MSI GeForce GTX 1050 Ti GAMING X vs AMD Radeon RX 470 @ [H]

This is a fantastic result for anyone looking to buy a sub $200 GPU, but I can't help but feel like its all still the result of AMD being tangibly behind Nvidia in some critical ways, and I'm not sure this "competition" is as great as it looks on the surface. Is anyone else here just a little bit confused about what exactly is going on with the GPU market, especially between $100-$300? I can't help but feel like AMD's strategy and chip configurations seem a bit odd, particularly Polaris 11, which is really far behind Polaris 10. Either the explanation is that Polaris 11 is really about trying to win back some mobile designs, or Polaris in general isn't doing what they'd hoped.

Compare die size of the different chips and their relative stack position (note, not saying all examples compete with each other)

AMD (mm2, transistors)
460 (123, 3B )
470 (232, 5.7B)
480 (232, 5.7B)

NV (mm2, transistor)
1050/Ti (135, 3.3B)
1060 6/3GB (200, 4.4B)
1070 (314, 7.2B)

The 470 beats the 1050 by 40-60+%. It also does so at 72% more die area and transistors. To expect any result other than a walkover by AMD would have just been blind fanboyism, so I'm not sure why people are so surprised. The fact that Nvidia forced the conversation and a reaction from AMD with a chip thats about half the size (and therefore half the cost to fab) means it doesn't seem like an unqualified success for AMD to me though. The most direct comparison is the one [H] did originally (460 vs 1050) and its not even close, which means that AMD has to sell more expensive chips cheap to pick up the slack and protect the only market where they probably have a real chance to gain share. RX480 vs 1060 is a tossup based on performance, but again AMD needs tangibly more die and transistors to do it. Not that these things are the be-all end-all of "whos winning" but AMD's repeated inability to deliver gaming performance with equivalent designs is going to keep hurting them in the long run. They're providing value to customers in the short term by giving them a nice performance premium for the price, but they're doing it at a disproportionate cost to themselves. Is this what we can expect out of Vega? A GPU that competes with the 1070 and 1080 but is 400mm or 9+B transistors on a brand new process? Or one that is the same size as a 1070 and is 20% slower?

Still, great job on the review Brent, its nice to see you guys turn it around so quickly to silence the haters.. I'm sure it took a lot of hard work. The 470 is definitely intriguing and attractive for the price, I just wish it used a little less power or else I'd be considering putting it in a living room gaming PC that currently has a 750Ti in it. As it stands the 1050Ti will probably end up being that replacement.
 
Last edited:
This is a fantastic result for anyone looking to buy a sub $200 GPU, but I can't help but feel like its all still the result of AMD being tangibly behind Nvidia in some critical ways, and I'm not sure this "competition" is as great as it looks on the surface. Is anyone else here just a little bit confused about what exactly is going on with the GPU market, especially between $100-$300? I can't help but feel like AMD's strategy and chip configurations seem a bit odd, particularly Polaris 11, which is really far behind Polaris 10. Either the explanation is that Polaris 11 is really about trying to win back some mobile designs, or Polaris in general isn't doing what they'd hoped.

Compare die size of the different chips and their relative stack position (note, not saying all examples compete with each other)

AMD (mm2, transistors)
460 (123, 3B )
470 (232, 5.7B)
480 (232, 5.7B)

NV (mm2, transistor)
1050/Ti (135, 3.3B)
1060 6/3GB (200, 4.4B)
1070 (314, 7.2B)

The 470 beats the 1050 by 40-60+%. It also does so at 72% more die area and transistors. To expect any result other than a walkover by AMD would have just been blind fanboyism, so I'm not sure why people are so surprised. The fact that Nvidia forced the conversation and a reaction from AMD with a chip thats about half the size (and therefore half the cost to fab) means it doesn't seem like an unqualified success for AMD to me though. The most direct comparison is the one [H] did originally (460 vs 1050) and its not even close, which means that AMD has to sell more expensive chips cheap to pick up the slack and protect the only market where they probably have a real chance to gain share. RX480 vs 1060 is a tossup based on performance, but again AMD needs tangibly more die and transistors to do it. Not that these things are the be-all end-all of "whos winning" but AMD's repeated inability to deliver gaming performance with equivalent designs is going to keep hurting them in the long run. They're providing value to customers in the short term by giving them a nice performance premium for the price, but they're doing it at a disproportionate cost to themselves. Is this what we can expect out of Vega? A GPU that competes with the 1070 and 1080 but is 400mm or 9+B transistors on a brand new process? Or one that is the same size as a 1070 and is 20% slower?

