of increasing clocks? Or louder where they both can go up? They picked the number they tested for a reason.

/done caring

They picked a noise level based on what people find a good reference point. Here is a poll thread to figure out what people wanted:
http://www.tomshardware.com/forum/id-2022222/poll-reference-noise-level-graphics-cards-tested.html

 

[taltamir]

I have never seen a sight take a predefined max noise level, and see the max performance you can achieve with that budget. It is an interesting way to test. I could have used such testing in the past to better choose a card based on my noise needs.

ok, some creative application, as I said, doesn't answer the rest of my points.

Here is the microstutter review in 2011, prior to PCper's FCAT review: http://www.tomshardware.com/reviews/radeon-geforce-stutter-crossfire,2995-4.html

Example of amature writing. They are calling macro-stutter micro-stutter.
micro-stutter is when the average FPS seems to remain constant, but the user notices something is wrong... and if you go look at at the individual frame render times you notice that instantaneous FPS on each frame rendered varies a lot even though the average remains semi-constant.
All they are doing is looking at the average FPS chart and noticing that from second to second it stutters like heck, and falsely calling this micro-stutter because they don't know what they are talking about usually.
Also, the charts are not even labeled properly. I have seen worse from them though

 

[VulgarDisplay]

So. These websites being biased because of their so-called discoveries.

Correct me if i'm wrong, these discoveries were absolute legitimate issues that were discovered and acknowledged by AMD, with CF microstuttering being a prime example. I just don't get it. These discoveries make these sites biased, but these were real, factual, and acknoledged bugs. What gives? Should users just be left in the dark? Apparently, that's what some people here want. Everyone left in the dark. Do you want to be left in the dark with other consumer products? Like an automobile with faulty brakes? The bias review site that makes mention of faulty brakes. How dare they do that. Shills. Jerks. For mentioning a product fault. Excuse me while I roll my eyes.

Come on. I can't believe i'm reading this stuff. Real issues. Get over it. That doesn't make them bias. Funnily enough, HardOCP was actually the website that gave AMD cards the hardest time with crossfire microstutter particularly with the 79xx cards. They just did not have a means to test it with FCAT, and they don't plan to do that, but they made clear mention of microstutter problems with AMD cards for years.

Were they right for doing that ? Yes. You don't leave consumers out in the dark by omitting product faults. Product faults should be front and center. Again I can't believe i'm reading that mentioning product issues are the fault of the review site and not the fault of the product. The level of persecution syndrome to get to that level is quite ridiculous. Brand doesn't even matter here. If NV had an egregious issue that affected their GPUs on a wide level which ruined immersion or experience, I WOULD want to know. I would expect sites to do that. And they have. The 590 VRM issues being a prime example, that was publicized everywhere. GTX 480 heat and noise issues. Review sites had a field-day with it, and I didn't disagree. While I do really like what NV has these days in terms of GPUs and prefer them, I was definitely not impressed by the 480 due to heat and noise issues. And review sites were incessant in slagging the 480. Should they have slagged it? Absolutely. I agreed. I feel similarly about the 290/290X reference design, and AMD was made aware of this as numerous websites made mention of the heat and noise issues. Guess what. The successor to the 290? How much do you want to bet that AMD will improve the reference design of whatever Hawaii's successor will be as a DIRECT RESULT of reactions to their 290X reference design? I'm sure it's going to happen. Product improvement due to issues being brought to their attention FRONT AND CENTER on review sites.

People need to know about these things. It isn't a brand thing. It's a consumer protection thing. It also improves subsequent product. How can anyone not see this. Fermi GTX 480 was slagged for being hot and loud. Nvidia went to the drawing board for the much improved 580, and then the very efficient Kepler and Maxwell architectures. AMD will undoubtedly improve their 290 successor in terms of reference cooling. And even on the 290, frame pacing was fixed. WHY? These product improved are MADE BECAUSE OF DISCOVERIES just like the ones mentioned here. Yet despite this, cries of bias website. Shills. What-ever. These types of reviews directly improve the products you get. These sites doing these discoveries are helping you with subsequent products being better. Not hurting you. Just stop with the persecution syndrome stuff by blaming the website for a product issue. Place the blame where it lies. Has got nothing to do with brand. Everything to do with consumer awareness and protection.

And the good news is the 290 series of GPU fixes these problems with microstutter. Would that have happened had these revelations not been discovered? Who knows. Maybe not. But no one is complaining now are they? They have a better product with smoother frametimes, that's the end result of these discoveries. Yet these sites are biased for the end result of an improved product, by making the powers to be at AMD aware of the issues.

What does this giant wall of text have to do with the article at toms being trash?

 

[bystander36]

ok, some creative application, as I said, doesn't answer the rest of my points.


Example of amature writing. They are calling macro-stutter micro-stutter.
micro-stutter is when the average FPS seems to remain constant, but the user notices something is wrong... and if you go look at at the individual frame render times you notice that instantaneous FPS on each frame rendered varies a lot even though the average remains semi-constant.
All they are doing is looking at the average FPS chart and noticing that from second to second it stutters like heck, and falsely calling this micro-stutter because they don't know what they are talking about usually.
Also, the charts are not even labeled properly. I have seen worse from them though

Clearly you misread their charts. Read this page, and it clearly is talking about micro-stutter: http://www.tomshardware.com/reviews/radeon-geforce-stutter-crossfire,2995-3.html

You obviously have made up your mind, so what ever.

 

[VulgarDisplay]

Its a myth that needed to be debunked that when you stop a GPU from running within spec performance suffers according to toms. I stand by what I said about the article being trash.

 

[Headfoot]

It's strange how somehow recently noise has become weighed more and more heavily. I've noticed that given that noise tolerance is subjective, it is applied more frequently when the reviewer/commentator seeks a particular result...

 

[MongGrel]

I always take anything off Tom's with a grain of salt more or less, and I've seen people argue on a few forums about stuttering on various forums because of it and just going way over the top.

Even when people have proven they are doing other things wrong.

 

[taltamir]

Clearly you misread their charts.

Gee, its almost as if they didn't actually bother labeling the things properly, which proves my point. The one on the specific page you linked is even worse... half wedges, half circle, that half circle separated into bands, no labels, no explanation of what is what on the chart. WTF are they even saying.

Also... I just googled
http://www.guru3d.com/news_story/amd_anti_micro_stutter_crtossfire_driver_delayed_to_late_july.html

I thought you said tomshardware discovered the crossfire microstutter.
Yet there are tons of forums discussing it before and here is an actual published news story mentioning it in 2010, while the toms article is from 2011.

EDIT: PS I am not saying they are WORTHLESS (another guy called them trash, not me). I am saying there are better sites out there. but to be honest, there are much much much worse sites out there too.

 

[norseamd]

http://www.tomshardware.com/reviews/thermal-paste-heat-sink-heat-spreader,3600.html

would you say this article is legit considering i used it quite a bit when building my computer? so i have the thermalright 140 and am wondering if it was right to spread the thermal paste first since that seems what the instructions for chill factor 3.

 

[Cerb]

It's strange how somehow recently noise has become weighed more and more heavily. I've noticed that given that noise tolerance is subjective, it is applied more frequently when the reviewer/commentator seeks a particular result...

It's weighted more heavily because we can have high performance within a reasonable SPL limit. It used to be you got one or the other, so aside from differentiating jet engine coolers from acceptable ones, it was of minor importance. Now, cards are preemptively throttling themselves, so it's getting more complicated. It's a huge factor in my purchasing, and I'm glad sites are testing it in varied ways.