Wondering about the performance difference between 1050 & 1060 variants?

[dsaf6229]

Why is there such a big performance difference between 1050 & 1050ti but not between 1060 3gb and 1060 6gb even though the difference between the 2 1060 & 1050 variants is 1 SMM (128 cores and 48 TMU)?

 

[captaincharisma]

because the TI version of the 1050 has more differences then just vram. there is no Ti version of the 1060

 

[k1114]

Percentage of increase. Going from 640 to 768 is a 20% increase. Going from 1152 to 1280 is a 11% increase.

 

[Karadjgne]

A 1050 has set clocks, memory speeds, larger amounts of internals etc. The 1060-6 is essentially a 1060-3 with more ram, so will perform the same in games only using 3Gb ov vram or less, but will have the advantage in games using more than 3Gb.

 

[k1114]

1060 3gb and 6gb have more differences than just vram and I know you guys know this. But forget the other specs because while it does affect it, it's to a lesser degree and complicates this unnecessarily.

A pascal sm contains 128 cuda cores and 8 tmu.

1050 - 5 sm: 640 cores, 40 tmu
1050ti - 6 sm: 768 cores, 48 tmu

1060 3gb - 9 sm: 1152 cores, 72 tmu
1060 6gb - 10 sm: 1280 cores, 80 tmu

Both are a difference of 1 sm: 128 cores, 8 tmu, yet the performance difference between 1050 vs 1050ti is large in comparison to 1060 3gb vs 6gb. The answer is what I've explained: increase percentage.

 

[dsaf6229]

The differences are still the same. If that extra 128 cores and 48 TMU's are present, they provide approximately 3-5 more fps given that memory isn't a issue. However I see that most games today use more than 2gb of vram and generally use about 2.8-3gb of vram. Some games even exceed the 3gb Vram but those are not in the majority. The 1050 is bottlenecked by it's measly 2gb frame buffer but the Ti isn't. In the 1060 case the 3gb is just enough for it to not bottleneck. In the games when the memory isn't beyond 2gb the 1050 performs very similarly to the Ti, about 3-5 fps less. This was seen really well in santigo santigo's comparison between the 1050 and 1050ti and Tom's hardware review of the 2 cards. The percentage difference doesn't make sense since the difference is caused by the hardware and the difference of 128 cores and 48 TMU's is the same between the 1050 and 1060 models. So it seems like it's the memory more than anything.

 

[dontlistentome]

The main reason is money.

The 1050 is positioned, in price, as an entry level gaming card. The 1060 6gb is positioned as the mid level 1080p gaming card. Those are the two price points, low to high.

The 1050 Ti is the mid range between 1050 and 1060, so it has to offer enough of a performance increase over the 1050 to entice customers into spending more on it.

The 1060 3gb is positioned differently. It's for people who want the much higher performance level you get with a 1060, but don't want to or need to buy a 6gb card. So the 3gb has to be fairly close to the 1060 6gb in performance or no one would buy it, they'd just buy the 1050 Ti instead.

 

[k1114]

The performance difference between the 1050s vs the 1060 is not 3-5fps each. For the 1060s the average is 3fps. For the 1050s, the average is 6fps. 10% to 20%, 3fps to 6fps. Both double. Odd.

 

[dsaf6229]

The main reason is money.

The 1050 is positioned, in price, as an entry level gaming card. The 1060 6gb is positioned as the mid level 1080p gaming card. Those are the two price points, low to high.

The 1050 Ti is the mid range between 1050 and 1060, so it has to offer enough of a performance increase over the 1050 to entice customers into spending more on it.

The 1060 3gb is positioned differently. It's for people who want the much higher performance level you get with a 1060, but don't want to or need to buy a 6gb card. So the 3gb has to be fairly close to the 1060 6gb in performance or no one would buy it, they'd just buy the 1050 Ti instead.

The issue is that when memory isn't a bottleneck the 1050ti doesn't offer much of a performance increase. It's because of extra 2gb of video memory that the 1050ti gets 10-12 fps more as it's not constrained by memory. Also the 1050 is usually clocked a bit higher than the 1050ti due to having less cores thus clocks of each core can go up provided both the 1050ti and 1050 have the same power supply and are restricted to the 75W power limit.
If you watch Santigo Santigo's performance comparison between the 1050ti and 1050, there is about a 5 fps difference as worst case scenario. In face in one of the benchmarks he got a difference of 1 fps of average fps between the 1050 and 1050ti. The 1060 3gb still offers a massive performance boost over the 1050ti of about 15-20 more fps. Also the 1060 3gb are relatively close to the price of a 1050ti while offering much better performance. So the 1050ti is in a strange spot where unless it's really close to a 1050 price but significantly cheaper than a 1060 3gb it's worth it. This may be the case in some countries but not in all countries.

 

[dsaf6229]

The performance difference between the 1050s vs the 1060 is not 3-5fps each. For the 1060s the average is 3fps. For the 1050s, the average is 6fps. 10% to 20%, 3fps to 6fps. Both double. Odd.

