AMD will launch AM4 platform in March 2016 says industry source

[Shivansps]

I don't remember AMD management complaining about Intel aggressiveness on the bottom market when ULV Celeron and Pentium were launched, but I do remember Devinder Kumar doing so by the time of Bay Trail launch. I also saw cat laptops disappear and be replaced by Bay Trail laptops two quarters after Bay Trail launch.

Thats because BT killed off AMD attempts to recover with Temash, they even wanted to get inside tablet market with the A4-1200.
But AMD had already lost the market to Celeron ULV by that time.

 

[ShintaiDK]

Temash was never tablet potential and AMD didn't back it up either. BT or not, it would fail.

1Ghz dual core at 3.9W and 225Mhz IGP on a single channel. Rest of the line was 8W.

 

[Shivansps]

UlV Celerons weren't as great since they were as fast as the cat family... the Cat family was killed by the worst piece of sh** ever released... the MONOcore AMD E1 2100, who was lagging on NES emulator, far worse than VIA C3 and was infamous worse than the C50 and used in almost every AMD laptop. As infamous as Superman 64 and Hong Kong 97.

Nah, slow as hell procesors existed since Bobcat was launched, E-240, C-50, C-30, etc...

Intel responded to Bobcat that with the Celeron 847, them AMD launched E-450, C-60, C-70... then Intel started to speed up the Celeron with 867, 877 and 887.

After that, Ivy based ULV Celeron started to show up, and with the 1007U and 1017U it was game over, before BT launched it was already very hard to find any AMD cat based product, OEM just started to use the trash of the trash, like the E1-2100 just to offer very very cheap products.

 

[cbn]

That is true, but AMD can do this with 2C 384sp and Athlon x 4 Bristol Ridge processors.

To hell with 1m/2t desktop chips on AM4.

1M/2T harvested Bristol Ridge chips are going to exist though.

And with a "big" 384sp (or even 448sp/512sp) iGPU they really need the higher TDP headroom of a desktop platform. This in order to get enough clockspeed on the CPU to feed the larger iGPU.

 

[Shivansps]

Temash was never tablet potential and AMD didn't back it up either. BT or not, it would fail.

1Ghz dual core at 3.9W and 225Mhz IGP on a single channel. Rest of the line was 8W.

it was a good product, until BT showed up, they attempted to refine the Z-60, the problem was that BT ZXXXX series was tailored for tablets, and there was no way that a Temash could match that.

Asus was one of the OEMs that fell intro the trap, im sure the T100 was always meant to have the A4-1200, they later decided to get rid of the A4-1200 by selling them on 15" notebooks.

 

[dark zero]

Nah, slow as hell procesors existed since Bobcat was launched, E-240, C-50, C-30, etc...

Intel responded to Bobcat that with the Celeron 847, them AMD launched E-450, C-60, C-70... then Intel started to speed up the Celeron with 867, 877 and 887.

After that, Ivy based ULV Celeron started to show up, and with the 1007U and 1017U it was game over, before BT launched it was already very hard to find any AMD cat based product, OEM just started to use the trash of the trash, like the E1-2100 just to offer very very cheap products.

But Bobcat can be OCed somehow.... but Temash... Temash was the MASSIVE dissaster from AMD.

 

[DrMrLordX]

1M/2T harvested Bristol Ridge chips are going to exist though.

At the bottom of the barrel. Or on FP4 for craptops/awful AiOs.

 

[cbn]

1M/2T harvested Bristol Ridge chips are going to exist though.

At the bottom of the barrel. Or on FP4 for craptops/awful AiOs.

Assuming AMD is able to get all the nice Bristol Ridge dies/yields (eg, 4C + 512sp and 4C + 384s) being used by OEMs in 35W FP4 format, I don't think there would be a perceived conflict with having a 2C Bristol with a good deal of iGPU enabled (384sp, 448sp, 512sp) on AM4.

So with Bristol Ridge 2C and 384sp/448sp/512sp why use that on the TDP limited FP4?

Don't you think that would be excessively limiting? (I would want the highest clocked two cores I could get if dealing with 384sp/448sp/512sp for gaming)

 

[cbn]

For 35W FP4, I would think 2C + 192sp (Stoney Ridge) works though.

 

[The Stilt]

Intel has the same memory bandwidth problem as AMD, only their $300-400+ iGPUs Iris Pro have eDRAM, the rest are only using dual channel DDR-4. At the same memory bandwidth AMD is still 30-50% faster than current Intel Skylake. As you said next year Bristol Ridge will increase performance by lets say 10-20% ???
But in 2017 with ZEN APUs at 14nm, high end APU SKUs will definitely need HBM or any other high bandwidth memory otherwise they will be under performing like crazy.

