Slight issue with review testing ,, opinion

[TheoneandonlyMrK]

So im a big fan of TPU GPU reviews but one thing about them I think is not so good and worth either ditching or improving is the mining performance tests, either ditch them because the figures in no way match what someone will actually mine with them at, If they were to buy them and do so.

For example, Vega 64 is said by you to do 33.14Mh/s ,but this is not how it will be run though ,at all , my card and most others I've noted do 42-44Mh/s at admittedly custom tuned settings and low watts(no bios mod at all), point is though that, that's how it would be run.

as it is now its not very informative for a potential buyer and not interesting to others.

there would still be something left to ponder, in that most miners will bios mod amd cards to get more, but I would accept as sufficient a stock bios , lowest watts for max hashes type testing methodology and it also has to have the system configured right to mine, ie most of the RX580 scores look like compute mode was not enabled because my stock bios saphire nitros do 29.5Mh/s with a slight memory tweak oc to 2100.

 

[qubit]

at admittedly custom tuned settings and low watts(no bios mod at all), point is though that, that's how it would be run.

I think your answer is right here, in bold and hence the reviews are correct.

Any review has to test the performance of a stock card, not one that could be modded in one of hundreds of unspecified possible ways. As it is, there are overclocked tests to give one an idea of how far the card could be pushed, so that covers any extra performance you may see with your card.

 

[TheoneandonlyMrK]

I think your answer is right here, in bold and hence the reviews are correct.

Any review has to test the performance of a stock card, not one that could be modded in one of hundreds of unspecified possible ways. As it is, there are overclocked tests to give one an idea of how far the card could be pushed, so that covers any extra performance you may see with your card.

so when overclocking's tested are sliders not slid??

I used wattman to tune the card ,nothing other then the manufacturers driver oc section , stock latest drivers too so no , im sorry qubit i dissagree.

the overclock bench testing is done right if a bit un-enthusiastically ,i believe id do better , im just saying only a fool would mine as these tests are run,, and no one either in the know or other wise just sets a stock config card mining , they are always tweaked , and should be.

in effect the figures agree with nothing else i've seen in the real world , whereas all the other tests are fairly representative of what joe bloggs can do , and does do on his pc.

I should say i have had experience with 8 types(1060,1080,1080ti ,rx480, 580 ,vega and a few bits of old tat) of card from about 6 different Oem's so im not just talking about the vega's figures either.


Do note im saying this because as a small scale miner ,when i look at that bit, what im really wanting to know is what the memory over clocks like, v power use and whats its reasonable max memory oc/ core under clock and power setting, its well known that the max memory oc for games is not the max for compute ,since the load is lighter and more stable.

 

[Divide Overflow]

Reviews are done at stock and are very representative of what can be "seen in the real world". There are no guarantees with overclocking and results will vary wildly.

 

[TheoneandonlyMrK]

Reviews are done at stock and are very representative of what can be "seen in the real world". There are no guarantees with overclocking and results will vary wildly.

no sir you are wrong, the overclocking section is not at stock.

see my post above yours regarding that.

and your wrong on point two , buy a batch of cards of the same type ,brand and version and the results these days are not so varied, and while I certainly accept there is a silicon lottery to account for in all cases , while mining, the lottery moves to the memory chips...... .fullest of stops I assure you though all my refs clock the same, all my nitros clock the same, and my pulses have all clocked the same.

As i said ,no need to balls to the wall test this but i think more could be done, or less??

 

[newtekie1]

None of the review scores are meant to give an idea of what exact performance you should expect. The game tests are not supposed to be used to tell you what FPS to expect in each game, and the mining test is not supposed to give you what exact hash rate to expect. You can't match W1z's setup exactly, and for the game tests you don't even know what exact settings and scenes were used in the testing.

The tests are supposed to be used to compare the card to other cards, to give an idea of how the card performs relative to other cards. And I think the mining results do a very good job of showing that.

 

[hat]

The issue here is that most miners do not run their cards like that. I certainly don't. So, for most miners, it is a useless performance metric. Of course, most people even viewing tpu reviews likely aren't mining, or looking for those results here if they are.

It would take a good bit more work to include that in the reviews in the way most miners would want to see it, though. I found my settings based on what other people said about the 1070 mining and some of my own tinkering. It would be a tough process to do that for every card that gets reviewed.

 

[the54thvoid]

It's a benchmark against other cards. Always has been. It's to inform of relative performance over another card. Simple really.
You want mining data that's accurate, you need an appropriate forum.

 

[TheoneandonlyMrK]

None of the review scores are meant to give an idea of what exact performance you should expect. The game tests are not supposed to be used to tell you what FPS to expect in each game, and the mining test is not supposed to give you what exact hash rate to expect. You can't match W1z's setup exactly, and for the game tests you don't even know what exact settings and scenes were used in the testing.

