You thought GTX 1060 was confusing? You've seen nothing. Enter RTX 2060 and 6 variants of the same card

[MysteriousAeon]

25 minutes ago, Master Disaster said:

Got a source on this. I've followed CEMU from pretty much day one, I've even had conversations with the developer (back before he left GBATemp) and I've never heard anyone ever suggest you run the emulator through WINE on Linux instead of Windows. Adding an extra layer of complexity to an emulator is never a good idea plus AMDs Linux drivers are terrible so why anyone would suggest running anything AMD based in Linux over Windows is beyond me.

 

OK so I did some googling before posting this and it turns out your partially correct. Using some custom AMD drivers on Linux its possible to get upto a 40% speed boost on CEMU but only if your GPU is weak and causing a bottleneck, if your GPU isn't bottlenecking then at best you see no difference. That said I can find nowhere the CEMU developer recommending Linux over Windows, the origins of this method come from Reddit.

Not sure if a i7-4790k bottlenecks a R9 390 but I had over 45 FPS in Linux while I was hardly getting 20 in Windows. It's not just about a bottleneck, Linux's AMD drivers are open-source so their OpenGL implementation is better overall.

 

The drivers I'm talking about are the Mesa ones. Not official.

[cj09beira]

this is the hardware version of a harrem 

[GoldenLag]

5 minutes ago, huilun02 said:

Of course this comes from non other than the same company that made GPP

And the 1050 and 1060 lineup

[Salv8 (sam)]

[leadeater]

5 hours ago, cj09beira said:

this is the hardware version of a harrem 

I dunno, you'd actually have want all those GPU SKUs for it to be that 

[S w a t s o n]

I see jensen is trying to give Intel a run for it's money on the sku count huh?

[mr moose]

I like the way every complains before we even know if any difference in the GPU's is just the memory controller or not.

 

More options aren't intrinsically evil unless they are intentionally obfuscating details and charging more. 

 

 

[leadeater]

7 minutes ago, mr moose said:

I like the way every complains before we even know if any difference in the GPU's is just the memory controller or not.

 

More options aren't intrinsically evil unless they are intentionally obfuscating details and charging more. 

I can't see why that many are necessary, having a few sure but 3 different models with the same memory type isn't going to reduce the cost that much so any cost difference is mostly going to be artificial or minimal so then why should it exist.

 

The GDDR5 vs GDDR6 models make sense as GDDR6 is obviously going to be more expensive but this is silly:

15 hours ago, NunoLava1998 said:

• RTX 2060 GDDR5X 3GB
• RTX 2060 GDDR5X 4GB
• RTX 2060 GDDR5X 6GB
• RTX 2060 GDDR6 3GB
• RTX 2060 GDDR6 4GB
• RTX 2060 GDDR6 6GB

 

I question the need for two 6GB variants but I guess having a cheaper higher end option is nice if the GDDR6 model is going to be that much more. But 3GB and 4GB of both options.... dumb.

 

If Nvidia really wants that many here's a crazy idea RTX 2060 = GDDR5 and RTX 2060 Ti = GDDR6.

[8uhbbhu8]

[aezakmi]

Where are the 3.5GB versions? 0/10 wouldn't get it

[Ashley MLP Fangirl]

i feel sorry for all of the nvidia fanboys who will need to memorize the whole thing. 

[SenKa]

Holy balls nVidia. I was already getting confused when you used to rebrand your last gen higher tier chips as budget ones for the next gen. Now this!? Thank god I haven't bought an nVidia card since the GTS 450

[Princess Luna]

It's pretty much a shit show by now.

[mr moose]

2 hours ago, leadeater said:

I can't see why that many are necessary, having a few sure but 3 different models with the same memory type isn't going to reduce the cost that much so any cost difference is mostly going to be artificial or minimal so then why should it exist.

 

The GDDR5 vs GDDR6 models make sense as GDDR6 is obviously going to be more expensive but this is silly:

 

I question the need for two 6GB variants but I guess having a cheaper higher end option is nice if the GDDR6 model is going to be that much more. But 3GB and 4GB of both options.... dumb.

 

Again, if there is no difference other than the amount/type of ram and that works out to give more options then why is it such an issue? 

Quote

If Nvidia really wants that many here's a crazy idea RTX 2060 = GDDR5 and RTX 2060 Ti = GDDR6.

 

But there is a GPU difference outside of ram on the Ti model.   

 

So long as the products are clearly marked and they aren't changing the number of cores like they did with the 1060 then I don't see it as a problem.

 

 

EDIT: and besides that, people are so busy looking for something to dump on they haven't even started arguing that the 2060 isn't big enough to actually make any use of the RTX:

 

[D13H4RD]

1 hour ago, mr moose said:

Again, if there is no difference other than the amount/type of ram and that works out to give more options then why is it such an issue? 

I think the real issue is whether there will be other differences besides VRAM count. 

 

It should be made clear in order to avoid confusion. 

[leadeater]

1 hour ago, mr moose said:

But there is a GPU difference outside of ram on the Ti model.  

And so was there with the 1060 3GB and 1060GB and so will there actually be for the 3GB and 4GB models, not a lot I hope but for sure the memory bus width because it's actually impossible there won't be so there is going to be memory bandwidth difference even within the same GDDR type used.

 

1 hour ago, mr moose said:

Again, if there is no difference other than the amount/type of ram and that works out to give more options then why is it such an issue? 

Yes there is a difference. "Which one should I pick, there's too many options and I don't understand them all". Was a lot easier to explain the 1060 3GB and 1060 6GB even with the slight GPU core difference.

