AMD's chance to catch up

  • Thread starter Deleted member 93354
  • Start date
AltTabbins said:
AMD might not be "winning" for me, but it's sure as hell winning for the mid range market. I have 2 brothers in law who are both mid-range pc gamers. They both play at 1920x1080 and run their hardware for several years before even considering and upgrade. Right now one is using a 460 gtx, and the other using a gtx 760. For $200, they get a huge upgrade and are able to play the newest games at ultra/high settings just fine. Most people here don't like that AMD doesn't have the top card, but for people like my brothers, it's not even close.. AMD all day for them.
Click to expand...

The 1060 will be $250 and be a much better card overall. So you can't say AMD is winning anything because they aren't. In eight days from now when AIB 1060s hit the market for $250, the RX 480 will be relegated as another AMD fail just like the Fury X. They already had a large PR disaster with the PCI-E power draw issue and many people just don't trust them now.
 
AltTabbins said:
AMD might not be "winning" for me, but it's sure as hell winning for the mid range market. I have 2 brothers in law who are both mid-range pc gamers. They both play at 1920x1080 and run their hardware for several years before even considering and upgrade. Right now one is using a 460 gtx, and the other using a gtx 760. For $200, they get a huge upgrade and are able to play the newest games at ultra/high settings just fine. Most people here don't like that AMD doesn't have the top card, but for people like my brothers, it's not even close.. AMD all day for them.
Click to expand...

They could have purchased 380Xs for $200 last year and played games at 1080P in high settings. For someone using a GTX 460, that would have been a upgrade worthy of God Rays coming down from the sky.

Hell, I just pulled a 960 from my PC and that even played Doom just fine at 1080P.
 
5150Joker said:
The 1060 will be $250 and be a much better card overall. So you can't say AMD is winning anything because they aren't.
Click to expand...

What sources do you have on that? You are speculating. But since you are, how long do you think it will take before you see a gtx 1060 for $250?
 
AltTabbins said:
What sources do you have on that? You are speculating. But since you are, how long do you think it will take before you see a gtx 1060 for $250?
Click to expand...

NVIDIA claims we'll see it on day one of release with AIB cards. I'm guessing just like the RX 480, with supply being very sparse, the price will go up initially and then come down in time when more stock is available. As for it being a much better card, NVIDIA is the concrete source as they claim 75% more efficient and 15% on average faster. They said the 1070 would be faster than Titan X and that was true so I don't doubt they're telling the truth with regards to the RX 480. I mean the RX 480 eats as much power as a 1070 so it's not exactly difficult to beat in power efficiency and then if you factor in the 1060 hitting 2 GHz quite easily (there's leaked images of that on VideoCardz), it will put even more distance between it and the 480.
 
5150Joker said:
NVIDIA claims we'll see it on day one of release with AIB cards. I'm guessing just like the RX 480, with supply being very sparse, the price will go up initially and then come down in time when more stock is available. As for it being a much better card, NVIDIA is the concrete source as they claim 75% more efficient and 15% on average faster. They said the 1070 would be faster than Titan X and that was true so I don't doubt they're telling the truth with regards to the RX 480. I mean the RX 480 eats as much power as a 1070 so it's not exactly difficult to beat in power efficiency and then if you factor in the 1060 hitting 2 GHz quite easily (there's leaked images of that on VideoCardz), it will put even more distance between it and the 480.
Click to expand...

Ah, ok, Nvidia said so. And the others were talking about using Sapphire as a source...
 
Remon said:
Ah, ok, Nvidia said so. And the other people are talking about Sapphire...
Click to expand...

Except Sapphire is speculating NVIDIA was scared which is too funny. NVIDIA wrote the following in their comparison to the 480:

Game list includes BioShock Infinite, Crysis 3, Grand Theft Auto V, Rise of the Tomb Raider, Star Wars Battlefront, Witcher 3, Ashes of the Singularity, The Division, and more. All games tested at 1080p resolution. GTX 1060 was tested with driver version 368.64 with a TGP measured at 120 watts. As of July 5, 2016, AMD RX480 competitive product was tested with driver version 16.6.2 across the same games and resolution with a measured TGP of 185 watts. - See more at: A Quantum Leap for Every Gamer: NVIDIA Unveils the GeForce GTX 1060

So what's going to happen when the above is validated on the 19th? Lots of excuses I'm sure. And like I mentioned, the 1060 can hit 2 GHz vs the 480 that can't OC at all and needs special binned parts by AIB to compete.

