Vega/Navi Rumors (Updated)

[T1beriu]

AMD Radeon RX Vega with Quake Championship

Source

 

[Crumpet]

I hope Vega launches soon, my R9 290 is starting to artifact

Please hold on <3

 

[Valantar]

I hope Vega launches soon, my R9 290 is starting to artifact

Please hold on <3

Time to walk back the overclock?

 

[T1beriu]

For f***'s sake I hope AMD gets the voltage dialed in right this time. There's no need for another RX580 situation where most chips can be undervolted by about 0.1v or more with a corresponding >20-30w power drop and therefore be run at top boost clocks all the time with efficiency nicely going up... save for the balls to the wall 1450MHz out of the box models that actually need the voltage. I understand this is done to maximize usable chips at x frequency, but come on. It hurts them more than it helps when it's review time.

Yeah AMD has been terrible about that. Hawaii, Fiji and Polaris are all overvolted out of the box and it really kills their efficiency.

It doesn't work like that. If there are a few cards that can go 0.1V less that doesn't mean all can go this low. If only 80% can do 0.1V less and 20% can't, what are you going to do with them?! Throw them away? No, you increase the voltage tolerance to increase the yields to keep making money.

There's a another way. The Nano (low TDP) way. Having 2 versions of the same card - RX 580 (180W) and RX 580 LE (Low Energy) - 85% of performance, but with only 60% of TDP.

 

[Crumpet]

Time to walk back the overclock?

The gpu is already at stock clocks and undervolted to try and preserve its life.

Perhaps i'll try raising the voltage back up but I don't think thats the issue.

 

[nathanddrews]

Quake Champions + Vega will definitely move some product... if the price and performance are right. I'm really hoping they pull a Ryzen on NVIDIA.

$349 for 1080FE performance would be sweet, $499 for 1080Ti perf.

(let me dream)

 

[Mockingbird]

Quake Champions + Vega will definitely move some product... if the price and performance are right. I'm really hoping they pull a Ryzen on NVIDIA.

$349 for 1080FE performance would be sweet, $499 for 1080Ti perf.

(let me dream)

The difference is that Nvidia didn't sit on its *** like Intel.

I don't agree with Nvidia prices, but I do have to say that Nvidia isn't slacking.

 

[zinfamous]

The difference is that Nvidia didn't sit on its *** like Intel.

I don't agree with Nvidia prices, but I do have to say that Nvidia isn't slacking.

Intel wasn't exactly sitting on its petard. They've more or less hit the wall in thermal design and, well, "no one told them it wasn't easy!" to move from 16, to 14, and to 10nm. At least, they are far closer to physical limits in design than nVidia or AMD will soon be when compared to GPU uArch.

 

[Dave2150]

Intel wasn't exactly sitting on its petard. They've more or less hit the wall in thermal design and, well, "no one told them it wasn't easy!" to move from 16, to 14, and to 10nm. At least, they are far closer to physical limits in design than nVidia or AMD will soon be when compared to GPU uArch.

...........

Suggesting Intel has 'hit the wall' in thermal design is a ludicrous suggestion, please educate yourself on this subject before posting nonsense.

Intel's mainstream parts (which are far more popular/affordable than their HEDT chips) still use terrible thermal paste to transfer heat between the die and IHS. They could use solder, which would greatly improve the thermals (and thus performance) of their mainstream parts (educate yourself on the benefits of delidding 7700k's for example).

 

[Crumpet]

...........

Suggesting Intel has 'hit the wall' in thermal design is a ludicrous suggestion, please educate yourself on this subject before posting nonsense.

Intel's mainstream parts (which are far more popular/affordable than their HEDT chips) still use terrible thermal paste to transfer heat between the die and IHS. They could use solder, which would greatly improve the thermals (and thus performance) of their mainstream parts (educate yourself on the benefits of delidding 7700k's for example).

Yes, but go from 4 cores up to 8 cores, and see where they get even with soldering. That's a lot of extra heat.

 

[Dave2150]

Yes, but go from 4 cores up to 8 cores, and see where they get even with soldering. That's a lot of extra heat.

Intel's mainstream parts feature a maximum of 4 cores, not 8. Read the post your quoting.

 

[Crumpet]

Intel's mainstream parts feature a maximum of 4 cores, not 8. Read the post your quoting.

The current versions do.

Coffee Lake will be increasing to 6 in theory, and we'll see from there. I expect we'll see Intel in the 8 core mainstream sooner rather than later.

 

[zinfamous]

...........

Suggesting Intel has 'hit the wall' in thermal design is a ludicrous suggestion, please educate yourself on this subject before posting nonsense.

Intel's mainstream parts (which are far more popular/affordable than their HEDT chips) still use terrible thermal paste to transfer heat between the die and IHS. They could use solder, which would greatly improve the thermals (and thus performance) of their mainstream parts (educate yourself on the benefits of delidding 7700k's for example).

eh, it's not like I'm the first person invoking "the approaching end of Moore's Law." I didn't say they hit, I said they are increasingly limited with each generation. It's just physics. At least, is there some other way to explain how we went from 20%+ increase in performance from generation to generation, to now being happy that we see 5%?

as far as delidding...why is Intel shipping out unsoldered chips and charging customers a rather ridiculous price for such? Seems ludicrous, to me. There's the lack of competition, I guess.

 

[swilli89]

The difference is that Nvidia didn't sit on its *** like Intel.

I don't agree with Nvidia prices, but I do have to say that Nvidia isn't slacking.

