So apparently, the Intel/Radeon marriage is a thing that's happening

[Notional]

Just now, LAwLz said:

I wonder where he is going. Apple again? They are designing their own GPUs now so it would make sense.

 

 

I've heard several people say this but I don't get why. Raja re-joined AMD in 2013.

He was at AMD for years before Fiji was done, and even longer before Vega was done. AMD were not sitting on a finalized GPU for 2-4 years before releasing it.

It takes at least a full year from finished architecture to a product on the shelf in stores. If Raja had any influence on the architecture of Fiji I doubt it was a lot. Then again, if he was responsible for using HBM on Fiji, then he is definitely to blame for most of the problems with that chip. AMD's focus (fixation) on HBM has sabotaged their high end cards a lot.

[ARikozuM]

Just now, djdwosk97 said:

*Insert two GPUs duct taped together* 

 

  Hide contents

I spent the last five minutes looking for a picture, but I just couldn't find one.....so use your imagination. 

 

You mean the R9 295X2? That thing had to be welded together. 

[dizmo]

25 minutes ago, JobbelsNL said:

I am also team blue dont worry but since games are getting more optimalized by the day team red will slowly start winning with more Threats and cores

Haha, no. The consumer market isn't even that big when you compare the corporate one, where AMD doesn't really have a standing at all.

12 minutes ago, ARikozuM said:

They could do it if they targeted the low- and mid-tier (entry and mainstream to some) rather than the high-end for a while. The sales could buy them the time required to get a CCX-like GPU (Crossfire 2.1) that uses the same cores for the high-end. I'd be stoked to see the Ryzen engineers make something out of Polaris chips.

I don't think that'd help them that much. Sure they could own it, but the profits are a lot smaller down there, and really how many people are buying 550/560 GPUs? I'd imagine not many.

[PocketNerd]

2 hours ago, Dylanc1500 said:

It's like going to Walmart, you have the electronics section and clothing section. Entirely different products but still Walmart. (I'm im Walmart currently so it was first to mind lol)

Except those departments don't run independently of Walmart. Whereas Radeon Technologies Group has at least some independence.

Besides, who in their right mind would turn down a deal like this because of some weird grudge?

[AMDNeko]

Personally I think Lisa Su will be a better head to RTG than Raja so no major loss..

[Notional]

2 minutes ago, PPCs-Kat said:

Personally I think Lisa Su will be a better head to RTG than Raja so no major loss..

Doubt it. She said it herself that it is too much for her to oversee. She will probably do great as an interim, but I'm sure AMD will find a proper replacement soon.

[ARikozuM]

Just now, dizmo said:

I don't think that'd help them that much. Sure they could own it, but the profits are a lot smaller down there, and really how many people are buying 550/560 GPUs? I'd imagine not many.

The top selling range for graphics cards is ~$100 to ~$350 (from past years). If AMD can own that space, without a question like 1070 versus 580 (AMD can literally just drop their naming scheme down by 10), they would get a very large portion of the market. 

 

I think they could make two single chips like the 560 and 570 (labeled as 550 and 560) leaving the 570 and 580 to be two of those GPU cores. 

 

Obviously we, the forum, aren't a good representation part of that market. Also, I'm surprised how many of us have 1080's. 

[Techicolors]

19 minutes ago, LAwLz said:

I wonder where he is going. Apple again? They are designing their own GPUs now so it would make sense.

is it for the next Mac Pro that's rumored to have a Vega chip? 

[LAwLz]

20 minutes ago, Notional said:

It takes at least a full year from finished architecture to a product on the shelf in stores. If Raja had any influence on the architecture of Fiji I doubt it was a lot. Then again, if he was responsible for using HBM on Fiji, then he is definitely to blame for most of the problems with that chip. AMD's focus (fixation) on HBM has sabotaged their high end cards a lot.

