Geforce GTX 1060 Thread: faster than RX 480, 120W, $249

[Sweepr]

This is a thread about the 1060. Keep your disdain for AMD out of here and stop trying to start an Nvidia vs AMD argument.

User Bacon1 was extremely surprised about the fact that AIBs can release overspecced mainstream VGAs with 8-pin / 2x 6-pin. I'm just pointing out that's not exclusive to NVIDIA and doesn't imply less than expected efficiency.

 

[nurturedhate]

User Bacon1 was extremely surprised about the fact that AIBs can release overspecced mainstream VGAs with 8-pin / 2x 6-pin. I'm just pointing out that's not exclusive to NVIDIA and doesn't imply less than expected efficiency.

Amazing how we already saw pre-order for an extra $10, right? I'm sure AMD was hoping for $299+ and several months advantage to market.

And this post? It would appear to most that you have an issue with certain posters and AMD, you keep bringing them up in a specific manner as quoted above. Again, this thread is about the 1060, not AMD.

 

[Bacon1]

Let's ignore the fact that Sapphire Nitro RX 480 is listed at 175W TDP with a minor OC and there's plenty of overspecced models on AMD's side as well. This is obviously some sort of conspiracy from NVIDIA.

These are listed as 120/150w, not 175+, so why do they need 8pin or dual 6 pin? That is what I'm asking. I mean even the 1070 only has a single 8pin, so why would a much more efficient perf/watt card need the same, or dual 6?

 

[jpiniero]

These are listed as 120/150w, not 175+, so why do they need 8pin or dual 6 pin?

Overclocking.

The two Inno3D cards are probably using the same board so it would make sense they would both use 8-pin, since 150W on 6-pin is pushing it.

 

[Sweepr]

Again, this thread is about the 1060, not AMD.

If AMD knew 6GB GTX 1060 would cost $249 and deliver GTX 980 performance (or close) at 120W TDP in advance, I think they would try to undercut NVIDIA even more ($199-219 for 8GB RX 480).

These are listed as 120/150w, not 175+, so why do they need 8pin or dual 6 pin? That is what I'm asking. I mean even the 1070 only has a single 8pin, so why would a much more efficient perf/watt card need the same, or dual 6?

Zero surprise if AIB (factory overclocked) models go past 150W with extra OC, but 2x 6 pin does sound overspecced, just like some crazy cooling solutions we're seeing.

 

[Bacon1]

If AMD knew 6GB GTX 1060 would cost $249 and deliver GTX 980 performance (or close) at 120W TDP in advance, I think they would try to undercut NVIDIA even more ($199-219 for 8GB RX 480).

Why? 480 is already doing that but 8GB @ $239 or 4GB @ 199. ~4% slower than 980 @ 1266

Just think its odd there are so many different TDP's for the 1060 and some with more power plugs than the 1070.

 

[Sweepr]

Why? 480 is already doing that but 8GB @ $239 or 4GB @ 199. ~4% slower than 980 @ 1266

That's nearly the same perf/$ (vs 8GB model) if 6GB GTX 1060 matches the GTX 980. Coupled with their stronger brand and maybe some positive launch reviews, it means it will steal the thunder from its closest competitor.

Just think its odd there are so many different TDP's for the 1060 and some with more power plugs than the 1070.

Higher TDP for custom cards at higher clocks is no surprise. I don't think GP106 will match GP104 perf/watt, but you can't draw any meaningful conclusions about its efficiency based on AIB choices.

 

[Silverforce11]

It's better to have extra power pins than it's needed, rather than just have barely enough.

And yes, AIBs love to emphasize VRMs & pin count, it's the new epeen.

 

[Sweepr]

New rumor about the 3GB model. Grain of salt required:

Geforce GTX 1060 3GB to cost $149 (AIB) / $199 (FE)?

If you’re willing to take a VRAM cut however, Nvidia is reportedly targeting an impressive $149 price point for partner cards, while its GTX 1060 3GB Founders Edition will price match the RX 480 4GB at $199.

...The 3GB VRAM indicates a 192-bit memory bus once again, while it’ll draw on slightly less power than its bigger brother at 100W TDP, 20W less than the GTX 1060 6GB.

www.game-debate.com/news/20782/3gb-geforce-gtx-1060-incoming-rumoured-to-cost-just-149


Zotac's Geforce GTX 1060 X-Gaming OC

1569-1784 / 8008MHz
heat pipe silent dual fan
4 + 1 phase power
a total length of 22.9cm
120W
the PWM automatic speed control
retail price: 1999 yuan

http://weibo.com/1804024624/DELpDB5Bh?type=comment#_rnd1468624604460

Some good news from Zotac, this custom OCed model costs the same as RX 480 8GB (reference) in China. They confirmed it's the 6GB model in the comments section.

It's better to have extra power pins than it's needed, rather than just have barely enough.

And yes, AIBs love to emphasize VRMs & pin count, it's the new epeen.

