Navi Rumors

Status
Not open for further replies.
It was a bit of a logistical nightmare for AMD more than anything else. Hard to predict cause well they not are manufacturing the memory and the manufacturer of course will say everything is looking good for their end, since the exclusivity contract is in place, there is nothing AMD could do about it.

Not strictly true. Anything is possible through a good contract in terms of deadlines, penalties, etc. In fact if Vega RX hits with Samsung HBM then I expect they exercised such a provision.

Of course, that would only offset their damage since once they have finalized the design and especially after tape-out they are stuck with that memory tech regardless of any shortcomings (unless they hedge with multiple memory controllers but that's not cheap either)
 
Not strictly true. Anything is possible through a good contract in terms of deadlines, penalties, etc. In fact if Vega RX hits with Samsung HBM then I expect they exercised such a provision.

Of course, that would only offset their damage since once they have finalized the design and especially after tape-out they are stuck with that memory tech regardless of any shortcomings (unless they hedge with multiple memory controllers but that's not cheap either)


Exclusivity contracts based on volume sales usually don't have strict deadline, what AMD probably did was based on how much volume they will need for Vega, already partitioned of what they will need from Hynix and get the rest from Samsung over the life span of Vega's manufacturing.
 
Navi and then Suni, obviously. #waitforSuni

Suni_4.jpg
 
My 4 predictions.
1. It will be on time and 75% faster than top end nVidia, with day-one perfect drivers.
2. It will mine like a mutherfucker on all crytocunts and so quickly, that it makes crytocurrency value depreciate geometrically. One million Etherum or bitcoin won't buy you a stick of gum.
3. It will have an advanced AI and new user interface that can predict if you'll have a bad day and purchase BOTH hookers AND blow, before you knew you needed them.
4. Oh, and it plays games real well.
 
BY MY POWERS OF PROPHECY...

I WILL PREDICT THE LAUNCH OF NAVI!!!


1. IT WILL BE LATE

Navi will not hit when AMD first hints it to launch. In fact, it will probably be 2-3 quarters later than what even the most pessimistic of roadmaps predict. It will also be teased, announced and launched multiple times before anyone can actually get their hands on it.

2. IT WILL BE INEFFICIENT COMPARED TO NVIDIA'S THEN-CURRENT CARD

It won't hit the energy efficiency of Nvidia's cards. In fact, it may not even match Pascal at the high end. It's die-size will be bigger than similarly priced Nvidia cards, and it will not be as easy to OC.


3. IT WILL NOT BE ALL THAT IMPRESSIVE
Not to say it will be bad, rather, it won't break new ground, and it won't disrupt the market much at all. Dispite AMD's suggested MSRPs, the retail cards will rise in line to match simillarly performant Nvidia cards. It will not take the performance crown. It may have more VRAM per dollar than Nvidia's products, but that is about it.

4. IT WILL NOT HAVE ANY DISRUPTIVE TECHNOLOGY, REGARDLESS OF WHAT AMD CLAIMS
Regardless of what RTG says, the cards' memory architecture and core improvements won't count for much. It's most lauded features will really just result in 'being faster than Vega'. If anything, these new tech bits will only provide more leeway for AMD to make "drivers aren't ready" excuses.

HERE SPEAK'TH THE [H]ARD PROPHET.
MAY HIS WORDS BE SET IN STONE

plus the corollary:

5. VIRAL MARKETERS WILL FLOG NAVI TO DEATH IN MULTIPLE FORUMS USING THE SAME POSTS
Regardless of the performance of the product - even after metrics have been released - there will always be another goalpost to move to (*cough* "dual gpu!", "Mining rates!", "power usage is only cents per day!")
 