Bristol Ridge top SKU for FP4 should be around 20% quicker in 3DMark Fire Strike for example, if allowed to run without TDP restrictions. It will be a 35W TDP chip with cTDP up to 45W. The TDP however won´t allow it to run at anywhere near the nominal speed. The peak performance (with either CPU or the GPU stressed) will be impressive, however the real world (e.g gaming) improvements will be dulled by the TDP limit.

To max out a FX-8800P (3.4GHz CPU, 800MHz GPU) you´ll need around 65-75W power budget D:

 

[coercitiv]

The peak performance (with either CPU or the GPU stressed) will be impressive, however the real world (e.g gaming) improvements will be dulled by the TDP limit.

To max out a FX-8800P (3.4GHz CPU, 800MHz GPU) you´ll need around 65-75W power budget D:

Mobile systems tend to reach GPU or overall cooling bottlenecks faster than APU max TDP bottleneck, with the sole exception being ultra mobile solutions (15W TDP).

For illustration purpose only
My CPU can reach it's rated 45W TDP while fully loaded (CPU load only), yet I can easily limit it's TDP to 20-25W and observe no subjective drop in performance while playing games on dGPU (GeForce 750M w/ DDR3 aprox. 50W TDP). So I could argue the iGPU performance at 20-25W should be close to GeForce 750M before I were to worry about my CPU being TDP limited.

 

[mrmt]

Thats because BT killed off AMD attempts to recover with Temash, they even wanted to get inside tablet market with the A4-1200.
But AMD had already lost the market to Celeron ULV by that time.

I don't think AMD had any illusions regarding Temash, especially because their cost problem vis a vis their ARM counterparts was even worse than Bay Trail, and since AMD didn't have a billion in spare cash lying around...

 

[Fjodor2001]

It is supposed to be the other way around, isn't it? If some cheapskate site is saying that AMD official information is no longer valid then they should put either credible sources or have a lot of credibility to put it at stake, this site has none of it and Kumar does not have to comment on every pie-in-the-sky AMD news.

Well, we haven't received any comment from AMD, so we can't say they are denying it. That's the point. It's not uncommon for companies like Intel and AMD to comment on these issues. And Reuters etc is no "cheapskate site". It's apparently credible enough for them to publish it. So calling it "baseless" like you did is outright ridiculous and in fact a lie.

I'm not questioning that Samsung is manufacturing chips for AMD, but that AMD was free to make that choice and will control any allocation at all. Given the previous disclosures about the WSA, that would make a departure of the previous terms, and since AMD didn't disclose any amendment as of lately then this possibility is highly unlikely.

So how exactly do you think the setup would work otherwise? Who would otherwise determine where the wafers get produced (at Samsung or GF) and based on what conditions?

Also, I get a feeling you have misunderstood my post to begin with: Note that AMD will be producing wafers at TSMC too (but on 16FF+ process, and for other non-Zen chips). So then that should conflict with the WSA too according to you? Because note that I'm not saying that wafers produced at Samsung necessarily should be counted as part of the WSA. I'm just saying that AMD do not have to be tied to producing all 14nm LPP wafers at GF (which is the process they have selected for Zen). So they can start producing them at Samsung, and then later quite easily move production to GF since they both use the same 14 nm LPP process. Didn't you read my previous post on this where I quoted my previous statement saying "AMD will likely prefer to produce the 14 nm chips at GF anyway, since they have a WSA with GF"?

 

[Fjodor2001]

I am also interested in this.

@ShintaiDK/mrmt: What are your sources to back up your claims?

Third attempt at getting an answer to this. Anybody home?

Since neither of you have responded I take it you do not have any sources for your claims? I.e. the claims maddie requested sources for in this post.

Then you accuse others that provide multiple well regarded sources backing up their info of doing "baseless speculation". Talk about double standards.

 

[maddie]

Third attempt at getting an answer to this. Anybody home?

Since neither of you have responded I take it you do not have any sources for your claims? I.e. the claims maddie requested sources for in this post.

Then you accuse others that provide multiple well regarded sources backing up their info of doing "baseless speculation". Talk about double standards.

ShintaiDK/mrmt appear to be misleading everyone.

After a day+ of no response, I have to assume both of you have no hard evidence as to the superior properties of TSMC 16nm process vs the Samsung 14nm one.

When you make big claims and then ignore requests for data, you only expose yourselves as unreliable for truthful information.

 

[ShintaiDK]

Ask yourself what performance products is made on 14LPP. Even A9X is TSMC only. Xilinx is TSMC. SPARC CPUs are TSMC.

As I said before, look at the customer base.

I know 14LPP is praised to the sky and beyond due to Zen. but even AMDs GPU division isn't fond of 14LPP. And it seems they only use it for low end products due to the WSA.

 

[Fjodor2001]

Ask yourself what performance products is made on 14LPP. Even A9X is TSMC only. Xilinx is TSMC. SPARC CPUs are TSMC.