GTX-1060-2-GHz.jpg
 
Last edited:
The reason why the RX 480 doesnt always beat 390 is probably bc it has less Rops, but it is doing equal work with less hardware. There is definitely IPC improvement from gcn 4.

For the 1070/1080 Nvidia is kicking ass purely bc of the more mature tsmc process that allows them insane clockspeeds. There is nearly zero ipc improvement from maxwell to pascal and there have been some write ups looking at those numbers.
If Global foundries can improve the 14nm process AMD will be see some huge improvements , of course there is no guarantee on if or when that will occur at global foundries.
 
even better is when you are Nvidia and you strike gold because you chose tsmc over global foundries and through nearly zero improvements from nvidia engineers internally.
Poor amd.... imagine if the RX 480 was made at tsmc. RX 480 would probably be running @ 1500 mhz stock and 30% less power consumption
 
sleepybp said:
even better is when you are Nvidia and you strike gold because you chose tsmc over global foundries and through nearly zero improvements from nvidia engineers internally.
Poor amd.... imagine if the RX 480 was made at tsmc. RX 480 would probably be running @ 1500 mhz stock and 30% less power consumption
Click to expand...

Where you coming up with this zero IPC improvement b.s. from?
 
sleepybp said:
even better is when you are Nvidia and you strike gold because you chose tsmc over global foundries and through nearly zero improvements from nvidia engineers internally.
Poor amd.... imagine if the RX 480 was made at tsmc. RX 480 would probably be running @ 1500 mhz stock and 30% less power consumption
Click to expand...

5150Joker said:
Where you coming up with this zero IPC improvement b.s. from?
Click to expand...

yeah you don't spend enough money to got to mars just for a new process node alone.
 
lmao 5150Joker

Why you mad bro?

Pascal, A New King?? - Clarence Spurr

how much faster is 1080 than 980ti?
980ti is 2816 cores @ around 1ghz stock 1080 is 2560 cores @1700 ish stock

You only need to have middle school math to see where these gains are coming from
hint : its the mhz ;)

I don't see how this is not obvious to everyone. I'm not an amd retard I'm gaming on gtx 770 atm.

Pascal kills it in power efficiency and clock speed clearly I see that
 
Pascal has great performance gains because it's clocked so high. The ability for it to clock so high whilst drawing relatively little power, though, is because it was engineered to do so. You don't just shrink an architecture and presto! you have greatness. AMD needed to do their homework better. Maybe Vega will benefit from the lessons learned with Polaris.
 
sleepybp said:
lmao 5150Joker

Why you mad bro?

Pascal, A New King?? - Clarence Spurr

how much faster is 1080 than 980ti?
980ti is 2816 cores @ around 1ghz stock 1080 is 2560 cores @1700 ish stock

You only need to have middle school math to see where these gains are coming from
hint : its the mhz ;)

I don't see how this is not obvious to everyone. I'm not an amd retard I'm gaming on gtx 770 atm.

Pascal kills it in power efficiency and clock speed clearly I see that
Click to expand...

I guess you think improvements in preemption for DX 12, scheduling, more warps per SM, faster instruction sets etc don't translate to IPC gains? You don't have to be an engineer to see NVIDIA made some pretty distinct changes to Pascal's architecture that improves IPC. Yes a big chunk of the gains are from the node change but they did a lot of tweaking to the architecture to reach those clocks. AMD's 14nm FinFET is based on Samsung's design who in turn stole from TSMC so the process excuse just doesn't fly.
 
Last edited:
sleepybp said:
lmao 5150Joker

Why you mad bro?

