Geforce GTX 1050 / 1050 Ti Thread

[MadOver]

For sure will beat 460 easily. Well, that was not a very hard thing to do I guess.. Let's just see how much nvidia tax will be imposed

 

[whm1974]

This!!!

PS. Thanks for this thread, Sweepr.

Edit: For the 1050 to truly "replace" the GTX950, it would have to support VGA monitors. Does it?

VGA and DVI is dead, manufacturers need to stop putting them on new displays and video cards.

 

[Sonikku]

Kinda disappointed about the clocks, with the architecture these should be capable of like 1.8GHz at decent power consumption should they not, possible justify a 6-pin version.

The lower clocks coupled with a likely higher price will probably do much to sour the added performance. This gen has kind of sucked historically speaking.

 

[raghu78]

Why are the boost clocks of GP107 300+ Mhz lower than GP106 and GP104. Is GP107 manufactured at Samsung 14 LPP ? Heck if this is true then we can clearly see that TSMC 16FF+ is more like a half node ahead in terms of transistor performance vs 14LPP. If this is true then it would be interesting to see Vega built at TSMC 16FF+ which I think is likely to be the case as TSMC has superior transistor performance and much better yields especially for large die GPUs.

 

[Arachnotronic]

Why are the boost clocks of GP107 300+ Mhz lower than GP106 and GP104. Is GP107 manufactured at Samsung 14 LPP ? Heck if this is true then we can clearly see that TSMC 16FF+ is more like a half node ahead in terms of transistor performance vs 14LPP. If this is true then it would be interesting to see Vega built at TSMC 16FF+ which I think is likely to be the case as TSMC has superior transistor performance and much better yields especially for large die GPUs.

Can we not turn this into a Vega discussion? Thread's about the GTX 1050

 

[raghu78]

Can we not turn this into a Vega discussion? Thread's about the GTX 1050

Anyway even on topic whats the reason for these much lower boost clocks. I am thinking the 14LPP process is the factor at play here.

 

[96Firebird]

Why are the boost clocks of GP107 300+ Mhz lower than GP106 and GP104. Is GP107 manufactured at Samsung 14 LPP ? Heck if this is true then we can clearly see that TSMC 16FF+ is more like a half node ahead in terms of transistor performance vs 14LPP. If this is true then it would be interesting to see Vega built at TSMC 16FF+ which I think is likely to be the case as TSMC has superior transistor performance and much better yields especially for large die GPUs.

The chart says 16nm for the 1050.

 

[Arachnotronic]

Anyway even on topic whats the reason for these much lower boost clocks. I am thinking the 14LPP process is the factor at play here.

I would guess power consumption, quite frankly. They need to get it comfortably sub-75W so that it not only doesn't require a PCIe connector but, out of the box, has overclocking headroom for enthusiasts/geeks to mess around with.

The lower power consumption should also help for thin and light gaming capable notebooks.

That said, I hope NVIDIA sticks with 16FF+ on this one, it's clearly the superior process to 14LPP. But, who knows. We'll find out soon-ish.

 

[jpiniero]

Anyway even on topic whats the reason for these much lower boost clocks. I am thinking the 14LPP process is the factor at play here.

I know there were rumors that nVidia were going to do GP107 and GP108 at SS. But I don't think that's the case. Pascal Refresh OTOH...

 

[Arachnotronic]

I know there were rumors that nVidia were going to do GP107 and GP108 at SS. But I don't think that's the case. Pascal Refresh OTOH...

There is no Pascal refresh. Also 14LPP is an inferior process, so why would a hypothetical (and non-existent) Pascal refresh be manufactured on that process over 16FF+?

 

[jpiniero]

There is no Pascal refresh. Also 14LPP is an inferior process, so why would a hypothetical (and non-existent) Pascal refresh be manufactured on that process over 16FF+?

I'm sure they want to release something new to consumers next year, and it's not going to be Volta. It is denser so they would be able to stick in more cores in the same die area.

SS's process is worse, but it's not as bad as you think. It's GloFo's bastardization of Samsung's node that's bad.

 

[b-mac]

I would guess power consumption, quite frankly. They need to get it comfortably sub-75W so that it not only doesn't require a PCIe connector but, out of the box, has overclocking headroom for enthusiasts/geeks to mess around with.

The lower power consumption should also help for thin and light gaming capable notebooks.

That said, I hope NVIDIA sticks with 16FF+ on this one, it's clearly the superior process to 14LPP. But, who knows. We'll find out soon-ish.

I am very interested to see how this performs and which thinner laptops it can find its way into.

 

[Arachnotronic]

I'm sure they want to release something new to consumers next year, and it's not going to be Volta. It is denser so they would be able to stick in more cores in the same die area.

SS's process is worse, but it's not as bad as you think. It's GloFo's bastardization of Samsung's node that's bad.

They could also just increase die area a little bit on 16FF+ to do refreshed products. A few mm^2 of silicon in the context of multi-hundred dollar dGPUs is peanuts.

But, it's a moot point, since there is no Pascal refresh.

 

[antihelten]

Doesn't matter, this is just an example where shading performance doesn't translate to overall gaming performance. And you don't have data to prove that shading (or bandwidth) will be the bottleneck regarding 1050 vs 1060, as you carefully stated here:

You're just assuming the former is the case, and doing some pessimistic performance predictions based on it. I think it will be closer to a 960 than you expect.

The point is that I said that when shader performance is the main bottleneck then gaming performance will follow shading performance, you then said that you disagreed with that and used an example where shader performance is not the main bottleneck, which really didn't make any sense at all.