The tests are supposed to be used to compare the card to other cards, to give an idea of how the card performs relative to other cards. And I think the mining results do a very good job of showing that.

Again I disagree, The whole point of reviews is to give the reader an idea of what performance to expect when they buy that product?? How can it not be??

 

[mtcn77]

Again I disagree, The whole point of reviews is to give the reader an idea of what performance to expect when they buy that product?? How can it not be??

You cannot change the opinion of bigots. That is not how a forum is supposed to work. Their white-washed detachment from tweaking has turned overclocking on its head. Now, it is the mainstream who claim that title and the spoils of glory thereof, albeit the achievement.

 

[hat]

You cannot change the opinion of bigots. That is not how a forum is supposed to work. Their white-washed detachment from tweaking has turned overclocking on its head. Now, it is the mainstream who claim that title and the spoils of glory thereof, albeit the achievement.

I don't see how this has much to do with tweaking and overclocking. Yes, tweaking is involved, but closer to the heart of the matter is obtaining results that look more like what we would find in the real world, because I don't see many people running these cards at stock settings while mining. That said, I don't see anyone trashing overclocking in this thread, either...

 

[Caring1]

The way I see it, graphics cards are mainly bought for gaming, so any further charts and data is extraneous and irrelevant.
I would be happy if hash rates tables were left off.

 

[mtcn77]

I don't see how this has much to do with tweaking and overclocking. Yes, tweaking is involved, but closer to the heart of the matter is obtaining results that look more like what we would find in the real world, because I don't see many people running these cards at stock settings while mining. That said, I don't see anyone trashing overclocking in this thread, either...

It is happening by having no prospect in place.

 

[TheoneandonlyMrK]

Yes

It's a benchmark against other cards. Always has been. It's to inform of relative performance over another card. Simple really.
You want mining data that's accurate, you need an appropriate forum.

And you believe im not looking at those forums? , or that I don't get reviews, read the thread, i get reviews.
But a ref vega64 on latest drivers with no compute mode available on balanced ,default mode does 36.2 Mh/s you have to switch to power save to get 33.3 Mh/s and i tweak just voltages and frequency via wattman too, just with care for the application instead of just because, see how scientific that seams.
I can easily do better than the 1080 score without sweating a 1060?? Too ,without any risk at all to the hardware using manufacturers packaged oc software.
I get that w1zzard is busy ,and possibly not that interested in mining ,im just trying to be subjective to increase the worth of and indirectly the use of those charts, i think of him as a reasonable and reliable source of gpu info but I think a little change and possibly a chat with Rtb and he could get more views and visitors.

 

[qubit]

so when overclocking's tested are sliders not slid??

I used wattman to tune the card ,nothing other then the manufacturers driver oc section , stock latest drivers too so no , im sorry qubit i dissagree.

the overclock bench testing is done right if a bit un-enthusiastically ,i believe id do better , im just saying only a fool would mine as these tests are run,, and no one either in the know or other wise just sets a stock config card mining , they are always tweaked , and should be.

in effect the figures agree with nothing else i've seen in the real world , whereas all the other tests are fairly representative of what joe bloggs can do , and does do on his pc.

I should say i have had experience with 8 types(1060,1080,1080ti ,rx480, 580 ,vega and a few bits of old tat) of card from about 6 different Oem's so im not just talking about the vega's figures either.


Do note im saying this because as a small scale miner ,when i look at that bit, what im really wanting to know is what the memory over clocks like, v power use and whats its reasonable max memory oc/ core under clock and power setting, its well known that the max memory oc for games is not the max for compute ,since the load is lighter and more stable.

I see the others have replied since you posted this and it still hasn't convinced you, so there's not that much I can add really, but I'll try.

The fact you've used wattman to alter the cards performance means you've moved the card away from stock. All reviews must be done at the stock performance of the card (that means taking it out of the box, installing and running it with no performance adjustments of any kind) for a meaningful comparison with the competition. If the reviewer wants to add in overclocking results, which many do, that's purely optional, since cards all overclock differently, ie the performance isn't guaranteed. It is guaranteed at stock however and comparable to the competition, and that's the difference. This applies to mining as much as any other task you give the graphics card.

As the others have said, if you want more detailed mining results, then a different site is for you.

Hope this helps.

 

[eidairaman1]

I don't see how this has much to do with tweaking and overclocking. Yes, tweaking is involved, but closer to the heart of the matter is obtaining results that look more like what we would find in the real world, because I don't see many people running these cards at stock settings while mining. That said, I don't see anyone trashing overclocking in this thread, either...

Considering the results are catered to the Gamer, I'm sure W1zzard or one of his staff isn't going to go through bios modding a card for mining on a review and just gave mining at stock, considering the samples they get they typically have to return.