Page 1 : Late is to be determined by AMD at its sole discretion regardless of any roadmaps or teasing, express or implied, of release date or availability or launch, in any multiple, in any market or capacity, for any purpose. Efficiency is based on measured memory devices + interface power consumption for whatever we like at the time. Size is a term that shall be used only appropriately and in correct context; it matters. Easy is a relative term and is meaningless in the scope of overclocking and is included for completeness only. The product as defined by AMD may be impressive to some but not to you and we are not liable for differences in opinion about impressiveness. Ground may not be new, or broken, regardless of claim by any claimant. MSRP is only a suggestion. Similarly performant may vary upon region, locale, time of day or type of brownie consumed. A crown may refer to any type of headwear and despite resemblences is not a trademark of Burger King. VRAM per dollar is subject to interpretation in comparison to Nvidia if not all of the VRAM is usable. Note that it is only about. Disruptive does not mean liable. Technology may mean magic to some and that's alright with AMD. Claim only refers to something we said not actual or alternative fact. Features may include things you call bugs or performance that we consider acceptable. A driver or drivers are ready when we say they are and always function correctly for the designed target whether or not the outcome is desired by the customer or anyone other than AMD. AMD does not and has never made excuses; in fact, that technology is not being developed.
 
You're expecting it to hit with something else other than Samsung HBM?

Supposedly they were locked in with Hynix but there had to be some kind of end date on it. If Hynix could not deliver what was agreed upon on time then.
 
Supposedly they were locked in with Hynix but there had to be some kind of end date on it. If Hynix could not deliver what was agreed upon on time then.


Hynix had no issues delivering product outside of the the actual first manufacturing of the it. The second part of it is AMD buying enough HBM memory to fulfill that contract, and Fiji doesn't seem to be enough for that. Fiji didn't sell, why would AMD push more into it? Also think about Raja saying they invested a lot of money to get the HBM pipeline down and now have to recoup that money, sounds like they need to sell more products than it being fully Hynix's fault, initially it could have been the delay of HBM (which is likely) but the the other half of the issue is AMD selling more products so they can maintain or increase their orders. Samsung has been delivering HBM 2 to nV for over a year now, so if the exclusivity contract ended because Hynix could not deliver HBM memory to AMD on a timely basis, they would have switched over to Samsung long ago. That didn't happen.
 
Last edited:
"AMD has not provided information about many aspects of its upcoming Vega-based products, including the capacity and layout of the memory on board, but SK Hynix's catalog notes that each 1.6Gbps stack offers 204.8 GB/s of aggregate memory bandwidth. If we assume that some of the first consumer Vega cards use 8GB of HBM2 memory, as our experience suggests they will, a pair of those 4GB stacks would be required. That configuration would provide a hypothetical graphics card with a total of 410 GB/s of memory bandwidth.

For comparison's sake, the four HBM stacks in the R9 Fury X provided 512 GB/s of peak bandwith, Nvidia's high-end GeForce GTX 1080 can access its GDDR5X memory at a rate of 320 GB/s, and the Pascal Titan X's 12GB of GDDR5X memory has a staggering aggregate bandwidth of 480 GB/s on that card's 384-bit bus. If the highest-end Vega card does use a pair of these 1.6Gbps modules, the cards would lose the bandwidth crown to Nvidia's most expensive desktop offering."

http://techreport.com/news/31357/rumor-changes-in-sk-hynix-catalog-could-hint-at-vega-hbm-speeds

True Story or Cool Story?
 
TR just guessed most likely, I think the bandwidth will be 483 gb/s just like Vega FE
 
Last edited:
Discussing Nvidia architectures aren't nearly as exciting because they're on such a regular, predictable cadence that we already know what we're going to get. AMD on the other hand, it's like a traveling circus that is always different when it comes around. Sometimes it has some awesome stuff, but sometimes the clowns forget to perform.
 
Lmao.

People who exhibit blind loyalty to any given brand/company are just....well, sad.

Unless someone owns stock in the company they vehemently defend, what is the point? Buy whatever you god damned need not whatever you're given by the corporate dick you got in your mouth.
 
That will just scale up with Navi or be moved onto the die or both, before mentioning hbm3, being like you said earlier, in not so many words that it just isn't a memory bandwidth issue. It isn't an issue for Nvidia either, if I understand correctly.
 
Lmao.

People who exhibit blind loyalty to any given brand/company are just....well, sad.

Unless someone owns stock in the company they vehemently defend, what is the point? Buy whatever you god damned need not whatever you're given by the corporate dick you got in your mouth.

Well even if you own stocks, what is posting online going to do?