In other words you basically said that you disagree with gaming performance following shading performance when shader performance is the main bottleneck, because in cases where shader performance isn't the main bottleneck gaming performance doesn't follow shading performance. This is of course completely obvious and also completely irrelevant to the point, it's basically the same as saying that something is true when it is true and false when it is false.

And yes I did run with the assumption that shader performance would be the main bottleneck in my estimates, and yes that very well may end up being a pessimistic assumption, in fact I think it may very well be so, because if shader performance is in fact the main bottleneck then Nvidia could have gotten away with simply using 6gbps memory instead of 7gbps (6gbps would have given the 1050 50% less bandwidth than the 1060, which would have been in line with the drop in shading performance). So all in all performance somewhere around 960 may very well be more realistic.

 

[ConsoleLover]

It would <profanity redacted> if it was actually released at 950 price point, that card was massively overpriced at $160 for the worthless 2GB and &190 for the 4GB version. Something like this needs to be in the price range of $100 to $110 to compete with the RX 460.


Profanity is not allowed in the technical forums,
Markfw900

 

[Dribble]

It would <profanity redacted> if it was actually released at 950 price point, that card was massively overpriced at $160 for the worthless 2GB and &190 for the 4GB version. Something like this needs to be in the price range of $100 to $110 to compete with the RX 460.

Why would nvidia lower the price point of the 1050, making it cheaper then the 950, which was a pretty successful card? As for the RX 460, it's just about able to compete with the current 950, so it's hardly like AMD is going to be putting much pricing pressure on the significantly faster 1050. Sure we'd all like the 1050 to be cheaper, but I've got to be realistic - without sufficient competition I expect the price to go up not down.

 

[Arachnotronic]

Why would nvidia lower the price point of the 1050, making it cheaper then the 950, which was a pretty successful card? As for the RX 460, it's just about able to compete with the current 950, so it's hardly like AMD is going to be putting much pricing pressure on the significantly faster 1050. Sure we'd all like the 1050 to be cheaper, but I've got to be realistic - without sufficient competition I expect the price to go up slightly.

Yeah, if the 1050 outperforms RX 460, it's going to be priced higher. Period.

 

[SPBHM]

Yeah, if the 1050 outperforms RX 460, it's going to be priced higher. Period.

if? with those specs it's impossible not to, also the 460 4GB is not all that well priced, let's not forget the 1060 is only $10 higher than the 480 8GB while being clearly faster, so I don't expect the 1050 to be to badly priced in relation to the 460 4GB, I expect it to be better perf / $

 

[BeauCharles]

It would <profanity redacted> if it was actually released at 950 price point, that card was massively overpriced at $160 for the worthless 2GB and &190 for the 4GB version. Something like this needs to be in the price range of $100 to $110 to compete with the RX 460.


Profanity is not allowed in the technical forums,
Markfw900

The GTX 950 never had a 4GB version. You must be thinking of the GTX 960.

 

[VirtualLarry]

Anyways, I'll be the first to tell you that I'm not really a big NVidia fan, but I'm going to wait to see how this 1050 turns out. Was considering a RX 460 4GB Gigabyte dual-fan model, but they canned the VGA output capability, apparently.

 

[dark zero]

For sure will beat 460 easily. Well, that was not a very hard thing to do I guess.. Let's just see how much nvidia tax will be imposed

That's because the RX 470 is the competitor of the 1050.... not the 460. And the competitor of the 460 is supposed to be the GT 1040 and I see nVIDIA releasing it along the 1030 as the bottom of the barrell.

 

[whm1974]

Anyways, I'll be the first to tell you that I'm not really a big NVidia fan, but I'm going to wait to see how this 1050 turns out. Was considering a RX 460 4GB Gigabyte dual-fan model, but they canned the VGA output capability, apparently.

Can't you use a DVI to VGA adapter? And besides what is so great about VGA anyway?

 

[Madpacket]

Can't you use a DVI to VGA adapter? And besides what is so great about VGA anyway?

FW900

 

[mohit9206]

I'm expecting $150 for 2GB and $180 for 4GB and GTX 960 performance.
Problem is being $180 puts it way too close to RX470 4GB and 1060 3GB for way less performance.

 

[antihelten]

That's because the RX 470 is the competitor of the 1050.... not the 460. And the competitor of the 460 is supposed to be the GT 1040 and I see nVIDIA releasing it along the 1030 as the bottom of the barrell.

The RX 470 is not the competitor of the the 1050 if the leaked specs are correct.

If shader performance is the main bottleneck then the GTX 1050 will be sitting at roughly 59% in the below graph, and if memory bandwidth is the main bottleneck then it will be sitting at 71%. Either way it's nowhere near an RX 470 (100%). Basically it will be equal to a 4GB RX 460 in the worst case scenario and a GTX 960 in the best case scenario, but either way it will be a hell of a lot closer to the RX 460 than the RX 470.

PS. The GTX 1060 6GB is at 121%.

I'm expecting $150 for 2GB and $180 for 4GB and GTX 960 performance.
Problem is being $180 puts it way too close to RX470 4GB and 1060 3GB for way less performance.

$180 would put it way too close to the RX 470 as you mention, and when the RX 470 is 40-70% faster that simply isn't tenable. I would expect the 4GB to be $150, and the 2GB to be $120-130. That would make it $10-20 more expensive than the corresponding 2 and 4GB RX 460 models, but also 0-25% faster, so Nvidia could probably get away with that.