 

[sneekypeet]

All this back and forth and nobody tagged @W1zzard to get an answer straight from the source?

@eidairaman1 the cards dont go back. If they did, w1zz could not retest when new drivers show up.

 

[TheoneandonlyMrK]

I see the others have replied since you posted this and it still hasn't convinced you, so there's not that much I can add really, but I'll try.

The fact you've used wattman to alter the cards performance means you've moved the card away from stock. All reviews must be done at the stock performance of the card (that means taking it out of the box, installing and running it with no performance adjustments of any kind) for a meaningful comparison with the competition. If the reviewer wants to add in overclocking results, which many do, that's purely optional, since cards all overclock differently, ie the performance isn't guaranteed. It is guaranteed at stock however and comparable to the competition, and that's the difference. This applies to mining as much as any other task you give the graphics card.

As the others have said, if you want more detailed mining results, then a different site is for you.

Hope this helps.

That's not an agreeable stand point to me so no it doesn't.
It's one section of a large review just like overclocking.
And just like overclocking , it's a tangential use of (using the exact same sliders no less), and unnecessary addition to a review, but they are liked.

And continuously pointing traffic to other sites, when they likely moved to here to get more credible facts and figures from amounts to comunity suicide to me so I'll ignore that shit, i look around,, believe me.

And im purely talking about software overclocks.

Sorry i forgot that detail sneeky

 

[mtcn77]

Again, is there such a thing as a stock gpu these days? Some of us old-schoolers haven't caught up with standard of the day industry practices, it seems. The fact that you can, or cannot voltmod your card is very telling of the landscape. Some people are conveniently ignoring the other half of the proposition...

 

[eidairaman1]

All this back and forth and nobody tagged @W1zzard to get an answer straight from the source?

@eidairaman1 the cards dont go back. If they did, w1zz could not retest when new drivers show up.

So you are telling me they keep each and every card they get even if it is the same exact die but from different card brands?

@mtcn77 stock means manufacturer did the footwork, not the end user.
Example My CPU, AMD had of 4.0GHz with 4.2 Turbo core, that is stock. I did the footwork to find 5.0GHz with turbocore off so it applies to all 8 cores. On about cards being software volt moded or lack of-they are doing it to reduce RMAs due to users over doing it on volt mods, plus marketing.

 

[rtwjunkie]

So you are telling me they keep each and every card they get even if it is the same exact die but from different card brands?

No, he doesn’t. You are correct. @W1zzard has stated many times he does not keep all the different brands of a GPU model. He keeps the base, stock, reference version, if one exists. This is why, for instance if you see a new card come out from a competitor six months later and it beats “your” card, it may not actually be beating it.

W1z only goes back and tests that plain Jane version, not your hyped up superclocked version. That’s why you have to take comparison tests later with a grain of salt.

 

[eidairaman1]

No, he doesn’t. You are correct. @W1zzard has stated many times he does not keep all the different brands of a GPU model. He keeps the base, stock version. This is why, for instance if you see a new card come out from a competitor six months later and it beats “your” card, it may not actually be beating it.

W1z only goes back and tests that plain Jane version, not your hyped up superclocked version. That’s why you have to take comparison tests later with a grain of salt.

Considering a 580 oc to 1450 on a reference board is the exact same as on a non reference...

 

[mtcn77]

Considering a 580 oc to 1450 on a reference board is the exact same as on a non reference...

This couldn't be further from the truth - what about the gpu memory parts supplier? Are all vendors, series, tiers the same?

 

[qubit]

No, he doesn’t. You are correct. @W1zzard has stated many times he does not keep all the different brands of a GPU model. He keeps the base, stock, reference version, if one exists. This is why, for instance if you see a new card come out from a competitor six months later and it beats “your” card, it may not actually be beating it.

W1z only goes back and tests that plain Jane version, not your hyped up superclocked version. That’s why you have to take comparison tests later with a grain of salt.

+1 to only keeping the reference cards. I remember back in late 2016 he helpfully offered to sell me one of his GTX 1080 non-reference review cards when I was looking to buy a GTX 1080 (no Ti back then) and I couldn't decide which one to get. I got a different model in the end, but I still appreciated the help.

 

[sneekypeet]

So you are telling me they keep each and every card they get even if it is the same exact die but from different card brands?

@mtcn77 stock means manufacturer did the footwork, not the end user.
Example My CPU, AMD had of 4.0GHz with 4.2 Turbo core, that is stock. I did the footwork to find 5.0GHz with turbocore off so it applies to all 8 cores. On about cards being software volt moded or lack of-they are doing it to reduce RMAs due to users over doing it on volt mods, plus marketing.

Indeed, typically reviewers for top tier sites do not send samples back. What happens to them down the line got pointed out by qubit.