Even if we assume that that's true, then he still had a lot of time. He didn't sit and roll his thumbs for an entire year. Depending on when development started he might even have been involved in the beginning, worked for Apple in the middle of development, and then gotten back for the last year.

He also worked on the earlier versions of GNC which Fiji was heavily based on.

 

I think saying that he had very little to do with Fiji is a very big assumption that's most likely wrong.

 

 

4 minutes ago, Technicolors said:

is it for the next Mac Pro that's rumored to have a Vega chip? 

Apple don't really need Raja for that. Even if they needed to do some custom design (for like the slot design they used in the trash can) it would be cheaper to just include that work in their contract with AMD.

My guess is that if he goes to Apple it's to design GPUs for their mobile SoCs.

[Shreyas1]

50 minutes ago, JobbelsNL said:

AMD needs a year or 2 then intel will be overgrown by AMD

Intel makes more than just desktop cpus

[Dylanc1500]

8 minutes ago, PocketNerd said:

Except those departments don't run independently of Walmart. Whereas Radeon Technologies Group has at least some independence.

Besides, who in their right mind would turn down a deal like this because of some weird grudge?

I wasn't going to go too in-depth as I said, I was in walmart at the time. My point was more so that many companies have different divisions within them, even Intel themselves has division that operate separately. "Radeon Technologies Group" is a name that AMD applied to a specific division within the company. A better example would be Intel and their divisions for CPU, Storage, Modems, and etc or even Microsoft with Windows, Xbox, Cloud, Surface and so on. The company I work for even has many different divisions but all of them operate as one entity and company.

 

You are correct though it was a chance for AMD and Intel both to make money and most likely AMD had the lower bid so they went that route. With the way Intel has it set up they theoretically could put any GPU with it, including NVidia.

[Jaredtheamazinglewis]

http://www.game-debate.com/news/24011/intel-officially-reveals-kaby-lake-g-cpus-with-discrete-amd-radeon-graphics

 

 

I just can't believe it but it's true.

[Jurrunio]

Wasnt it posted already?

[Thinkfreely]

Im so going to do a thing about this! Also cashgrab is real.

[Trik'Stari]

10 hours ago, cj09beira said:

ya, the feasibility of navi in my mind really needs hbm3 to be ready for action because some cards will end up with only one hbm stack, which needs to be as fast as possible.

vega is top of the line the issue is that its bottleneck in other parts of the pipeline while rendering games, its compute perf is actually really good, been basically as fast as p100 while being a smaller die, it seems amd is focusing more on the markets where they think vega will sell for its merits more often.

not exactly intel is developing its own igpus the deal with nvidea (now with amd) was just so that intel doesn't get sued, as right now amd and nvidea have so many patents that its impossible to make a  fast efficient gpu without inflicting on those patents 

It's not top of the line, compared with the 1080ti.

 

I don't mean top of AMD's line, I mean top of the market line. As in "the most powerful gaming card out today". Perhaps I should have been more specific. Also non of it matters when the card just isn't available.

[vanished]

topics merged, again

[Trik'Stari]

Let's just hope they can find someone who actually knows what the company needs to do.

 

Stop talking and marketing, and just make a great friggin GPU. Not Vega, just make a badass R9 590 or 690 or whatever fucking number you want to call it, and have it actually be a realistic competitor to the 1080ti or whatever Nvidia has out when the AMD GPU comes out.

[Dylanc1500]

So I guess the real question comes down to, are they going to have the gall to put a vega 64 die and the 18 core die on one.

[vanished]

12 minutes ago, Dylanc1500 said:

So I guess the real question comes down to, are they going to have the gall to put a vega 64 die and the 18 core die on one.

I highly doubt it.  That kind of hardware in a laptop is just unheard of even in the thickest of beasts

[cj09beira]

9 minutes ago, Trik'Stari said:

Let's just hope they can find someone who actually knows what the company needs to do.