It's not like you can single handedly improve the public perception of the product one post at a time.
 
"AMD has not provided information about many aspects of its upcoming Vega-based products, including the capacity and layout of the memory on board, but SK Hynix's catalog notes that each 1.6Gbps stack offers 204.8 GB/s of aggregate memory bandwidth. If we assume that some of the first consumer Vega cards use 8GB of HBM2 memory, as our experience suggests they will, a pair of those 4GB stacks would be required. That configuration would provide a hypothetical graphics card with a total of 410 GB/s of memory bandwidth.

For comparison's sake, the four HBM stacks in the R9 Fury X provided 512 GB/s of peak bandwith, Nvidia's high-end GeForce GTX 1080 can access its GDDR5X memory at a rate of 320 GB/s, and the Pascal Titan X's 12GB of GDDR5X memory has a staggering aggregate bandwidth of 480 GB/s on that card's 384-bit bus. If the highest-end Vega card does use a pair of these 1.6Gbps modules, the cards would lose the bandwidth crown to Nvidia's most expensive desktop offering."

http://techreport.com/news/31357/rumor-changes-in-sk-hynix-catalog-could-hint-at-vega-hbm-speeds

True Story or Cool Story?
The specs revealed show the memory on RX Vega 64 operating at 1.89 GT/s over a 2048-bit bus, meaning bandwidth of 483 GB/s. Vega 56 has its memory running at 1.6 GT/s over the same bus for your quoted 410 GB/s.

http://www.pcworld.com/article/3212206/computers/radeon-rx-vega-64-pricing.html
http://techreport.com/review/32320/amd-radeon-rx-vega-64-and-rx-vega-56-graphics-cards-revealed
https://www.pcper.com/reviews/Graph...ga-64-and-Vega-56-Specs-Prices-Power-Detailed

Your article is from January this year. I don't know what you're trying to say.
 
Better luck next time, RTG. I do think HBM/HBM2 is overkill, regardless GDDR5X, GDDR6 or GDDR6X may prove adequate for high levels of performance and without the supply concerns (that may now be ironed out). I wonder what 2018/2019 will bring in terms of AMD GPUs...
 
Given AMD's hardwork on HBM, it's ironic that Nvidia is the one who's reaping the rewards.
 
Better luck next time, RTG. I do think HBM/HBM2 is overkill, regardless GDDR5X, GDDR6 or GDDR6X may prove adequate for high levels of performance and without the supply concerns (that may now be ironed out). I wonder what 2018/2019 will bring in terms of AMD GPUs...

Now that we know how much HBM2 costs, we need to talk about why AMD decided to use it. Like most of AMD’s hardware, the company is partly trying to make a long-term technological play in the market. This started with Fiji and has progressed through Vega.

There’s more to it, though. HBM2 critically allows AMD to run lower power consumption than GDDR5 would enable, given the Vega architecture.

Speaking with Buildzoid, we know that Vega: Frontier Edition’s 16GB HBM2 pulls 20W max, using a DMM to determine this consumption. This ignores the voltage controller’s 3.3v draw, but we’re still at 20W memory, and no more than an additional 10W for the controller – that’s less than 30W for the entire memory system on Vega: Frontier Edition.

We also know that an RX 480 uses 40-50W for its 8GB, which is already a significant increase in power consumption per-GB over Vega: FE. The RX 480 also has a memory bandwidth of 256GB/s with 8GB GDDR5, versus Vega 64’s 484GB/s. The result is increased bandwidth, the same capacity, and lower power consumption, but at higher cost to build. In order for an RX 480 to hypothetically reach similar bandwidth, power consumption would increase significantly. Buildzoid calculates that a hypothetical 384-bit GDDR5 bus on Polaris architecture would push 60-75W, and an imaginary 512-bit bus would do 80-100W. For this reason alone, HBM2 saves AMD from high power budget that would otherwise be spent solely on memory. This comes down to architectural decisions made years ago by AMD, which are most readily solved for with HBM2, as HBM2 provides greater bandwidth per watt than GDDR5. HBM is effectively a necessity to make Vega at least somewhat power efficient while keeping the higher memory bandwidth. Imagine Vega 56, 64, or FE drawing an additional 70-100W – the world wouldn’t have it, and it’d be among the hottest cards since the GTX 480 or R9 290X.