As I said before, look at the customer base.

I know 14LPP is praised to the sky and beyond due to Zen. but even AMDs GPU division isn't fond of 14LPP. And it seems they only use it for low end products due to the WSA.

The are multiple reasons for why various customers chose different foundries. Cost, volume, yields, business relationships and so on. Also, 14LPP is a very new process, and the customer base may change over time. So what you write above does not prove anything.

To sum it up, I take it you had no solid data to back up your claim that "TSMCs 16FF+ is vastly superior to 14LPP" after all, only speculation?

 

[PPB]

Ask yourself what performance products is made on 14LPP. Even A9X is TSMC only. Xilinx is TSMC. SPARC CPUs are TSMC.

As I said before, look at the customer base.

I know 14LPP is praised to the sky and beyond due to Zen. but even AMDs GPU division isn't fond of 14LPP. And it seems they only use it for low end products due to the WSA.

Your argument is invalid since these contracts arent only made based on the performance metrics of the processes disputed. You probably wont ever know the performance metrics the contractors asked to the fabs to see if their process are eligible for their designs, neither the economical terms to make those contracts happen or the ramp up timeframes.

So basically, you dont have proof at all, just guesstimations based on your biased perception on how the fab industry works. The "x process of y company got more contracts, it must be better in all metrics!!1one!1eleventy" mantra is just a childish attempt to grasp how said industry works.

 

[maddie]

Ask yourself what performance products is made on 14LPP. Even A9X is TSMC only. Xilinx is TSMC. SPARC CPUs are TSMC.

As I said before, look at the customer base.

I know 14LPP is praised to the sky and beyond due to Zen. but even AMDs GPU division isn't fond of 14LPP. And it seems they only use it for low end products due to the WSA.

It seems this is the problem with so many today. Everything is confrontational and argued to the last as if we're attorneys in court.

Is there no room for a good hard discussion/debate without the spite and false statements?

Argue something that disagrees with someones core belief and you get a few here going rabid/manic.

What a mess. Some can't say "I stand corrected" or "I was wrong"

To keep in mind, "I never learned anything from an argument I won"

 

[Sweepr]

To sum it up, I take it you had no solid data to back up your claim that "TSMCs 16FF+ is vastly superior to 14LPP" after all, only speculation?

Even if he provided any data you'd still question him, asking for more evidence. We all know you're hyping this process because AMD will be using it.

 

[Arachnotronic]

Even if he provided any data you'd still question him, asking for more evidence. We all know you're hyping this process because AMD will be using it.

100% true. The only reason that there seem to be so many "Samsung" supporters on this board is that these people think that Samsung is AMD's ticket to achieving process node parity with Intel.

 

[deasd]

Even if he provided any data you'd still question him, asking for more evidence. We all know you're hyping this process because AMD will be using it.

100% true. The only reason that there seem to be so many "Samsung" supporters on this board is that these people think that Samsung is AMD's ticket to achieving process node parity with Intel.

Very funny forum. Everyone is arguing a person's position rather than technology or fact. Now I really understand why studying psychology is much more interesting in AT. I'd better be a bystander afterwards.D:

 

[mysticjbyrd]

Even if he provided any data you'd still question him, asking for more evidence. We all know you're hyping this process because AMD will be using it.

If TSMCs 16FF+ was superior, then why aren't they using it? AMD has all the data we lack. Are we to believe they are just that incompetent, or is the argument the 14 LPP is being used, because it's better all around chip for mobile that sacrifices high end performance?

 

[ShintaiDK]

AMD is using it for higher GPUs.

The only reason AMD use 14LPP anywhere is the WSA. And that one first expires no sooner than March 2nd, 2024.

 

[mysticjbyrd]

AMD is using it for higher GPUs.

The only reason AMD use 14LPP anywhere is the WSA. And that one first expires in 2024.

I didn't know they were still under that agreement.
2024... l didn't know it was that bad! GF is going to kill AMD's cpu division at this rate, even if Zen can be somewhat competitive.


I am guessing you guys already knew this, but it's news to me.


April 17th, 2015 at 3:58 am - Author Anton Shilov
Advanced Micro Devices on Thursday revealed that it had signed a new wafer supply agreement (WSA) with GlobalFoundries for 2015. Based on the disclosed terms of the agreement, AMD will significantly reduce purchases from GlobalFoundries, which indicates that the company projects further declines of PC-related product shipments this year.
Under the terms of the new agreement, AMD expects wafer purchases from GlobalFoundries to be approximately $1 billion this year on a take-or-pay basis, which is a decline from 2014. Last year AMD planned to spend $1.2 billion on purchases from GlobalFoundries, but its actual purchases from the company in 2014 were approximately $1 billion due to lower fourth quarter purchases.

http://www.kitguru.net/components/g...alfoundries-expects-further-apu-cpu-declines/