We can definitely conflate "sitting on its ass" with "milking the customer". To us consumers these basically amount to the same thing. nVidia has most definitely sat on its ass for a few generations by intentionally not releasing its fastest to market. I know the insta-line people spew is that why should they given a lack of competition.. but their internal incentives for doing this are a moot point when we know for a fact they do this in the open marketplace. Case in point, they released the basically unattainable $1200 Titan X when everyone and their mother knew a $700 (still very profitable) 1080ti was planned all along. This is them de facto sitting on their asses, while not born from complacency, its born from corporate profit motives.. same result.

...........

Suggesting Intel has 'hit the wall' in thermal design is a ludicrous suggestion, please educate yourself on this subject before posting nonsense.

Intel's mainstream parts (which are far more popular/affordable than their HEDT chips) still use terrible thermal paste to transfer heat between the die and IHS. They could use solder, which would greatly improve the thermals (and thus performance) of their mainstream parts (educate yourself on the benefits of delidding 7700k's for example).

Yes so you are basically highlighting the fact that because Intel milked the mainstream segment with 4 cores and cheap thermal transfer material they got caught with their pants down.

 

[Harmaaviini]

So, I'm going to put out a bold prediction Vega reviews tomorrow!

There was that tweet from Ryan Smith posted earlier.


nVidia released a new driver today so any online troll/fanboy/youtuber/viral marketer has the latest performance improvements for popular benchmarks available for them.
http://fudzilla.com/news/graphics/43481-nvidia-releases-geforce-381-89-game-ready-driver

A super moderate posting (probably) a silly question. Almost as if taunting us...



So there. Maybe there's a pattern here or maybe I'm just off my meds...

Swing and a miss.

 

[Mockingbird]

We can definitely conflate "sitting on its ass" with "milking the customer". To us consumers these basically amount to the same thing. nVidia has most definitely sat on its ass for a few generations by intentionally not releasing its fastest to market. I know the insta-line people spew is that why should they given a lack of competition.. but their internal incentives for doing this are a moot point when we know for a fact they do this in the open marketplace. Case in point, they released the basically unattainable $1200 Titan X when everyone and their mother knew a $700 (still very profitable) 1080ti was planned all along. This is them de facto sitting on their asses, while not born from complacency, its born from corporate profit motives.. same result.

One the CPU side, look at how little things have improved since Sandy Bridge.

Then look at the GPU side, and see how things have improved since Fermi

 

[Valantar]

Intel wasn't exactly sitting on its petard.

Well, I sure hope Intel wasn't sitting on a French 16th century door breaching charge. That sounds rather dangerous. Not that I quite understand why it would own one in the first place, though.

 

[IllogicalGlory]

One the CPU side, look at how little things have improved since Sandy Bridge.

Then look at the GPU side, and see how things have improved since Fermi

For one, you can just add more cores to make GPUs faster, which means every node shrink really is a doubling of performance at the same power. GPU architecture is fundamentally different from CPU. You can't just make that comparison without considering the differences.

IBM, Intel, AMD and others have been working to make CPUs faster since the 1970s, GPUs have been niche for a long period and only recently starting to become really important. They have a lot more potential.

 

[raghu78]

One the CPU side, look at how little things have improved since Sandy Bridge.

Then look at the GPU side, and see how things have improved since Fermi

If you look at multi core server CPUs where adding more cores increases performance in server workloads which scale extremely well with more cores then CPUs are doing well too. In fact you should compare the latest Skylake Xeon with 28C/56T and AVX-512 against Sandy-Bridge EP with 8C/16T and AVX-128.

https://en.wikipedia.org/wiki/List_of_Intel_Xeon_microprocessors#Xeon_E5-26xx_.28dual-processor.29

https://en.wikipedia.org/wiki/List_...kylake-SP.22_.2814_nm.29_Scalable_Performance

The problem is improving single thread performance is very hard on these massive out of order cores especially while keeping power under control. Intel, AMD, Apple and any company which designs very high performance CPUs face the same challenges.

 

[Glo.]

http://www.tweaktown.com/news/57313....it&utm_medium=twitter&utm_campaign=tweaktown

 

[yellow_eggplant]

http://www.tweaktown.com/news/57313....it&utm_medium=twitter&utm_campaign=tweaktown

"The card will rock 9.8 TFLOPs..."

So small Vega? We already know Vega would be around 12.5 TFLOPS based on the Instinct M125 preview. Oh and why would it be at 1000mhz-1200mhz? Would be weird for AMD's focus on higher clock speeds take a step back. Probably just placeholder numbers or something

 

[ItsAlive]

Do you think Vega will really be 10.5"? Fiji based cards were like 7.5" and I thought that one of the perks of HBM is that they can make the PCB much smaller due to the smaller memory footprint.

 

[Mopetar]

http://www.tweaktown.com/news/57313....it&utm_medium=twitter&utm_campaign=tweaktown

Embarrassing if true.

Poor Volta indeed. It will be all alone in the market.

But this just seems implausible or there's some key information missing, such as the test title being something that seriously favors NV or just runs like donkey on AMD GPUs.

 

[caswow]

i cant believe people in this forum actually believe this peace of ish article in the slightest. its even sad.

 

[keymaster151]

http://www.tweaktown.com/news/57313....it&utm_medium=twitter&utm_campaign=tweaktown

It says right in the image "Radeon RX Vega performance estimated based on architecture, shader count and clocks." This is nothing but rumors and speculation, not actual benchmark results.