Geforce GTX 1050 / 1050 Ti Thread

[Arachnotronic]

FW900

This monitor? https://www.amazon.com/Sony-GDM-FW9...00004YNSR/ref=cm_cr_arp_d_product_top?ie=UTF8

 

[whm1974]

This monitor? https://www.amazon.com/Sony-GDM-FW9...00004YNSR/ref=cm_cr_arp_d_product_top?ie=UTF8

Why in hell would someone spend almost $4000 on a CRT?

 

[Arachnotronic]

Why in hell would someone spend almost $4000 on a CRT?

Dat infinite contrast ratio & extremely fast response time!

 

[Bacon1]

Why in hell would someone spend almost $4000 on a CRT?

Its a super old monitor, was much more reasonable back in the day . Same reason you see other old computer parts for crazy prices these days that aren't worth it at all.

 

[crisium]

Looks good, besides the low clocks for Pascal. But if it keeps it in the no-pin <75W spec I won't complain. I'm sure we'll see 6-pin versions with a healthy factory OC. A version that can boost to where the other Pascal cards go will definitely beat the 960, but I really doubt these "leaked" default boosts will. Still, I'm looking forward to have a new pinless king launch.

 

[antihelten]

Looks good, besides the low clocks for Pascal.

It's interesting to note that using the normal GPC setup with 5 SMs per GPC and thus 640 cores at the "normal" Pascal frequency of 1.7GHz would have almost the exact same performance as the 768 cores setup at 1.38GHz that they went with. Nvidia most likely elected to go wide and slow due to better efficiency.

Of course this does open up the rather interesting question of whether or not any of the AIB partners will make a model with the necessary power setup to allow the card to clock up to the normal 1.7-1.8 GHz. If so, it could turn out to be the best overclocker of the generation (although it may still be memory bandwidth bottlenecked).

 

[Sweepr]

Nvidia sees tight graphics card supply; to release GTX 1050 in late September

Nvidia's Pascal-based GeForce GTX 1080, 1070 and 1060 graphics cards are seeing strong demand from the market and therefore are currently in tight supply. To further expand into the market, Nvidia is planning to release its mid-range GTX 1050 graphics card at the end of September at the earliest.

Graphics card players including Asustek Computer, Gigabyte Technology, Micro-Star International (MSI) and Colorful have all been aggressively striving for more GPU supply from Nvidia and their profits from the graphics cards and shipments are both expected to reach new highs in the third quarter.

http://www.digitimes.com/news/a20160909PD209.html

 

[sm625]

Given how efficient pascal gpus are, you would think they should be able to pack more than 768 into a 75W envelope. Kind of disappointing... but with no real competition I guess they can do whatever they want.

 

[nathanddrews]

FW900

This is the exact reason I returned my R9 290 and got the GTX 970 instead. 1920x1200@96Hz, zero input lag, great calibration.

Why in hell would someone spend almost $4000 on a CRT?

I'm not sure that anyone would, but they often sell for around the $500 mark. I bought my first when they were still in production, but the most recent one I bought was off a friend for $50.

Modern fixed-pixel displays are truly getting better - there's no denying the advantages of LCD monitors when it comes to weight, heat, power, glare, and geometric uniformity. Likewise, some higher end LCD monitors can rival the FW900 in contrast or color and usually far exceed it in brightness. The advent of adaptive sync also makes them more attractive than CRT in some situations, whereas a CRT can offer an extremely wide range of resolutions and refresh rates to statically adapt to any game/application/content.

IMO digital displays just need that final push to finally kill off CRT - input lag and motion blur. OLED could be that technology, but so far it has failed to deliver on that front. We really need OLED monitors suited for PCs/gaming instead of movies and TV.

Wow, that was super OT, sorry.
-----------------------------------------------------
I came in here to talk about the 1050. I kinda hope its performance is worse than the 460 so that it will be priced lower, but it's likely that I'll end up using it to power my next HTPC build, assuming that it covers all the bases that its bigger brothers do (DP1.4/HDMI2.0/HDR/PlayReady3.0).

An undervolted, passively cooled, single slot, low-profile version would be tempting at any price...

