radeon 7

What’s up folks !!. How’s the weather. Chilly I hope. Or whatever it is you like. I mean the point is hope you are doing A-okay. First off, I would like to point out that this post is almost entirely speculation or a perspective into what Radeon 7 looks like.

Reasons are obvious of course.

Lisa Su of AMD unveiled Radeon 7 very recently and we don’t have the card yet :P.

We will have solid ground on what the card actually is only when we have benchmarks about it.

Although, AMD displayed their own benchmarks for the graphics card but let’s just say we will be satisfied once we have seen them first hand.

Radeon 7 : The “What Is”,

So what is it that we are going to focus on in this post.

Well, we will get into some key details about this card.

Details being as in why wasn’t this card created with GDDR6 ?.

Why does it have 16 GB of memory head-space ?.

Why are they deploying HBM2 memory which reportedly increases the overall manufacturing cost ?.

So why are we taking these questions ?.

That’s a good question !!. :P.

Because as always I was browsing through the internet looking to find what people’s tech problems are.

Don’t worry I look at life problems too.

But in this post let’s look at these queries that Radeon 7 launch posed in the minds of our beautiful people.

So I decided it’s an interesting topic and decided why not address all these issues at large through one post.

I wish to tell you that this post won’t become stale with time.

You see I am an evergreen fellow. Atleast I like to think I am.

So the issues I am going to address in this post are more learning oriented and not just I being a smartass.

You follow ?.

You are going to take something incredible from this post whether you are reading this post immediately or 5 years hence :P.

So let’s dive in, alright.

GDDR6 VS HBM2,

First off, you can’t get 1 TB/s bandwidth on GDDR6.

With the current sense of tech that is deployed in most consumer grade cards, you just can’t have 1 TB/s bandwidth made available using GDDR6.

Will that be a possibility in the future ?.

We don’t know.

But as of now, 1 TB/s can only be made possible with HBM2.

So let’s say if you actually created a GDDR6 memory with 16 GB, which isn’t in production yet.

Trying to achieve 1 TB/s bandwidth by incorporating 512 bit memory configuration around the GPU, that would be a PCB designing nightmare.

So the PCB layout manufacturing cost for something like 1 TB/s bandwidth over GDDR6 would be hazardous to say the least.

When it comes to gaming workloads the 1 TB/s bandwidth will just be a massive overkill.

The performance scaling for the Radeon 7 tops at about 700 GB/s bandwidth theoretical.

So it seems like the Radeon 7 has 1 HBM stack more than what it would need if this card was just aimed at gaming.

Now you may ask why not create an exclusive gaming oriented Radeon 7 card with GDDR6 memory ?..

Yeah !!. It ain’t that easy.

GPU dyes, yes the core GPU dyes are very very expensive to make.

And the GPU dye created for Radeon 7 is compatible with HBM2 and not with GDDR6.

So depending on the number of Radeon 7 units AMD looks to sell, creating a GDDR6 version of Radeon 7 could be very very expensive from a business perspective.

Also the primary purpose of Vega 20, the processor of Radeon 7, is a compute accelerator for data centers.

That is the primary reason why in Radeon 7 we are seeing this massive compute capability of 912 GFLOPS (1:2) [FP64 (double) performance6].

This ain’t to be used for gaming son. It is supposed to be used by those freaky scientists for computing something really freaky.

I am just kidding. I love scientists. But you get the point, right.

What if AMD goes the good guy route ?.

Let’s consider another scenario.

What if AMD can take up the R & D cost to re-create Vega 20 so that it supports GDDR6 and cuts of FP64 ?.

Well in that case, I don’t think that the 295 Watts TDP will go down by a lot.

That is ofcourse after giving due consideration to how much bandwidth GDDR6 will look to provide.

Will it make it worse, I can’t say for sure.

So the 295 W TDP is unlikely to change if they made a gaming oriented dye out of Vega 20.

By the way I am using all these specs from here if you would like to see them yourselves.

There is one other thing that needs to be considered to keeping the business strategies in mind.

Putting things into perspective,

Unfortunately, it seems we still think that the only way for businesses to strive is via competition and not by collaboration and innovation.

Sometimes, I think we are surrounded by all these fancy companies, comforts and conveniences but still our thinking is still suspended in survival mode.

Anyhoo, we cannot do anything about it YET.

If if if, considering a hypothetical situation where AMD takes the pain of investing in R & D for re-creating Vega 20 chip that provides GDDR6 support.

And we also slash of the unnecessary FP64 as far as gaming scenario is considered.

You are looking at a massive stock of graphic card that may or may not sell all that well.

The reason for that is the competitor card from the other side i.e., Nvidia RTX 2080 won’t have any problem in slashing down the prizes to compete well with this newly formed card.

Because of the oh so simple fact that they are holding the consumer grade graphic card market situation Venn diagram closely tucked towards themselves. That is putting it lightly. :P.

So with this kind of market situation in mind, it will just be a massive “Let’s just not do it vibe” prevalent in the production department of AMD.

And putting up my previous statement into perspective that we are still somewhere stuck in survival mode and most decisions are just triggered by “Does it make me money or not ?.”, going out of the way to just cater to what the consumer really really wants takes a hit.