The Vega architecture is clearly starved by memory bandwidth, too: Overclocking HBM2 alone shows this, as its gains are greater than just core clock increases. AMD didn’t have another choice but to go with HBM2, even though costs would be roughly one-third on the memory. GDDR5 might be possible, but not without blowing power consumption through the roof or losing on performance by limiting bandwidth.

http://www.gamersnexus.net/guides/3032-vega-56-cost-of-hbm2-and-necessity-to-use-it
 
will come out shortly after the 2080ti , cost the same as the 2080ti and perform close to 1080ti and heat your office at the same time, and you wont be able to find one for the first 6 months
 
will come out shortly after the 2080ti , cost the same as the 2080ti and perform close to 1080ti and heat your office at the same time, and you wont be able to find one for the first 6 months

...but how much power does it need?
 
BY MY POWERS OF PROPHECY...

I WILL PREDICT THE LAUNCH OF NAVI!!!

HERE SPEAK'TH THE [H]ARD PROPHET.
MAY HIS WORDS BE SET IN STONE

Hard to say. I don't think they will have anything that will hands down beat Volta. But they may close the gap even more. Remember this is the FIRST true new architecture under Raja's leadership. Vega was an extension of Fury. Raja might pull off a Lisa Su with Ryzen. You don't know. Some of the theories they are applying here for Navi are interesting and quite a departure.

And whiile some of you knocked Vega, it does have some interesting extras NVIDIA doesn't have. But these mostly apply to large datasets and packed math. While not useful for most gamers now, they are useful to people who use their cards for number crunching and editing.
 
Given AMD's hardwork on HBM, it's ironic that Nvidia is the one who's reaping the rewards.


Its because they have an architecture that can actually use HBM to the fullest. They aren't using it for power savings, they are using it with a 1024 bit bus which they need all that bandwidth for DL and HPC apps.
 
Did they give a reason? Your post wasn't negative at all.

Yes, for supposedly "trolling".

On Anandtech, if you talk about topics that AMD (or more specifically, RTG) is uncomfortable with, it gets sweep under this broad "trolling" category.

I will give you another example. The moderator also closed down the thread about Koduri going on a sabbatical even though it doesn't seem to violate any rules.

https://forums.anandtech.com/thread...sa-su-to-head-up-rtg-in-the-meantime.2518554/
 
Yes, for supposedly "trolling".

On Anandtech, if you talk about topics that AMD (or more specifically, RTG) is uncomfortable with, it gets sweep under this broad "trolling" category.

I will give you another example. The moderator also closed down the thread about Koduri going on a sabbatical even though it doesn't seem to violate any rules.

https://forums.anandtech.com/thread...sa-su-to-head-up-rtg-in-the-meantime.2518554/
That was honestly weird. No one goes on vacation the same time their major product releases.
 
If it launches in 2018 I find it really hard to believe that it won't have HBM2. We haven't even gotten a rumor of a new memory standard. Not that that's a bad thing.
 
Here we go again :depressed: !!

-Let's give a big-welcome to the beginning of "#Wait for NAVI" era !!! ( https://www.techpowerup.com/237725/amd-navi-gpu-by-q3-2018-report )

-One note though: let's keep in mind that it took AMD about 7 months delay, since their first VEGA-demonstration at gaming ( DOOM@4K : https://hardforum.com/threads/amd-v...stem-rocks-doom-at-4k-ultra-settings.1921718/ , - January 2017 -), until we actually saw VEGA ( - July/August 2017- ).
And right now we don't even have a game, so ....... you get the point :p !!!
 
About time to wake this thread up.
Personally I'm hoping it gets as juicy as the Vega Rumors thread:LOL::LOL:

#WaitForNavi
 
Seriously have nothing better to post?

Actually re-reading the first few posts made me laugh a bit. I'd forgottten it was initially rumoured to be released in early 2018. AMD needs to update that. The last I've heard from them is 2020.
 
Status
Not open for further replies.
Back
Top