True, it's almost comical. Zotac's Geforce GTX 1060 'Supreme Plus OC' for example (2x 8-pin):

 

[Arachnotronic]

It's better to have extra power pins than it's needed, rather than just have barely enough.

And yes, AIBs love to emphasize VRMs & pin count, it's the new epeen.

It's honestly pretty wasteful. It adds additional cost without delivering any additional performance

The extra cost should go into whisper quiet coolers, IMO.

 

[Bacon1]

2x 8 pin??? Thats the same as Fury and 980 TI lol

Are there going to be many different core counts or something? I can't see how one chip will hit all these price points.

 

[Sweepr]

http://videocardz.com/62278/nvidia-geforce-gtx-1060-rumors-part-7-new-cards-more-benchmarks


List of some custom models announced so far:

Nvidia brings Pascal graphics chips for Mobile in August

Mobile Geforce GTX 1060
1280 CUDA cores
6GB GDDR5 192-bit
65W TDP

www.golem.de/news/notebooks-nvidia-bringt-pascal-grafikchips-fuer-mobile-im-august-1607-122153.html

 

[hsjj3]

Guys, about the pins. 120W is sufficient for the base/reference models. Then with a factory OC etc you need more. But, with my experience/research on GTX 960, not much.

My model is 128W power limit (EVGA 960 SC 4GB). I can run at 1541MHz + 7400MHz core. However, the power limit is hit in most games/benchmarks. For example, in Firestrike it downclocks 2-3 clock cycles due to power constraints. At stock settings, it downclocks 1-2 clock cycles as well.

However, this may not be a good example because the GTX 960 was initially rated for 2GB, and the 2GB extra for the 4GB models would have definitely increase power on their own by 5W.

Still, I have seen the performance of some GTX 960 like EVGA FTW (most extreme) that comes with 160W power limit, and here in heavy benchmarking use (minus Furmark, that virus!) it tops out at about 85% (i.e. 136W). So yeah, just because the listed power rating is higher, it's mostly headroom that never gets used up unless you run the Furmark virus.

EDIT: As can be seen with the current FE models for Pascals, they OC just as well vs the aftermarkets that come with 2x8pins for example for the GTX 1070. These extra pins really are just marketing gimmicks to fool the unsuspecting. Sure, at the high end it means the core clocks don't get throttled (even then usually just by a couple of boost cycles), but it's gimmicky as hell. I can understand an allowance for 20% extra power than the stock TDP, but anything above that is pure gimmicks.

 

[hsjj3]

[TDP] represents the maximum amount of power the cooling system in a computer is required to dissipate...... but the maximum average power (?) would draw when running "real applications".

Dude you're quite crafty. Nvidia didn't state any of that independently. That is quit clearly what they referred to as the Wikipedia definition. The parts they found relevant to them, they highlighted in RED.

On the next page, Nvidia stated THEIR OWN definition of TDP:

"TDP is a measure of maximum power draw over time in real world applications. It does not represent the maximum power draw in pathological cases such as Furmark. Please see [TDP wiki page] for more info."

http://cdn.videocardz.com/1/2016/07/NVIDIA-GeForce-GTX-1060-Reviewers-Guide-18_VC.jpg

Nvidia never on their own referred or highlighted the part about the cooling solution because that's not how they define TDP.

 

[zinfamous]

An FE 3GB 1060? I guess when you have legions of people that will buy it simply because it's green rather than pay the same price for a red 480 that will destroy it, or a red 470 that will beat it at up to $50 less cost, you can have the balls to release a piece of hardware that really has no business being around in 2016.

I mean, with low power draw, quiet, efficient, it is probably great for HTPCs and indy games if they can pump out a single slot short card for tiny cases. That strikes me as the perfect scenario for a 3gb 1060...but why not just get a better performing 4gb 470 and save up to $50 off the gauging FE edition?

 

[sze5003]

I saw some article about a gigabyte 1060 but it's tripple slot. Anyone know why it would need such cooling and take up that amount of space for a mid range card? I'm pretty sure gigabyte may do the same with a 480. But I've always wondered why if they are not super powerful to begin with.

 

[railven]

I saw some article about a gigabyte 1060 but it's tripple slot. Anyone know why it would need such cooling and take up that amount of space for a mid range card? I'm pretty sure gigabyte may do the same with a 480. But I've always wondered why if they are not super powerful to begin with.

If I had to guess, AIBs just mass produce x-style coolers and slap it on wherever it will fit.

I remember my friend bought an ASUS Direct CU II GTX 750!!! There was no way in hell that card needed that beefy cooler, but probably easier to use surplus that to design something else.

 

[hsjj3]

Damn, looks like EVGA just slapping its ACX 2.0 coolers on the GTX 1060. I thought we could see some improvement there...but no.

 

[antihelten]

On the next page, Nvidia stated THEIR OWN definition of TDP:

"TDP is a measure of maximum power draw over time in real world applications. It does not represent the maximum power draw in pathological cases such as Furmark. Please see [TDP wiki page] for more info."

http://cdn.videocardz.com/1/2016/07/NVIDIA-GeForce-GTX-1060-Reviewers-Guide-18_VC.jpg

Nvidia never on their own referred or highlighted the part about the cooling solution because that's not how they define TDP.

