ASUS, MSI, EVGA GTX 950 Reviewed

[El_Capitan]

Damn, reminds me of how good the GTX 460 was back in the day. Not bad.

[Black64]

At what place do you test crysis 3?

 

I just cannot believe 60FPS with high settings and 2xsmaa on that card.

Unless it is done at the very first level in the interior and facing a wall, then it could be possible.

If so, then the test is not really good imo, for that game at least.

 

Witcher 3 as well, it has higher frames than a gtx 780 lol.

If we need to compare with techspot's benchmarks (http://www.techspot.com/review/1006-the-witcher-3-benchmarks/page3.html)

 

Yea... I dunno that is kinda crazy after reading the review.

[Wesstron]

Here are Techspot's GTX 950 numbers. Does it make more sense now?

 

http://www.techspot.com/review/1049-nvidia-geforce-gtx-950/page7.html

[WarWeeny]

Here are Techspot's GTX 950 numbers. Does it make more sense now?

 

http://www.techspot.com/review/1049-nvidia-geforce-gtx-950/page7.html

 

No lol.

 

68FPS battlefield 4 on OCC and 47 on techspot, not really helping lol.

 

And the point still stands, WHERE do people benchmark those games that do not have a benchmark tool of its own.

Those numbers fluctuate massively, as seen with BF4. They could have done some crazy shit to get low frames, or stare at a wall to have increased frames, no one could say. Same with The Witcher 3, do they gallop full speed on the horse through the starting area, or are they basing the frames of of the beginning of the game where Gerald plays with Ciri? Huge difference in performance right there.

 

I'd say, pick only games with either a build-in benchmark tool, or just show a video of the said benchmark to clear confusion across the board. This will never happen, and i understand why, but still. Developers need to have a benchmark tool included the games it makes, they were always there with earlier games, why stop now? #lazydevelopersthesedays.

[SpikeSoprano]

I think the problem here is you have to take any cards benchmarks with a grain of salt, just because one site gives high frame rates and another site gives low frame rates doesn't mean one or the other is wrong, just different test rigs testing different parts of the game. I read as many reviews as possible on different sites to come to a conclusion. No one site can be right on with reviews 100% of the time.

[bp9801]

I think the problem here is you have to take any cards benchmarks with a grain of salt, just because one site gives high frame rates and another site gives low frame rates doesn't mean one or the other is wrong, just different test rigs testing different parts of the game. I read as many reviews as possible on different sites to come to a conclusion. No one site can be right on with reviews 100% of the time.

 

 

 

Exactly. Every site has a different setup, different testing conditions, different games, different everything. One site may have unexpected driver issues, another may have a card not overclock as high, and a third may have no problems with the test other than it uses Ultra settings on everything.

 

Getting all bent out of shape because of a disagreement over the tests doesn't serve anything in the long run. Look at other places, see what they show, and then come to your own conclusion instead of launching an attack.

[DanTheGamer11]

This review reminding me how much wasted potential there is with my 7950 in Crysis 3 because of my CPU :')

[WarWeeny]

I think the problem here is you have to take any cards benchmarks with a grain of salt

 

Then don't do them at all if they cannot be accurate, just say: "this card is ok and should perform anywhere between card x and card y. It is the same that way, people still have no clue what the card can and cannot do, and know what to expect from performance numbers based on cards it is compared to.

 

Just synthetic benchmarks are the way to go then (Unigene, 3dmark etc.) and just skip gaming all together, unless it has a benchmark tool integrated, otherwise it is just guesswork.

 

Just my 5 cents.

 

 

 

 

 

 

Getting all bent out of shape because of a disagreement over the tests doesn't serve anything in the long run. Look at other places, see what they show, and then come to your own conclusion instead of launching an attack.

 

 

So discussing and pointing out things is now launching an attack...riiight.

[bp9801]

So what you seem to want is a review that says: "This card is really great. It performs above the cards in its price bracket, but doesn't go above the bracket unless you run two in SLI/CrossFire." No game benchmarks, just a synthetic benchmark that does not show real world performance.

[WarWeeny]

So what you seem to want is a review that says: "This card is really great. It performs above the cards in its price bracket, but doesn't go above the bracket unless you run two in SLI/CrossFire." No game benchmarks, just a synthetic benchmark that does not show real world performance.

 

No, i said that to have good benchmark results, you can only rely on synthetic benchmarks since testing with games is just not realistic as each review plays it at a different location (if the game doesn't have an ingame bench of its own), although no one really tells you exactly where the test has started and stopped, so people still do not know where the actual testing has been done.

 

To give an example, Frank said that for crysis 3, he played a part of the opening sequence and just walked around a bit and shot a bullet or two (if even that). That is how he tested the game, and this should be included in the review if a game doesn't have an ingame benchmark of its own..in my opinion of course, nobody is forced to do this, but it does give the reader an idea where the test took place and can recreate the sequence so he/she can compare how their current card performs compared to the reviewed card.

 

It is of course a LOT easier to test it with a game that has a benchmark integrated like sleeping dogs for example. Just run the test, get the results and compare them. With tests like Crysis 3, that has no benchmark tool it is hard to compare, especially in areas that aren't that demanding like the opening area.

 

What i was trying to achieve was that reviewers should include a small explanation on how and where the reviewer did his testing with a game that has no benchmark tool, and on what basis he concludes what the results are (minimum frames? average?, max?).

I think this would give a lot clarity to the readers and get a better idea on how the card performs.