Still, great job on the review Brent, its nice to see you guys turn it around so quickly to silence the haters.. I'm sure it took a lot of hard work. The 470 is definitely intriguing and attractive for the price, I just wish it used a little less power or else I'd be considering putting it in a living room gaming PC that currently has a 750Ti in it. As it stands the 1050Ti will probably end up being that replacement.
We just have to see how Vega shakes out - if a rather new design it can have some nice surprises or not.

Agree that Pascal perf/Die size and perf/w is much better than AMD's - hence Nvidia will probably make more money off of each sell. I think P11 is in trouble of even being relevant - its selling price is way too high. Can AMD make a good enough profit cutting pricing on the 460? Current pricing on the 1050 is barely sufficient against a lower priced 470 - you get a hell a lot more performance percentage wise with the 470 then cost percentage increase so even the 1050 is being sold too high and should come down. That is if AMD can keep the 470 low.
 
In fact, here's a RX480 for $180 AR.

Get that green team bearded flannel hipster-vegan suspender-wearin' weaksauce outta here.




;)

P.s. the 470 needs a price drop if AMD is gonna be selling 480s (even if only 4GB variant) for ~$200. Seeing too many 470s @ $200.
 
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
The majority of end users dont care how many transistors it has or die area. They care about the price of it and the performance it gives. It is too bad AMD wasnt able to push full P11 out with the 16CU's to compete head on with the 1050Ti, but that 75W ceiling was too close for comfort i imagine as a result they really left a large gap between the 460 and the 470. 1050Ti fills that perfectly, good move on Nvidia's part. 470 is also a salvaged P10, so part of the die area is inactive, performance comparisons based on die size should be using the 480 as the benchmark shouldnt they as those are both fully enabled chips. Now cost per chip is something AMD is getting beat on for sure here. Not sure if that is made up in the $30 price premium or not. It really is too bad a RX465 didnt see the light of day, most likely viable for the respin thats been talked about. Great to have competition here though, keeps em both honest and really pushing performance in entry level and middle segments. Looking forward to Vega and renewed competition at the high end!
 
The majority of end users dont care how many transistors it has or die area. They care about the price of it and the performance it gives. It is too bad AMD wasnt able to push full P11 out with the 16CU's to compete head on with the 1050Ti, but that 75W ceiling was too close for comfort i imagine as a result they really left a large gap between the 460 and the 470. 1050Ti fills that perfectly, good move on Nvidia's part. 470 is also a salvaged P10, so part of the die area is inactive, performance comparisons based on die size should be using the 480 as the benchmark shouldnt they as those are both fully enabled chips. Now cost per chip is something AMD is getting beat on for sure here. Not sure if that is made up in the $30 price premium or not. It really is too bad a RX465 didnt see the light of day, most likely viable for the respin thats been talked about. Great to have competition here though, keeps em both honest and really pushing performance in entry level and middle segments. Looking forward to Vega and renewed competition at the high end!

To the end user, you're right it doesn't matter, and I alluded to that point in my post. I was speaking more to the fact that AMD is spending more money to deliver the same performance and what it means for their health as a company moving forward. It also speaks a bit to their ability to create efficient designs from a power and performance perspective. The 470 being a salvage part does mean that it at least provides them a way to recoup costs on bad dies, but when you look at the relative position of the 480 and the 470, they seem a bit too close together, which indicates to me that they may have hoped the 480 would be further up market. Again, for people actually buying GPU's looking to get the most performance per dollar, the situation is great, but poor margins aren't the route to success especially in a business that requires substantial re-investment into R&D to stay relevant.
 