In one of the benchmarks on youtube, https://www.youtube.com/watch?v=n9emI9RSxxY, you can see that the 1060 6gb is about 5-6 fps ahead of the 3 gb variant. About the same difference as between the 1050 and 1050ti. The performance difference comes out to be about 15% or so. And the 1050 and 1050ti also get about 8-15% performance difference depending on whether they are close to 30 or 60 fps. For example if the 1050ti gets 35 fps and 1050 gets 30 fps the percentage difference is 15%. But if the 1050ti gets 65 fps and 1050 gets 60 fps then the percentage difference is about 8%. If the small 2gb video memory isn't exceeded on a 1050 then it'll be about 10% on average slower than the 1050ti and not really 20% similar to the 1060 case. But if you exceed the frame buffer slightly then there will be a 1-2 fps addition to the 5-6 fps difference between the 1050's thus increasing the difference between the 2. The more vram you exceed the more the difference.

 

[Karadjgne]

You said everything in the very first sentence...

"In one of the benchmarks on YouTube...."

That pretty much invalidates any argument. 1. It applies solely to those particular games. 2. Those particular games are run on a certain setup that's going to be a totally unknown variable. Even the actual fps is suspect as there's a multitude of variables like ram speeds, cpu speeds, IPC, hdd/ssd/NVMe, Lan, every driver and version, windows, startups etc that are going to be different. And that's not including any nvidia control panel settings which most ppl don't even bother moving from default. The only thing YouTube shows is what's Possible, Not what's Probable.

Nobody in real life can see a 5fps difference, between any 2 fps counts. Apart from the fact that there is no single value for fps other than the extremes of minimum and maximum. And they only apply to 1 single instance. Giant firey explosion at end-game = minimum fps. Going up a flight of stairs = maximum fps. The rest of the 50 hours of game play spent somewhere in between. Benchmarks are good for only one thing, relative performance, the exact amount of difference is a useless figure

 

[dsaf6229]

You said everything in the very first sentence...

"In one of the benchmarks on YouTube...."

That pretty much invalidates any argument. 1. It applies solely to those particular games. 2. Those particular games are run on a certain setup that's going to be a totally unknown variable. Even the actual fps is suspect as there's a multitude of variables like ram speeds, cpu speeds, IPC, hdd/ssd/NVMe, Lan, every driver and version, windows, startups etc that are going to be different. And that's not including any nvidia control panel settings which most ppl don't even bother moving from default. The only thing YouTube shows is what's Possible, Not what's Probable.

Nobody in real life can see a 5fps difference, between any 2 fps counts. Apart from the fact that there is no single value for fps other than the extremes of minimum and maximum. And they only apply to 1 single instance. Giant firey explosion at end-game = minimum fps. Going up a flight of stairs = maximum fps. The rest of the 50 hours of game play spent somewhere in between. Benchmarks are good for only one thing, relative performance, the exact amount of difference is a useless figure

It's true that nobody can really see a 5 fps difference in real life. It's also true that the youtube benchmarks are for 1 particular set up. However when seeing seeing the benchmarks, relative performance is to be taken with a grain of salt. For example if a 1050 or 1060 3gb is playing a game at 30 fps and a 1050ti or 1060 6gb is playing that same game at 35 fps there's about a 15% difference relative to each other. Bump the fps to 60 and 65 fps respectively and the relative performance difference lowers to about 8%. Relative performance starts to differ and sometimes by significant margins. However the exact amount of fps difference will stay relatively constant up to about 60 -100 fps (it'll increase more with higher and higher fps but not as fast as relative performance). So if there is a 4-5 fps difference between the 2 1060 models when considering fps up to 60, it should be the same with the 1050 and 1050ti as well and not 9-10 fps which some benchmarks show. On you tube you can see what's possible and then use it to estimate or guess at what's probable.

 

[dsaf6229]

A 1050 has set clocks, memory speeds, larger amounts of internals etc. The 1060-6 is essentially a 1060-3 with more ram, so will perform the same in games only using 3Gb ov vram or less, but will have the advantage in games using more than 3Gb.

A 1060 6gb has more 128 more cores and 8 more TMU's than the 3gb 1060 model. The difference is not just memory. If a game uses 2 gb of vram then the 1060 6gb will still perform better than a 1060 3gb due to the additional cores and TMU's.

 

[Karadjgne]

So in layman's terms, since 60fps is the goal, and I'll make it easy with a 1060-3 running at 55fps and a 1060-6 running at 60fps, the difference in the two cards is so negligible as to be a pointless argument over a benchmark, because anyone viewing the 2 screens in real life couldn't tell the difference.

Realistically, for a 2Gb vram game, there's no difference. The real difference happens in a 5Gb vram usage game where the 1060-3 would be struggling with 45fps and bouncing the adaptive v-sync all over, and the 60fps of the 1060-6. Which will be a noticeable difference.

Take for instance the 3Gb varient of the 1050 vrs the 1060-3. Cores and tmu aside, that's going to be fps count. The 1060-3 getting a few fps more than the 1050-3. The vram will be the limiting factor. Both will get relative fps in equal amounts at the same settings, upto the point where the vram is throttled, then both plummet. Right about the same detail settings. The 1060-6 won't see that limit, so while it only gets a few more fps than the 1060-3, you get relatively higher detail settings for the same fps. So if the 1050/1060-3 gets @60fps on medium before throttle, the 1060-6 will get @60fps on high-very high before throttling. That'll be a considerable viewing difference.