It’s not a question of right or wrong.

Actually, it’s not even a question of what should be done.

It’s like is there an active effort put in from all the sides in discussion to produce something of fantastic value to the customer.

Because customers are the king.

Anyways, as I was saying that just creating a card that may not have a big sale value due to the reasons we just looked into is not looking a viable option at this moment.

Now you may ask what about the Radeon 7 ?. Why will it sell ?.

Incredible question actually.

What I think the driving factor in creating a Radeon 7 card is just this.

AMD is anyway are creating a Vega 20 chip.

Vega 20 is designed specifically for high computational workloads mind you, and let’s say Radeon 7 doesn’t make much money from the gaming segment that isn’t the end of Vega 20 clientele now is it ?.

So it doesn’t hurt AMD because Vega 20 was primarily brought into existence because of its compute accelerator power.

And as long as the data centers and scientific research centers keep on gobbling Vega 20 they don’t necessarily need to have Radeon 7 gobbled up as well.

Now consider this situation where you have effectively subtracted the data center clientele from the Vega 20 by stripping it down to have no FP6 and GDDR6 support.

It doesn’t guarantee that the upfront cost of development will be fulfilled by sales which will be reduced to about 500$ because of the strong hand of Nvidia.

That will leave AMD in the typical Meh situation now wouldn’t it.

So now I think you will understand why AMD hasn’t bothered to create a Radeon 7 just for gamers :P.

Oh don’t be sad.

World doesn’t stop here. As time progress, tech is only going to bloom. Irrespective of these non-nonsensical tug of war we educated people sugar coat as strategies :P.

About the 7 nm chip,

The Vega 20 is made up on 7 nm lithographic process.

And like all tech, with time the process will mature and get cheaper and with that the dye will get cheaper.

Now some great things that are evolving as we speak is many apps especially those that are used in data center applications for accelerator purposes.

There are many companies apart from Nvidia and AMD that are showing interest in using HBM2.

So what this means is as HBM2 featuring modules start finding their way into various uses, we will see HBM2 price coming down.

Well, Rajput I see all of that but 700$ for this card ?,

I know its sad.

It’s a calculated business strategy which I feel give them monies to AMD but it sure is true.

But consider this fact also. The ridiculous prices that we may have on our hands especially for cards like RTX 2080TI which are clocking at around 1200$, are we somewhere responsible for this.

I am not saying only we are.

But what happens when we buy 1000$ cards such as the Titan X Pascal which in no way is a compute oriented card.

If I bring you back to my original statement, that depicts on what is the basis of operation for most companies today.

And if it is just “how do I make money ?” then using Titan X Pascal you just got benchmarked as to how much you will pay for something better.

Now they have us by our tail.

We are never satisfied with what we have and the fact is we never will be. But does that mean that urge will just push us to do just about anything ?.

Companies must exist to serve consumer needs not to analyze consumer behavior and extract what they can.

Now do we jump to conclusion that some company is bad.

No.

We don’t know within ourselves that we are unconsciously driven.

If we know what we are and what is it precisely we want. Companies yield to you.

But we are already slaves to satisfy something that never can be satisfied.

That’s not even the bad part.

We know it can’t be satisfied but we still chose to be the same anyway.

So as I was saying Titan XP doesn’t seem top have any features such as the FPG64 that will clearly demarcate it as a compute capable card.

It is essentially a gaming card and its successful price point clocked at 1000$ just proved we are ready to buy cards at that price point, no problem.

So if this consumer unconscious behavior continues, don’t be surprised if with the passage of time the probable cards as in the 4080 Ti’s get clocked at 3500$.

You can look into the benchmarks compiled over by HardOCP here.

What it looks like from the benchmarks is that we can expect anywhere around 28-29% improvement over RX Vega cards.

So the existence of Radeon 7 the way it exists with respect to price point as such should not shock you now.

Having seen the groundworks and frameworks of how the things are, what can we perceive from here ?.

It seems like well it is like that actually, AMD is not trying to break any ground into consumer grade graphics market segment.

It is taking what it has and knows will work and is rebranding it to see how things progress from there.

So what will break grounds ?. Can grounds even be broken ?.

If grounds could not be broken it won’t be called ground at all now will it ?.

Okay that was kinda stupid statement. But you get the point right ?.

AMD which was an underdog broke many grounds that had not seen rain from a very long times by introducing Ryzen.

And with the wave of tech we are seeing all over the place. The room to break grounds is always there.

But the question comes down to this fundamental basis. Are we ready to look beyond the limitations that we have created on ourselves ?.

Are we ready to really invest our lives into creating something truly fantastic that enriches individual consumer’s lives ?.

Or are we so caught up in our own cycle of how to beat the other ?. Let me tell you this say you somehow managed to beat the other.

Don’t live in this folly that you won.

And the most fundamental thing that we as consumers need to understand.

Are we so unsatisfied within ourselves that we will just throw our bucks to get something flashy that provides something more at whatever price it is quoted ?.

If we as individuals aren’t capable of looking within ourselves from this perspective.

Let me tell you this very clearly, the basic way everything is working, nobody will get of their ass to break any grounds.

I love you all.

I will see you in the next one :).

Namaskaram _/\_ 🙂