Here in my neck of the woods, the most expensive 1050ti is $250 CDN (avg $200 CDN) and that's where a few 470's have their price after 10-15 dollar rebates. You'd have to be totally crazy to buy a 1050ti at 200 over a 470 at 250!
 
Interesting, I thought the 1050 cards were lackluster when they came out (although the RX 460 is even more disappointing), if you can actually get an RX470 for an extra $30 I can see a lot of people deciding they can skip a few McDonald's meals to get a better card.
 
Great review. Really good to see the competition in this bracket is hot, wish it was the same at the top end.
 
Those 470's are pretty good for the price and absolutely stomping the 1050ti in comparison. I was really expecting the 1050 series cards would be closer in performance to the 1060's. There's a big gap and the 470 is filling it well for only about $30 more.
 
Interesting, I thought the 1050 cards were lackluster when they came out (although the RX 460 is even more disappointing), if you can actually get an RX470 for an extra $30 I can see a lot of people deciding they can skip a few McDonald's meals to get a better card.
Yeah agreed.
But this scenario also plays out with the 460, if you can stretch go for the 1050ti single fan models with no auxiliary connector.
So I think the 470 when it has competitive pricing is stealing sales from the 1050ti, but then I think the 1050/1050ti are stealing sales from the 460 2GB and 4GB.

Worth remembering that all 1050ti are currently locked to 75W cap even with the auxiliary connector no matter what one tries with the OC software, so there may be a bit more performance to come from the model but then that needs to be balanced with at what point does it bottleneck due to its spec/BW.

Cheers
 
Here in my neck of the woods, the most expensive 1050ti is $250 CDN (avg $200 CDN) and that's where a few 470's have their price after 10-15 dollar rebates. You'd have to be totally crazy to buy a 1050ti at 200 over a 470 at 250!

It's not that simple. If the PSU in your PC can't cope you need to spend extra to replace it. So instead of an extra $50, you're looking at an extra $100 or more.
 
AMD using a bigger die is obviously not by choice. Process issues and they don't have the cash reserves to make a separate chip for each card like Nvidia.
Yes it's not as profitable, but that's not an issue for you, the consumer.
And lets be real, making an issue about 50-60W when a majority of gaming or casual end users have a normal 350-400WPSU, is really grasping for straws. It's okay to admit AMD has won the low end enthusiast market currently in this battle (looking at LeldraPrime1 and co here). Don't always have to shit on everything, I swear if Vega comes out and is the absolute bees knees, you'd bitch about the colour, or the fans being 2db louder than a reference 1080 like it's the end of the damn world.
 
AMD using a bigger die is obviously not by choice. Process issues and they don't have the cash reserves to make a separate chip for each card like Nvidia.
Yes it's not as profitable, but that's not an issue for you, the consumer.
And lets be real, making an issue about 50-60W when a majority of gaming or casual end users have a normal 350-400WPSU, is really grasping for straws. It's okay to admit AMD has won the low end enthusiast market currently in this battle (looking at LeldraPrime1 and co here). Don't always have to shit on everything, I swear if Vega comes out and is the absolute bees knees, you'd bitch about the colour, or the fans being 2db louder than a reference 1080 like it's the end of the damn world.

The technical issue is more than just general 50-60W, it is also those who are on a budget or with say older budget GPUs do not necessarily have the option or even want to consider the aspects of auxiliary power connectors, quite a few of these want to plug into the PCIe slot only and know it will work without any problems, while also giving the performance increase promised.
A more likely situation is that a lower price 470 will eat into 1050ti sales, but the 1050 and 1050ti will be taking sales from the 460 (definitely 2gb vs 2gb and similar scenario but possibly less numbers stretching to 1050ti) and core Nvidia client base or those that cannot stretch to the lower price 470/challenges using auxiliary connector will still buy the 1050ti.
 
Last edited:
It's not that simple. If the PSU in your PC can't cope you need to spend extra to replace it. So instead of an extra $50, you're looking at an extra $100 or more.

How many systems out there (OEM's) don't come with at least one six pin PCI-E power connector if the system ISN'T using the onboard graphics (AMD or Intel)? You can bet money that the PSU is in the 350 to 450 range with an extra PCI-E connector. The PSU is not a factor for the majority of people out there.
 