 

As it stands now, Crysis 3  can be ran at 60FPS on the 950 (with no indication where the bench has been done and how he did it), but in chapter 2, the gtx 970 struggles to get even 50fps, and that card is a lot stronger than that gtx950. So based on this review, that 950 should outperform the gtx 970 with ease. Only at the 3Dmark/Unigine benchmark can you really see how much weaker/stronger a card is because those tests are all the same if you use the same settings.

 

See where i am getting at here?

[ccokeman]

 

So what you seem to want is a review that says: "This card is really great. It performs above the cards in its price bracket, but doesn't go above the bracket unless you run two in SLI/CrossFire." No game benchmarks, just a synthetic benchmark that does not show real world performance.

 

No, i said that to have good benchmark results, you can only rely on synthetic benchmarks since testing with games is just not realistic as each review plays it at a different location (if the game doesn't have an ingame bench of its own), although no one really tells you exactly where the test has started and stopped, so people still do not know where the actual testing has been done.

 

To give an example, Frank said that for crysis 3, he played a part of the opening sequence and just walked around a bit and shot a bullet or two (if even that). That is how he tested the game, and this should be included in the review if a game doesn't have an ingame benchmark of its own..in my opinion of course, nobody is forced to do this, but it does give the reader an idea where the test took place and can recreate the sequence so he/she can compare how their current card performs compared to the reviewed card.

 

It is of course a LOT easier to test it with a game that has a benchmark integrated like sleeping dogs for example. Just run the test, get the results and compare them. With tests like Crysis 3, that has no benchmark tool it is hard to compare, especially in areas that aren't that demanding like the opening area.

 

What i was trying to achieve was that reviewers should include a small explanation on how and where the reviewer did his testing with a game that has no benchmark tool, and on what basis he concludes what the results are (minimum frames? average?, max?).

I think this would give a lot clarity to the readers and get a better idea on how the card performs.

 

As it stands now, Crysis 3  can be ran at 60FPS on the 950 (with no indication where the bench has been done and how he did it), but in chapter 2, the gtx 970 struggles to get even 50fps, and that card is a lot stronger than that gtx950. So based on this review, that 950 should outperform the gtx 970 with ease. Only at the 3Dmark/Unigine benchmark can you really see how much weaker/stronger a card is because those tests are all the same if you use the same settings.

 

See where i am getting at here?

 

Please do not put words in my mouth on this one. If you want to insinuate that I am falsifying the results I have a few letters for you. To insinuate that I as you put it in your quote above  "just walked around a bit and shot a bullet or two (if even that)." then we can take this off line for a more direct answer because you call into question my competence and my integrity. That I have no tolerance for. If it sounds callous then the point was made. 

 

Canned benchmarks are tweaked, drivers are tweaked for these canned benchmarks and it shows. The point you seem to be missing is that the where and how matter little. Its how the card compares to the rest of the test suite with the settings used and the benchmark sequence run at that point in time. I could be sitting on a dock dangling my feet in the water with the view skewed towards the sky to maximize FPS. If I run the same sequence from card to card to card then those results are comparable and the performance margin is comparable to be used to gauge differential performance and come to a conclusion that one is better than the other. Canned benchmarks have to be used in some instances but I prefer not to use them as actual game play is much better than a fly by benchmark.

 

But to answer your question and this is the only sequence I will detail is that  I start in the first sequence after entering the tower. I start at the second doorway and break the attendants neck, I then exit the room and walk down the stairs and dispatch the two guards in the main arena, I then continue down and around to the right and kill the guard firing from a distance then closer until he is dead, I then retrace my steps and as the door opens I kill the first guard and enter the room and chase the next one down and dispatch it with a volley of shots hitting him multiple times as he falls cold and lifeless on the deck. I then fir across the arena to shoot out a couple windows forcing shots to stop coming as I make my way back around the bay and up to the control room and kill that guard , again with multiple shots to the head neck and torso then exit stage left down the stairs  and spin to a stop. Wash rinse and repeat over and over and over and over. In the process reloading 4 to 5 times. 

 

Now if you want to call me out in private I can provide you a less sanitary version of my thoughts

[WarWeeny]

 

Please do not put words in my mouth on this one. If you want to insinuate that I am falsifying the results I have a few letters for you

 

Where do you see me saying that you are "falsifying" results?

Nowhere, don't go insinuating yourself please, before making a fit and soft-threat me.

You completely missed the point i was making and you make your own assumptions, don't do that.

Instead, just ask for more clarification, more friendly that way too.

 

 

I'l go in more detail, since you missed my point entirely in the previous post.

 

The point i was making here is that the benchmarks are not really realistic (keyword here since you apparently missed that in my previous comment) in terms of the load the GPU is given, giving a higher result in the end and an unrealistic view of the cards performance. Your description of the bench you do for Crysis 3 proves that as well, the indoor sequence and the small outdoor portion has nowhere near the same load as let's say the starting point of chapter two where you see the grasslands and let the mines explode for example.

 

When seeing the "60FPS" on the benchmarks, people assume they can attain those frames on those settings at all times when playing with that card, and are disappointed when they reach chapter two where the frames drop well below 40 at times. and will never reach that 60FPS again.

 

To prevent the above scenario, and give a more realistic view on things, it is smart to give the viewer an idea on what you did to achieve said frames/results. If you would have added that part about how you got those Crysis 3 frames, everyone would understand how you got those frames and how you do your benchmarks for that game, and people can recreate that for themselves to compare.

 

Nothing more, nothing less, i never meant any "harm" or "insult" to you as a reviewer, i just wanted to say that it may be a good idea to mention what you did for a certain game to get those results, so the readers won't be fooled by the frames with their own assumptions.

 

 

Hopefully my point is coming across now...