Pascal, A New King?? - Clarence Spurr

how much faster is 1080 than 980ti?
980ti is 2816 cores @ around 1ghz stock 1080 is 2560 cores @1700 ish stock

You only need to have middle school math to see where these gains are coming from
hint : its the mhz ;)

I don't see how this is not obvious to everyone. I'm not an amd retard I'm gaming on gtx 770 atm.

Pascal kills it in power efficiency and clock speed clearly I see that
Click to expand...
You do realize you can not just clock the gpu higher and get a linear performance increase there are clearly engineering changes, for starters the significantly better color compression.
 
Ocellaris said:
Using Sapphire as a source for Nvidia anything? This is like going to a Ford dealership and asking them about Chevy.
Click to expand...

You do realize, Sapphire and Zotac were under the same parent company right?

I don't know why Sapphire is not listed under PCPartner any longer, but yeah.
 
joker why such harsh words? you are way too angry

Nvidia is following the tick/tock strategy of intel, with this one being the process shrink. And they pulled it off quite well which is clear in their clockspeed and power improvements.

Seriously why are you so butthurt? Their strategy has worked well and I admit it. But you seem to be in denial about what their strategy was
 
sleepybp said:
joker why such harsh words? you are way too angry

Nvidia is following the tick/tock strategy of intel, with this one being the process shrink. And they pulled it off quite well which is clear in their clockspeed and power improvements.

Seriously why are you so butthurt? Their strategy has worked well and I admit it. But you seem to be in denial about what their strategy was
Click to expand...
I don't think it has anything to do with being "butthurt", but instead has to do with making things up like no IPC improvements with absolutely zero evidence to back up the statement.
 
If you go by Instructions Per Cycle (IPC) he's technically correct. Clock for clock, Maxwell has a slightly higher IPC than Pascal (ignoring architectural improvements and everything else 5150Joker mentioned in his post). But since Pascal cards are clocked much higher at stock settings gives them a performance advantage.
 
"I see, so in another words you're talking out of your ass. Thought as much. I guess you think improvements in preemption for DX 12, scheduling, more warps per SM, faster instruction sets etc don't translate to IPC gains? You don't have to be an engineer to see NVIDIA made some pretty distinct changes to Pascal's architecture that improves IPC. Yes a big chunk of the gains are from the node change but they did a lot of tweaking to the architecture to reach those clocks. AMD's 14nm FinFET is based on Samsung's design who in turn stole from TSMC so the process excuse just doesn't fly."

and who's "talking out of their ass" now?

Maybe you should do a little research of your own before you get butthurt next time
 
sleepybp said:
"I see, so in another words you're talking out of your ass. Thought as much. I guess you think improvements in preemption for DX 12, scheduling, more warps per SM, faster instruction sets etc don't translate to IPC gains? You don't have to be an engineer to see NVIDIA made some pretty distinct changes to Pascal's architecture that improves IPC. Yes a big chunk of the gains are from the node change but they did a lot of tweaking to the architecture to reach those clocks. AMD's 14nm FinFET is based on Samsung's design who in turn stole from TSMC so the process excuse just doesn't fly."

and who's "talking out of their ass" now?

Maybe you should do a little research of your own before you get butthurt next time
Click to expand...

Again, you ignored the IPC improvements made to the architecture. Just because that guys so-called efficiency calculation using canned benchmarks from AT and OC club hasn't reflected it yet doesn't make anything conclusive as the architecture is brand new and drivers aren't optimized. In addition, there may be relatively few new games that take advantage of Pascal vs Maxwell. And once again, you claim AMD is falling behind because they are using GloFlo and not TSMC yet ignored the fact that GloFlo is using what Samsung gave them who in turn stole from TSMC.
 
