• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Possible Radeon 390X / 390 and 380X Spec / Benchmark (do not hotlink images!!!!!!)

Status
Not open for further replies.
Boomstick777 said:
Exactly man, good to see some people with heads still screwed on. We can't compare 4GB HBM VS 4GB of GDDR5 because we just don't know anything about HBM yet. Will it hinder performance, will the games stutter when going from card to system.. We don't know yet.

Only 1 more day to go ! Looking forward to it if not just to end all the lameness.
Click to expand...

Considering the VRAM is on the GPU, then I would only assume the latency between the system ram and vram is going to be the same as the cpu to gpu. Minimal.

Certainly a lot faster than the current system ram -> memory controller -> vram setup.
 
There was an animation I saw linked on Reddit that showed the benefits of HBM, looked like it "might"/"does" work differently so may overcome "some " limitations.

I hold out hope that a 4GB Fury card will not be too limited.
 
chroniclard said:
There was an animation I saw linked on Reddit that showed the benefits of HBM, looked like it "might"/"does" work differently so may overcome "some " limitations.

I hold out hope that a 4GB Fury card will not be too limited.
Click to expand...

The data can be shunted out of the buffer much faster when it comes time to re-use cells so I expect hitching to be either much reduced or gone.
 
People sure lost a lot of faith in AMD. Let's paste this one more time.

When I asked Macri about this issue, he expressed confidence in AMD's ability to work around this capacity constraint. In fact, he said that current GPUs aren't terribly efficient with their memory capacity simply because GDDR5's architecture required ever-larger memory capacities in order to extract more bandwidth. As a result, AMD "never bothered to put a single engineer on using frame buffer memory better," because memory capacities kept growing. Essentially, that capacity was free, while engineers were not. Macri classified the utilization of memory capacity in current Radeon operation as "exceedingly poor" and said the "amount of data that gets touched sitting in there is embarrassing."

Strong words, indeed.

With HBM, he said, "we threw a couple of engineers at that problem," which will be addressed solely via the operating system and Radeon driver software. "We're not asking anybody to change their games."

So AMD acknowledges and is confident it can be solved.
 
bru said:
Your right we don't know anything about 4GB of HBM, but unfortunately we do know all about 4GB of game data, textures and the like. It doesn't matter whether you store them on a HDD, SSD, System ram or even GDDR5 video ram, they still take up 4GB of space. HBM will be no different.
Click to expand...

MadMossy said:
Considering the VRAM is on the GPU, then I would only assume the latency between the system ram and vram is going to be the same as the cpu to gpu. Minimal.

Certainly a lot faster than the current system ram -> memory controller -> vram setup.
Click to expand...

See this is the thing right here, both of these are logical statements. We just don't know how it will work real world in gaming.

If it's rubbish people won't buy it, if it's good people will. At least wait for reviews / user feedback before passing judgement. For the sake of competition and improving performance I hope Fury X is very good and gives Nvidia a run for their money against 980 Ti / Titan X.

Changing the subject, some sick games at E3. Need for Speed looks amazing (Hopefully not just smoke and mirrors), and Fallout. wow. Doom, OMG to many games !
 
Last edited:
SimBy said:
People sure lost a lot of faith in AMD. Let's paste this one more time.

When I asked Macri about this issue, he expressed confidence in AMD's ability to work around this capacity constraint. In fact, he said that current GPUs aren't terribly efficient with their memory capacity simply because GDDR5's architecture required ever-larger memory capacities in order to extract more bandwidth. As a result, AMD "never bothered to put a single engineer on using frame buffer memory better," because memory capacities kept growing. Essentially, that capacity was free, while engineers were not. Macri classified the utilization of memory capacity in current Radeon operation as "exceedingly poor" and said the "amount of data that gets touched sitting in there is embarrassing."

Strong words, indeed.

With HBM, he said, "we threw a couple of engineers at that problem," which will be addressed solely via the operating system and Radeon driver software. "We're not asking anybody to change their games."

So AMD acknowledges and is confident it can be solved.
Click to expand...

AMD cards using more VRAM than Nvidia equivalents on GDDR5.... They going to solve this one for sure;)

Note: This is to get above 10000
 
MadMossy said:
Considering the VRAM is on the GPU, then I would only assume the latency between the system ram and vram is going to be the same as the cpu to gpu. Minimal.

Certainly a lot faster than the current system ram -> memory controller -> vram setup.
Click to expand...

System RAM gets read speed of less than 200GB/s, GDDR5 GPU's have around 336GB/s of throughput... HBM cards have 512GB/s

It doesn't really matter how fast you make the GPU's vram, that isn't the bottleneck

Latency wont have changed much, its bandwidth and power that has been improved, the GPU still has a memory controller, the RAM isn't actually part of the GPU just the PCB traces have been removed
 
andybird123 said:
System RAM gets read speed of less than 200GB/s, GDDR5 GPU's have around 336GB/s of throughput... HBM cards have 512GB/s

It doesn't really matter how fast you make the GPU's vram, that isn't the bottleneck

Latency wont have changed much, its bandwidth and power that has been improved, the GPU still has a memory controller, the RAM isn't actually part of the GPU just the PCB traces have been removed
Click to expand...

System RAM at 200GB/s

NICE system that - what platform ?
 
andybird123 said:
System RAM gets read speed of less than 200GB/s, GDDR5 GPU's have around 336GB/s of throughput... HBM cards have 512GB/s

It doesn't really matter how fast you make the GPU's vram, that isn't the bottleneck

Latency wont have changed much, its bandwidth and power that has been improved, the GPU still has a memory controller, the RAM isn't actually part of the GPU just the PCB traces have been removed
Click to expand...

AMD's thoughts on the matter back when they started developing HBM (right after GDDR5 was ratified) was that capacity and power were chief among the concerns looking forward. I assume b/w is something you always want more of in HPC and the name itself implies the main overarching reason for doing it was indeed b/w.

I hate to use the term "new paradigm" but software is designed based off assumptions about hardware, when HBM becomes standard I think bandwidth will start to matter way more.
 
So is HBM is going to utterly game reliant then? or need the magic of DX12 support which isn't here let alone implemented into games so if the Fury X is everything we want it to be it will be like owning a Ferrari with no wheels, is this AMD just trying to be first reguardless? I mean they did the same with Dx11.
 
kazuya1337 said:
So is HBM is going to utterly game reliant then? or need the magic of DX12 support which isn't here let alone implemented into games so if the Fury X is everything we want it to be it will be like owning a Ferrari with no wheels, is this AMD just trying to be first reguardless? I mean they did the same with Dx11.
Click to expand...

There's an industry term "pipe cleaner" which refers to a product that serves as a money-making test bed for new technology you want to implement in the next product line.

750Ti was a pipe cleaner for Maxwell, for example. Fury serves as a pipe cleaner for HBM and interposers. That doesn't mean it won't be great but getting the HBM engineering work out of the way allows them to merely worry about the new architecture and silicon node in 2016 rather than all three.
 
Taken form HardOCPs 980 Ti review.

At 4K though 4GB of VRAM is clearly not enough. At 4K you want at a MINIMUM 6GB. It is possible though that more may actually help as you start increasing the number of video cards in SLI. 6GB might actually not be enough for some games in 4K when SLI is involved, we will see.
Click to expand...

AMD must have worked a miracle. :p
 
Status
Not open for further replies.
Back
Top Bottom