That is clearly not Nvidia's full definition of TDP. That sentence is simply an abbreviated explanation of TDP provided by Nvidia to enlighten underinformed reviewers. Most reviewers from larger sites (such as AT) will completely ignore that sentence, since they are already perfectly aware of the full and correct definition of TDP.

The definition of TDP is basically "maximum amount of power used by a chip (and thus dissipated) for a thermally significant period of time under normal operating conditions", and the whole purpose of defining this number (along with the thermal threshold) is to allow AIB partners to know how beefy the cooler needs to be. Any variation in measuring TDP between different companies derive from differences in how they define "thermally significant period of time" and "normal operating conditions".

So yes TDP absolutely pertains to cooling (hence the Thermal part of the name).

Bacon1 is of course wrong to argue that TDP is the maximum amount of heat a given cooling system can necessarily dissipate, since it is effectively the exact opposite (i.e. the minimum amount of heat a given cooling system must be able to dissipate). However as with all other things, AIB partners will try to avoid over engineering their cooling solutions since doing so can be costly (this is basically what the Wiki entry is explaining), of course an over engineered cooling solution can often serve as a selling point and as such be worth it to the AIB partner anyway.

 

[Erenhardt]

The definition of TDP is basically "maximum amount of power used by a chip (and thus dissipated) for a thermally significant period of time under normal operating conditions", and the whole purpose of defining this number (along with the thermal threshold) is to allow AIB partners to know how beefy the cooler needs to be. Any variation in measuring TDP between different companies derive from differences in how they define "thermally significant period of time" and "normal operating conditions".

I added some emphasis.

 

[positivedoppler]

I added some emphasis.

should that include memory and everything else on the board that generates heat?

 

[antihelten]

should that include memory and everything else on the board that generates heat?

That would be known as Total Board Power or TBP.

Note that memory chips (and other components) have TDP and thermal threshold values of their own that also needs to be taken into account when designing a cooling solution.

 

[hsjj3]

Well suit yourselves, Nvidia clearly highlighted the part that TDP is defined differently by different companies. If you search online, you will find that Nvidia has always rated TDP as the max power under high gaming load.

Also, if you bother to look into the BIOS of their reference cards, you will see the stock power limit correspond to the rated TDP.

In fact, AMD has stopped using TDP for their GPUs at least. If you look at the official RX 480 page and product documents, it clearly states "Board Power" to remove any and all doubt as to what it could mean.

Sure, logic from Nvidia's defition would extend to the cooling solution. If it is designed to run at 120W under heavy gaming load, the cooler will be designed to dissipate 120W worth of heat.

 

[antihelten]

Well suit yourselves, Nvidia clearly highlighted the part that TDP is defined differently by different companies. If you search online, you will find that Nvidia has always rated TDP as the max power under high gaming load.

Also, if you bother to look into the BIOS of their reference cards, you will see the stock power limit correspond to the rated TDP.

In fact, AMD has stopped using TDP for their GPUs at least. If you look at the official RX 480 page and product documents, it clearly states "Board Power" to remove any and all doubt as to what it could mean.

Sure, logic from Nvidia's defition would extend to the cooling solution. If it is designed to run at 120W under heavy gaming load, the cooler will be designed to dissipate 120W worth of heat.

Again Nvidia's short explanation is done for the benefit of reviewers, not to deliver their own formal definition of TDP.

Of course the limit in the BIOS corresponds to the TDP, why wouldn't it? If it's lower, then the TDP will also be lower. If it's higher, then the GPU can go above the TDP value, thus making the TDP value useless (since AIB partners can no longer rely on it when designing their cooling solution).

And AMD using TBP doesn't change anything (and you can be almost certain that they still report TDP numbers to their AIB partners), nor does it really remove all doubt (unless you can tell me exactly how AMD defines TBP).

Not only does Nvidia's definition of TDP extend to cooling solutions, the entire reason for Nvidia even defining a TDP value in the first place is for the benefit of the cooling solution (and properly designing it).

In other words, the entire purpose behind the TDP value is for it to serve as a guide to properly designing a cooling solution.

 

[tviceman]

http://videocardz.com/62278/nvidia-geforce-gtx-1060-rumors-part-7-new-cards-more-benchmarks


List of some custom models announced so far:

Nvidia brings Pascal graphics chips for Mobile in August

Mobile Geforce GTX 1060
1280 CUDA cores
6GB GDDR5 192-bit
65W TDP

www.golem.de/news/notebooks-nvidia-bringt-pascal-grafikchips-fuer-mobile-im-august-1607-122153.html

Coming in >10% above GTX 980. Of course Pascal does better in DX12 than Maxwell, but still boding well for this little chip. Equal to RX 480 in DX12, faster in DX11, 30-40 less watts in power consumption, and 30mm2 smaller.