 

More options are not better, options that don't make any difference is even worse. More options that inflates other products that would other wise not have been is terrible. BOM cost wise the 3GB and 4GB are doing to be minute, what we pay will controlled to give meaning to those two different models so there won't be a $10 difference which even then is likely more than the parts cost difference between 3GB and 4GB.

 

Never confuse or assume more options with being better, first you must show that having more is better to say it's better.

[pas008]

they been allowing multiple models on x60 variants since 200series

192 core vs 216 core 896mb vs 1.7gb

and I had 1.7gb 216 core versions which was nice

 

thinking about it I think they had multiple versions of 8600 too

[mr moose]

9 minutes ago, leadeater said:

Never confuse or assume more options with being better, first you must show that having more is better to say it's better.

I am not the one making assumptions here. No ones knows what the difference is other than memory.

[leadeater]

Just now, pas008 said:

they been allowing multiple models on x60 variants since 200series

192 core vs 216 core 896mb vs 1.7gb

and I had 1.7gb 216 core versions which was nice

 

thinking about it I think they had multiple versions of 8600 too

True there's often been more than one but the issue isn't that there is, it's that there is 6 and that's never been the case. You don't need 6 variants of the same GPU, we have model numbers so if there actually is a meaningful difference between those 6 options then use model numbers not sub naming. If there isn't a meaningful difference between the 6 then how many is actually needed? 2? 3? 4?

 

This is a mid range GPU correct, if GDDR6 isn't actually going to have a performance difference then why put it on the cards which increases the cost. It would make sense to use it on GPU configs that need the memory bandwidth but then you're moving away from variants to actual GPU core difference so should be using model numbers.

 

Nvidia is likely using the 2060 to increase the volume of GDDR6 to lower the cost so it makes business sense there, though with the GDDR5 options existing that will/should be cheaper and if that doesn't have a performance impact I doubt many are going to pay more for nothing.

[leadeater]

8 minutes ago, mr moose said:

I am not the one making assumptions here. No ones knows what the difference is other than memory.

It is literally impossible to have the same memory bus width with 3GB and 4GB options, that's not an assumption. You can know that from the size of memory chips available and how many are required to make up 3GB and 4GB.

 

Edit:

So it's not an assumption to know the that there is a BOM cost difference, it's not an assumption to generally know what the BOM cost of the memory chips are, and it's not that much of an assumption to know how that translates to a product price and how that difference would normally be small enough that to bother would result in price padding to differentiate the SKUs (I'm not saying that's bad unless you have 6 of the same basic thing).

 

So we are going to have 6 different RTX 2060 variants on the market, two different memory types, two different bus widths (minimum) even before knowing if there are GPU core differences or not. It's unnecessary under the same RTX 2060 model name.

 

It's not an argument against these 6 variants existing at all but to carry the same RTX 2060 model number.

[pas008]

1 minute ago, leadeater said:

True there's often been more than one but the issue isn't that there is, it's that there is 6 and that's never been the case. You don't need 6 variants of the same GPU, we have model numbers so if there actually is a meaningful difference between those 6 options then use model numbers not sub naming. If there isn't a meaningful difference between the 6 then how many is actually needed? 2? 3? 4?

 

This is a mid range GPU correct, if GDDR6 isn't actually going to have a performance difference then why put it on the cards which increases the cost. It would make sense to use it on GPU configs that need the memory bandwidth but then you're moving away from variants to actual GPU core difference so should be using model numbers.

 

Nvidia is likely using the 2060 to increase the volume of GDDR6 to lower the cost so it makes business sense there, though with the GDDR5 options existing that will/should be cheaper and if that doesn't have a performance impact I doubt many are going to pay more for nothing.

i'm not a market analyzer

I dont think you are either

but having 3 5 and 6gb models prollly paid off well for 1060

and we havent even dug into oem models or even considered mobile where many small variables can make huge difference on battery

[mr moose]

8 minutes ago, leadeater said:

It is literally impossible to have the same memory bus width with 3GB and 4GB options, that's not an assumption. You can know that from the size of memory chips available and how many are required to make up 3GB and 4GB.

 

That's memory, I specifically said:

 

15 minutes ago, mr moose said:

I am not the one making assumptions here. No ones knows what the difference is other than memory.

 

[leadeater]

4 minutes ago, pas008 said:

i'm not a market analyzer

I dont think you are either

but having 3 5 and 6gb models prollly paid off well for 1060

and we havent even dug into oem models or even considered mobile where many small variables can make huge difference on battery

The 5GB came later and from memory was not a global SKU. Only the variant with different amounts of memory chips or using different memory types would effect battery life and you can have mobile only SKUs, you don't have to skater shot the desktop market to solve a mobile platform problem.

 

And anyone can analyze the market, it doesn't have to be your literal job to be able to or be allowed to. But that's not even the problem or the point of view I'm looking at, that's the buyer wondering if they need 3GB or 4GB, should you pay more, what's the difference between GDDR5 and GDDR6. Good luck explaining all that even with benchmarks with data you can refer to.

[Fnige]

Havent seen it... so...

MOAR VARIANTS

[leadeater]

7 minutes ago, mr moose said:

That's memory, I specifically said:

 

33 minutes ago, leadeater said:

Never confuse or assume more options with being better, first you must show that having more is better to say it's better.

 

Either you are assuming more options are good or you are ignoring how multiple variants of the same card can be confusing for the consumer/buyer.

 

I'm not assuming more is bad, I know more are bad if they don't actually make a difference or inflate prices to accommodate their existence. If there is an actual performance difference between them then using the same model number is again confusing to the consumer/buyer and therefore bad.

 

Critical thinking applied moves the likelihood away from more options being good for the consumer/buyer to being worse.