How many systems out there (OEM's) don't come with at least one six pin PCI-E power connector if the system ISN'T using the onboard graphics (AMD or Intel)?

But that's just it. Many people do use the onboard graphics or bus-powered GPUs. Their PCs don't come with high-powered PSUs. They're often second-hand office PCs. Check out the Steam hardware survey some time. Note that 6 of the top 10 GPUs - 7 if you include the GTX 960M - require no additional power and 4 of that 6 are Intel onboard graphics. Overall, 18% of PC gamers use Intel graphics. That 18% is the main target market for this card.
 
i really had no idea the 470 was so much more powerful a card. (course it cost more lol)
 
But that's just it. Many people do use the onboard graphics or bus-powered GPUs. Their PCs don't come with high-powered PSUs. They're often second-hand office PCs. Check out the Steam hardware survey some time. Note that 6 of the top 10 GPUs - 7 if you include the GTX 960M - require no additional power and 4 of that 6 are Intel onboard graphics. Overall, 18% of PC gamers use Intel graphics. That 18% is the main target market for this card.
actually I think it was said that steam will read both the iGPU and dGPU therefore the numbers may be a bit skewed. It would be cool if it would query hardware for each game or just one like FO4. See what hardware is playing that game. Some of those in the survey may only be playing low system required games.
 
actually I think it was said that steam will read both the iGPU and dGPU therefore the numbers may be a bit skewed.

If that were the case, would we not see far higher numbers for Intel and a congruence between CPUs and GPUs?
 
i really had no idea the 470 was so much more powerful a card. (course it cost more lol)
Also worth noting all 1050ti are currently capped at 75w even if they have an auxiliary power connector and someone tries OC, but we will have to wait and see how much difference that can make in terms of performance once there is a bit more flexibility in terms of power management and clocks (especially memory).
So that is a 75W 1050ti cards (until 75W limit is removed from mainboard+auxiliary) against 144W 470 in same game for both.
Such a figure does not mean much to most consumers, especially here but worthy of note from a technical point.

Cheers
 
Last edited:
So in some reviews I've read the 1050ti was getting higher minimum fps than the 1060, what's the deal there?
 
I think we should note that custom designs are bordering on overkill for a card that draws 75W. You can get about as much mileage from the cheapest $130-140 1050 as you would the more expensive ones, so the pricing is more in the 1050ti's favor.
 
I think we should note that custom designs are bordering on overkill for a card that draws 75W. You can get about as much mileage from the cheapest $130-140 1050 as you would the more expensive ones, so the pricing is more in the 1050ti's favor.
Did you see the performance difference? Pricing isn't the issue as the 1050s seem locked to a max performance level (from what we have seen so far).
 
Did you see the performance difference? Pricing isn't the issue as the 1050s seem locked to a max performance level (from what we have seen so far).
I misspoke, I meant to say it slightly evens the scales, though the price-to-performance is still certainly more in the 470's favor, if we're looking at that, which you are right about.

That said, what I maintain is that there is a very real $30 or $40 difference (more if we're looking at custom 470s) between the reference 1050ti (which is negligibly different from custom) and custom 470s (where custom makes more of a difference because of the increased power draw and heat), which is not an insignificant amount for someone buying cards in this price range. All this is mostly in reference to the [H] review's conclusion on pricing, BTW.
 
How many systems out there (OEM's) don't come with at least one six pin PCI-E power connector if the system ISN'T using the onboard graphics (AMD or Intel)? You can bet money that the PSU is in the 350 to 450 range with an extra PCI-E connector. The PSU is not a factor for the majority of people out there.

The fact that 750 Ti is still the #3 GPU on Steam proves, for whatever reason, there's a sizeable market for a <75W TDP card, and NV is able to charge a premium for having the fastest card in this category. 470 being faster is a moot point for this segment since it doesn't satisfy the TDP constraint to begin with, and as AMD: Activate mentioned AMD's "win" in perf/price comes at a huge BoM cost.
 
Back
Top