Whose tests are more accurate?

[uco73]

Techpowerup vs Passmark

 

[W1zzard]

It has more cores than 3090, higher clock, much more memory, but slightly lower memory bandwidth

+11% over 3090 is probably a bit too optimistic by our algorithm, but no reason it should be slower than 3090 unless the power limit is very low

 

[TheoneandonlyMrK]

Techpowerup vs Passmark

It's not passmark.

 

[uco73]

It has more cores than 3090, higher clock, much more memory, but slightly lower memory bandwidth

+11% over 3090 is probably a bit too optimistic by our algorithm, but no reason it should be slower than 3090 unless the power limit is very low

I trust your tests more than any other.

 

[Toothless]

I trust your tests more than any other.

So what's the point of the thread? First you ask which is more accurate but then say you trust TPU's more.

 

[uco73]

So what's the point of the thread? First you ask which is more accurate but then say you trust TPU's more.

OK
You're free to delete this thread. I really didn’t mean anything bad. I was just interested in the opinion of most about these tests. I didn’t mean to provoke any ugly controversy.

 

[maxfly]

OK
You're free to delete this thread. I really didn’t mean anything bad. I was just interested in the opinion of most about these tests. I didn’t mean to provoke any ugly controversy.

All questions are valid in a forum, unless your purposefully trying to troll. You dont seem to be one of those.
Try to be a little more descriptive when your laying out your initial question or concerns. Many times when people ask vague questions theyre trying to stir the pot or create drama.

 

[uco73]

All questions are valid in a forum, unless your purposefully trying to troll. You dont seem to be one of those.
Try to be a little more descriptive when your laying out your initial question or concerns. Many times when people ask vague questions theyre trying to stir the pot or create drama.

I was just wondering why the results were different for the same graphics cards, just as I was wondering if TPU was testing graphics with some special tests or directly in games.

 

[mtcn77]

The general opinion definitely has a touch of frivolity. I can swear HW Unboxed guys used to be groundbreakers, yet they are currently up with shilling accusations. I don't get what news source today's pc users get their news from. It certainly feels like upside down - australia vs the world.

 

[W1zzard]

I was just wondering why the results were different for the same graphics cards, just as I was wondering if TPU was testing graphics with some special tests or directly in games.

Just to clarify, a lot of "standard" cards in the gpu database performance chart are based on my performance tests.

For cards with that text, the performance is estimated using an algorithm, there is no actual measurement

So on your screenshot A6000 is estimated, all other comparison cards are based on my actual data

 

[grammar_phreak]

I don't write much on here, but Passmark is one benchmark I have a bone to pick with.
According to Passmark the 980ti beats the 1070 and Vega 56 while slightly trailing the 1070ti. The GTX 970 is ahead of the RX 590. Then the 2080ti is slightly ahead of the 3070ti. A few years ago the i7-4790k enjoyed a lead over the i7-6700k, but that's changed.

It's not always accurate. A lot of what influences Passmark scores are scores submitted. So if a lot of people benchmark an overclocked 980ti, then it will throw off the overall score.
I wouldn't have a problem with Passmark but on other forums there are people who make hardware recommendations based on Passmark results, and this steers the person asking the question in the wrong direction. Granted, there are a handful of games where a 980ti would etch by a 1070 and a Vega 56. So in a way, looking at Passmark results is like looking at one game, at one screen resolution and then assuming the results for everything else will be the same.

 

[Tom Sunday]

In general I challenge all and any hardware testing done! No matter what program or technology is being employed. Everyone who follows the tech-sites knows that some sites have very obvious biases. What do you read in those reviews and tests, and who writes them, especially the reviews on the products that appear on the first few pages of the many tech-sites? Then who are these people who actually write them and give a product, a full five-star rating? These are the people that get PAID to write reviews and their employer makes money from the advertisers which subscribe to their channel. Paid advertisers in that respect get more recognition and or more consideration than non-paying companies. Or for that matter companies who provide free hardware samples. Testing methods and 'comparative charts' in many instances can also be sullied by either omitting direct competitors that are known to having a quality product or using testing techniques that specifically hamper the product of others. Case in point: "Would you pay someone as a fee paid advertiser in publicly testing your new to the market product with the expectation of receiving a poor review?" Capitalism always stomps on ethics more often than not. This is not to say that all (www) "Review Sites" are bad apples. But do note that many of them are not playing by the golden rule! So buyer be aware.

 

[mtcn77]

I don't write much on here, but Passmark is one benchmark I have a bone to pick with.
According to Passmark the 980ti beats the 1070 and Vega 56 while slightly trailing the 1070ti. The GTX 970 is ahead of the RX 590. Then the 2080ti is slightly ahead of the 3070ti. A few years ago the i7-4790k enjoyed a lead over the i7-6700k, but that's changed.

It's not always accurate. A lot of what influences Passmark scores are scores submitted. So if a lot of people benchmark an overclocked 980ti, then it will throw off the overall score.
I wouldn't have a problem with Passmark but on other forums there are people who make hardware recommendations based on Passmark results, and this steers the person asking the question in the wrong direction. Granted, there are a handful of games where a 980ti would etch by a 1070 and a Vega 56. So in a way, looking at Passmark results is like looking at one game, at one screen resolution and then assuming the results for everything else will be the same.

How passmark tests is, they pick a random goalpost say sse2 or sse4.2 and then run along with it. They thus come up with results that they generalize thereafter. This is what they do.

 

[lexluthermiester]

Techpowerup vs Passmark

If I may offer a point: You're comparing a single benchmark suite test to a whole series of test runs in a group. There's bound to be variations. However, I would say that Passmark's result compliments the testing results offered by TPU. But that's my interpretation.

Everyone who follows the tech-sites knows that some sites have very obvious biases.

Rubbish. Bias has almost nothing to do with results from reputable tech sites. The variations that happen are a result of differences, slight though they might be, in testing methodologies. There is nothing nefarious to be seen. The differences often highlight a particular strength or weakness of the hardware being tested. This is why it is important to seek out multiple sources of review data when researching a major hardware purchase.

 

[Tom Sunday]

....multiple sources of review data when research a major hardware purchase

"Multiple sources"...that is naturally the key with many purchase decisions. You are also on-point in having interjected the word "reputable" into your reply. Actually I do not remember ever having made a single purchasing decision from the simple reading of any tech-site review. Or my subscription to their instruments of 'testing tortures' light years away from real life circumstances and realities. And most important never to relying on the people that get PAID to write reviews. Bias? As already regurgitated here earlier, capitalism always stomps on ethics more often than not. This in many respects is where 'bias' enters my threshold. That's my interpretation in how I see the world turn at least for now!