 

Stop talking and marketing, and just make a great friggin GPU. Not Vega, just make a badass R9 590 or 690 or whatever fucking number you want to call it, and have it actually be a realistic competitor to the 1080ti or whatever Nvidia has out when the AMD GPU comes out.

amd really needs marketing as most people are oblivious to their products 

1 hour ago, ARikozuM said:

They could do it if they targeted the low- and mid-tier (entry and mainstream to some) rather than the high-end for a while. The sales could buy them the time required to get a CCX-like GPU (Crossfire 2.1) that uses the same cores for the high-end. I'd be stoked to see the Ryzen engineers make something out of Polaris chips.

navi is that, and no, they should not focus on the lower end as their main problem is lack of a good halo product and mind share.

56 minutes ago, Notional said:

It takes at least a full year from finished architecture to a product on the shelf in stores. If Raja had any influence on the architecture of Fiji I doubt it was a lot. Then again, if he was responsible for using HBM on Fiji, then he is definitely to blame for most of the problems with that chip. AMD's focus (fixation) on HBM has sabotaged their high end cards a lot.

hbm isn't keeping back amd cards at all, its actually helping to reduce power consumption by a good amount, the problem with gcn right now is that at 64 cus the rest of the gpu can't keep up, after 290x the rest of the gpu hasn't been increased in any significant way because they need to redo a big part of the gpu to allow for more than 4 pipelines, they probably would need 6 to make vega a great gpu/ 8 smaller ones,

and also most gpus from amd right now are bottlenecked by memory,

 

 

[AresKrieger]

I'm not remotely surprised in the least you don't take extended leaves for vacation after all, now I don't think this will change anything as AMD is in a situation on the gpu market side I do not think they can escape, as Nvidia unlike intel plans ahead with performance so they can quickly respond to whatever amd does.

[Dylanc1500]

Just now, Ryan_Vickers said:

I highly doubt it.  That kind of hardware in a laptop is just unheard of even in the thickest of beasts

Hey, I never said it had to be in a laptop lol. I just want to see some try to cool it with them that close together.

[sazrocks]

29 minutes ago, Ryan_Vickers said:

topics merged, again

lol, I have now gotten six separate notifications just for this thread.

[cj09beira]

50 minutes ago, LAwLz said:

Even if we assume that that's true, then he still had a lot of time. He didn't sit and roll his thumbs for an entire year. Depending on when development started he might even have been involved in the beginning, worked for Apple in the middle of development, and then gotten back for the last year.

He also worked on the earlier versions of GNC which Fiji was heavily based on.

 

I think saying that he had very little to do with Fiji is a very big assumption that's most likely wrong.

 

 

Apple don't really need Raja for that. Even if they needed to do some custom design (for like the slot design they used in the trash can) it would be cheaper to just include that work in their contract with AMD.

My guess is that if he goes to Apple it's to design GPUs for their mobile SoCs.

so i did some light duckduckgoing (like googling) and the fury x was released in 24/06/2015 and raja joined in 22/04/2013

that is 2 years before launch, we know that the gpu architecture is finished around a year earlier than the release of the final product, (given by the announcement of Vega's design finished around june 2016) so that means he had only a year to have any effect in the design at all, and considering the fix involves big design changes by increasing the amount of pipelines from 4 to 6/8 its not something they do in the last minute, no excuses for vega not having them though, and i would like to know why this wasn't done, maybe not enough R&D 

[cj09beira]

9 minutes ago, AresKrieger said:

I'm not remotely surprised in the least you don't take extended leaves for vacation after all, now I don't think this will change anything as AMD is in a situation on the gpu market side I do not think they can escape, as Nvidia unlike intel plans ahead with performance so they can quickly respond to whatever amd does.

their plan is to do the same they did with ryzen and have multiple gpus connected together, the difference is that in this case it actually kills two birds with a single strike as it allows for increased amounts of rops/cus, hbm3 will have a big impact on how good each die can be as they probably going to do what they did with ryzen where each die has one hbm3 stack,