There is no IPC improvement, there is a decreace in IPC from maxwell to pascal so stop saying otherwise or you will just look more uninformed.
Did they add features in pascal ? Yes, but not ipc improvement.
You think they added better support for dx12 or vulcan? here is the latest nvidia "win" in new apis such as dx12/vulkan

Doom vulkan support added:

New Patch Brings Vulkan support to Doom + Benchmarks
 
  • Like
Reactions: N4CR
like this
you just have to either be the bigger man/woman and ignore the bs or put them on ignore. its frustrating to the people that are trying to read and learn and pick up new info when every f'n thread if filled with people fighting and trying to prove they know more than the next guy. especially when NONE of us are the ones making the cards and KNOW everything about them!
 
It doesn't even matter if nVidia gave up 50% IPC, or break even or what. All that matters is they have a combination of technology that gives us the best performance and/or value.

I don't even understand why it was even brought up. If nVidia streamlines their arch to reach insane clocks, or increases IPC at the cost of clocks, that's each company's decision.

I do agree with Joker, nVidia did do a lot of engineering with Pascal and catapulted themselves further. I think there's more to this than AMD picking the wrong fab (if they even did.) If AMD did pick the wrong fab, that's another technical mistake on their part.

Nothing nVidia does is knee jerk reaction. Their decisions are planned and calculated out way in advance, hence why they are leading.
 
Last edited:
sleepybp said:
There is no IPC improvement, there is a decreace in IPC from maxwell to pascal so stop saying otherwise or you will just look more uninformed.
Click to expand...

Pray tell what qualifies as IPC improvements to us "uninformed" nvidiots?
 
"It doesn't even matter if nVidia gave up 50% IPC, or break even or what. All that matters is they have a combination of technology that gives us the best performance/value."

True statement I agree with this as a consumer this is what matters
performance/value
 
sleepybp said:
If Global foundries can improve the 14nm process AMD will be see some huge improvements , of course there is no guarantee on if or when that will occur at global foundries.
Click to expand...
GF has a long long history of a new process being craptastic to start then steadily improves to the point of being downright excellent.
 
§kynet said:
GF has a long long history of a new process being craptastic to start then steadily improves to the point of being downright excellent.
Click to expand...
If you want to be a perpetual also-ran in the chip industry, that's one way to do it.
 
sleepybp said:
lmao 5150Joker

Why you mad bro?

Pascal, A New King?? - Clarence Spurr

how much faster is 1080 than 980ti?
980ti is 2816 cores @ around 1ghz stock 1080 is 2560 cores @1700 ish stock

You only need to have middle school math to see where these gains are coming from
hint : its the mhz ;)

I don't see how this is not obvious to everyone. I'm not an amd retard I'm gaming on gtx 770 atm.

Pascal kills it in power efficiency and clock speed clearly I see that
Click to expand...
You do realise the 1080 is a node and architecture improvement over the 980 and not 980ti.
Sure compare to 980ti from a consumer buying perspective, but it is wrong to technically think the 1080 evolved from that model; look at the GPC-SM structure to see what I am going on about.
This is why the 980ti still does so well compared to the 980.
SO if you want actual IPC/perf-watt/etc then look at the 980 when talking about the 1080.
Cheers
 
Ocellaris said:
Using Sapphire as a source for Nvidia anything? This is like going to a Ford dealership and asking them about Chevy.
Click to expand...

You take it with a larger grain of salt, but it doesn't necessarily mean that everything they say about the competition is a lie. To me only idiots deal in absolutes like that, generally speaking.
 
DigitalGriffin said:
So the Maxwell edition cards have the same IPC as [edit]Pascal [/edit](Sorry it was late at night and I was tired) . And the 1080 actually starts to lose efficiency because memory on the board starts to bottle neck performance.

Here's your chance AMD...Don't blow it.
Click to expand...

AMD doesn't have to do anything... DX12 and Vulkan will :D

original.jpg


original.jpg
 
wantapple said:
AMD doesn't have to do anything... DX12 and Vulkan will :D

original.jpg


original.jpg
Click to expand...
Have to admit that FuryX performance jump is very impressive. Even the new 480 is looking good with Vulkan. Almost makes the "nearly 980Ti performance" remarks look true, in at least one game. LOL
 
Dam thats a huge performance jump! Nvidia cards getting raped top to bottom!
 
You must log in or register to reply here.
Back
Top