 

[Madpacket]

This monitor? https://www.amazon.com/Sony-GDM-FW9...00004YNSR/ref=cm_cr_arp_d_product_top?ie=UTF8

Yes that's it. I still use mine (paid $50.00) for competitive FPS. As much as I like my 144Hz Asus monitor it still pales when it comes to responsiveness compared to a CRT.

 

[tviceman]

Of course this does open up the rather interesting question of whether or not any of the AIB partners will make a model with the necessary power setup to allow the card to clock up to the normal 1.7-1.8 GHz. If so, it could turn out to be the best overclocker of the generation (although it may still be memory bandwidth bottlenecked).

7ghz memory speeds would yield 112 GB/s, almost 60% the bandwidth of GTX 1060, right in line with the shader count reduction. Given that GP107 has a slightly higher ROP / GPC ratio, clock for clock it should perform ever so slightly higher than 60% of GTX1060.

Perhaps Nvidia will plug a higher TDP based GP107 as the GTX 1050 TI when supply allows for it.

 

[Sweepr]

New details. Two models, 1050 and 1050 Ti.

1050 = cut down GP107 with 2GB RAM, late October launch, $119
1050 Ti = full GP107 with 4GB RAM, mid October launch, $149

http://www.hwbattle.com/bbs/board.php?bo_table=news&wr_id=50728

 

[tviceman]

I'll be surprised if the actual MSRP is $149 and $119. The RX 460 4gb is $140, and GP107's specs put it on par or above the GTX 960 - performance levels that easily trounce the RX 460 by ~20%. I'm betting the MSRP will be $159 and $129 respectively.

If Nvidia allows board partners to go above 75w in some custom variants, this little chip will have massive headroom.

 

[antihelten]

I'll be surprised if the actual MSRP is $149 and $119. The RX 460 4gb is $140, and GP107's specs put it on par or above the GTX 960 - performance levels that easily trounce the RX 460 by ~20%. I'm betting the MSRP will be $159 and $129 respectively.

If Nvidia allows board partners to go above 75w in some custom variants, this little chip will have massive headroom.

To be fair the RX 460 4GB is kind of overpriced as is. If it dropped down to ~$125 (to maintain a comparable perf/$ to the 1050 Ti at $150), then one might argue that that is really were it ought to have been from the start (the 2GB version by extension could drop down to $100).

 

[Arachnotronic]

I'll be surprised if the actual MSRP is $149 and $119. The RX 460 4gb is $140, and GP107's specs put it on par or above the GTX 960 - performance levels that easily trounce the RX 460 by ~20%. I'm betting the MSRP will be $159 and $129 respectively.

If Nvidia allows board partners to go above 75w in some custom variants, this little chip will have massive headroom.

AMD has been talking about how it wants to grab tons of market share from NVIDIA. Releasing faster, more efficient products at an equivalent price is probably a good way to try to stop that. GP107 should also have a good cost structure -- small die, narrow memory bus, etc. -- so it's not like NVIDIA needs to hurt its margins to hit these price points.

tl;dr -- they look like really good products that should be extremely competitive.

 

[DamZe]

Anything with a 128-bit bus in 2016 is DOA, I don't care if its an AMD or nVIDIA product. Save up for the 192+ bit cards.

 

[jpiniero]

Anything with a 128-bit bus in 2016 is DOA, I don't care if its an AMD or nVIDIA product. Save up for the 192+ bit cards.

It would be the fastest card you can get (at retail at least) that doesn't need a power connector though.

 

[tviceman]

Anything with a 128-bit bus in 2016 is DOA, I don't care if its an AMD or nVIDIA product. Save up for the 192+ bit cards.

Oh, you're a paper specs kind of person? I'll take real world performance metrics over die size, bus width, and mhz numbers any day of the year.

 

[DamZe]

Oh, you're a paper specs kind of person? I'll take real world performance metrics over die size, bus width, and mhz numbers any day of the year.

Me, oh no... I am the type of guy who owned a HD 5770, build rigs for friends who insisted on 960s, and thinks that discrete 128-bit video cards should be a thing of the past. Remember the GTX650Ti? It was subpar, the 192-bit BOOST variant was not. We see the same problems today, 128-bit is simply not ideal, I understand that it can help drive down power consumption, but it will definitely be the bottleneck for the 1050Ti, same as it was for the 650Ti.

 

[tviceman]

Me, oh no... I am the type of guy who owned a HD 5770, build rigs for friends who insisted on 960s, and thinks that discrete 128-bit video cards should be a thing of the past. Remember the GTX650Ti? It was subpar, the 192-bit BOOST variant was not. We see the same problems today, 128-bit is simply not ideal, I understand that it can help drive down power consumption, but it will definitely be the bottleneck for the 1050Ti, same as it was for the 650Ti.

Work the specs out and you'll see that math does not support what you say. With 7ghz vram on a 128-bit bus it will have 58% of the bandwidth of GP106 (GTX 1060 6gb). It so happens to have 60% the number of cores and ROPs but at a slower frequency, equating to about 55-57% of the total shader and throughput of the GTX 1060.

So in actuality, unless GTX 1060 has a significant bandwidth bottle neck right now - which it does not - the gtx 1050 TI will not either since they're the same architecture and scaled nearly 1:1 in specific ratios.

Also, considering that it has 25% the shader and ROP throughput of the GTX 1080 (based on core count and clock speeds) and 30% of the bandwidth, it's got more bandwidth available to its disposal relative to it's capabilities than the GTX 1080.

 

[frozentundra123456]

Oh, you're a paper specs kind of person? I'll take real world performance metrics over die size, bus width, and mhz numbers any day of the year.

Yep, seems like a lot of arbitrary decision making, either before any real data, or ignoring what we have already.

 

[DamZe]

Work the specs out and you'll see that math does not support what you say. With 7ghz vram on a 128-bit bus it will have 58% of the bandwidth of GP106 (GTX 1060 6gb). It so happens to have 60% the number of cores and ROPs but at a slower frequency, equating to about 55-57% of the total shader and throughput of the GTX 1060.

So in actuality, unless GTX 1060 has a significant bandwidth bottle neck right now - which it does not - the gtx 1050 TI will not either since they're the same architecture and scaled nearly 1:1 in specific ratios.

Also, considering that it has 25% the shader and ROP throughput of the GTX 1080 (based on core count and clock speeds) and 30% of the bandwidth, it's got more bandwidth available to its disposal relative to it's capabilities than the GTX 1080.

That may well be true, this does not seem to resemble the GTX650Ti/BOOST scenario I referred to earlier. The BOOST 650Ti variant did get a 33% increase in ROPs, 600mhz increase to the memory + the 192-bit bus over the standard 650Ti.

The 1050Ti is basically a modern 960 as far as bandwidth is concerned (give or take Pascal's DCC), with a slightly bumped up texture fill rate due to higher clock speeds. But for 50$ more you get the 1060Gimp3 edition with twice the Tflops, it's still a tough sell for people with a spare 6-pin PCI-E connector, except for HTPC users who want the tiny TDP footprint.

 

[tviceman]

That may well be true, this does not seem to resemble the GTX650Ti/BOOST scenario I referred to earlier. The BOOST 650Ti variant did get a 33% increase in ROPs, 600mhz increase to the memory + the 192-bit bus over the standard 650Ti.

The 1050Ti is basically a modern 960 as far as bandwidth is concerned (give or take Pascal's DCC), with a slightly bumped up texture fill rate due to higher clock speeds. But for 50$ you get the 1060Gimp3 edition with twice the Tflops, it's still a tough sell for people with a spare 6-pin PCI-E connector, expect for HTPC users who want the tiny TDP footprint.

I'm not arguing the value proposition - the low end cards this round seem to have absolutely terrible value vs. the $200-250 price point. The only thing I was doing was attempting to show you that the bandwidth available (bus size and memory speed) to it's theoretical performance is right in line, if not slightly better, than other Pascal models.

I don't think you'll ever see me recommending the 3gb 1060 version, though. Even if it's speed is more than capable for the next 18 months at 1080p, its limited vram size will become an issue before then. I'd recommend the RX 480 4gb at the same price over the 1060 3gb, while recommending the 1060 6gb over the RX 480 8gb.

 

[mohit9206]

1060 3GB is actually 1050Ti
1050Ti actually is 1050
1050 is actually 1040
Renaming it this way allows them to mark up the prices.This is what they have done.

 

[Sweepr]

It would be the fastest card you can get (at retail at least) that doesn't need a power connector though.

Indeed, and that's a strong selling point for these cards.

No doubt we will also see factory OCed versions with 6-pin as well.