AMD CPU speculation... and expert conjecture

[iceclock]

yeah maybe in arms and smaller microchips. but mainstream unshur of that.

 

[logainofhades]

At long last we finally have to answer that we have been waiting for. There will not be an BD FX successor for 2014. We will have to keep pushing our 8350s to higher clocks. I personally will be shooting for 5.0GHz base, 5.2GHz turbo, with cool and quiet and other power saving features enabled.

[/url] [/img]

I hope that maybe they make some architectural improvements instead of just making more space heaters like the FX 9xxx series. Some tweaks to improve performance would be nice.

 

[griptwister]

For those expecting a high end AMD APU because of the FM2+ gigabyte sniper board:

http://www.overclock3d.net/articles/cpu_mainboard/gigabyte_g1_sniper_z87_first_look/1

 

[de5_Roy]

this is the fm2+ sniper
http://www.newegg.com/Product/Product.aspx?Item=N82E16813128653
the cpu socket is a giveaway.

 

[juanrga]

noob2222, de5_Roy,

The article was clear about Temash being replaced. It failed about what would be the replacement.


gamerk316,

yes the performance quoted by AMD for the APU includes the GPU, because APU=CPU+GPU.

This Berlin server will be used for HSA workloads, because that is for what an APU such as Berlin has been designed.

 

[de5_Roy]

noob2222, de5_Roy,

The article was clear about Temash being replaced. It failed about what would be the replacement.

the article is not being questioned here, since it only dealt with rumors and the writer's personal opinions.
"temash being replaced" is a blanket statement. that succession is natural in the course of lifecycle. but that was not the claim you made. please stop refering to blanket statements. instead, present the proof of your claim of temash getting replaced by an arm soc.

 

[noob2222]

You never know; Via could shock us all some day and release a high end X86 chip for the desktop. 😛

apparently with Mantle and a GCN card, it will be ...

 

[Ags1]

Well, I predict that Haswell is a failure, and Intel are going to replace it in six months or so... 🙂

Watch this space!

 

[juanrga]

Spoiler

i)
They didn't miss 8% of CPU target, because AMD never revealed the frequencies. In my BSN* article I included a table with three different combinations of frequencies. Final Kaveri CPU freq. is only a 3% behind one of them.

The same about the GPU frequencies. AMD never gave them.

ii)
I already explained you that only the GDDR5 version of the HD7750 has 900MHz clocks. The version with DDR3 has 800MHz clocks. Kaveri has 720MHz. All them bulk.

The large drop in iGPU performance from original expectations is explained by AMD dropping the GDDR5 support. The original 1050GFLOP Kaveri was considering GDDR5 memory. BSN* has copies of the original docs reporting GDDR5 support. I already mentioned that someone has said me that the Kaveri die has a disabled GDDR5 IMC. Which means that the decision to abandon GDDR5 was taken very late.

iii)
Trinity and Richland are 100W. Kaveri is limited to 95W TDP. Limit both Trinity and Richland to 95W max and say me about how many you drop frequencies.

My bet is that a 95W rated Trinity would be ~3.7GHz and Richland ~3.9--4.0GHz

iv)
Kaveri has larger and powerful iGPU than Trinity/Richland. Those 100W TDP for the APU are distributed between CPU and GPU.

Take Trinity/Richland and substitute their iGPU with a Kaveri-level iGPU. If the total TDP continue being of 100W, you have to cut the CPU frequencies by some extension.

v)
Sure that Kaveri would achieve higher clocks in SOI. Has someone said the contrary here? NO But are you aware that Glofo has not ready 28nm SOI? AMD decision to go bulk has been a success. I don't understand why you guys don't accept it.

vi)
We don't know turbo frequencies nor its aggressivity. What if AMD follows Intel way of lower base clocks and higher turbo clocks?

The last rumor is that AMD is considering >4.0GHz turbo.

vii)
In my BSN* article I didn't considered a 20% IPC gain over PD, but a 20% minus a 5% for unknows. Palying safety, this means that my predictions of i5 level of performance are off by about a 3%.

Now, benchmarks leaked after I wrote my article say that SR is a 31% faster than PD and speculate about higher gains in the final silicon

It is still too soon. There are so many unknows.

viii)
It is evident than FinFET is the future. AMD is already tapping out chips on 14nm bulk FinFET and some foundries are running the first tests of 10nm FinFET.

But hey you are the expert that said us that Kaveri was SOI...

I. http://semiaccurate.com/2013/11/12/amd-misses-expectations-kaveri/

AMD did announce targets at the financial analysts day in the summer. 4.0 GHz and 900 MHz were the targets...how else did you arrive at your speculated numbers? You pulled them out of the sky? The GFLOPS number they gave was what you used...remember? Selective memory? Alzheimer's?

II. You're dismissing it because you don't want to admit they missed their targeted performance bracket on bulk. It would make your entire argument wrong.

III. That's right, a 95W Richland part would still have higher clocks on a less advanced process. See something wrong with that? Trinity level base clocks (i.e. 1.5 generations ago) are not acceptable.

IV. You're neglecting that Kaveri is also using GCN which is supposed to consume less power than VLIW4, which is what was on Trinity/Richland. Your argument holds no water.

V. Because it is a failure, and behind the scenes, AMD is basically almost admitting as much without outright saying it missed the performance targets because of process constraints. I want AMD to succeed as much, or more, than you do; however, we should be calling a spade what it is, a spade. Kaveri is a great APU, which tends to be a upper low end part in the flagship SKU. Berlin server APUs/CPUs are aimed at lower power/lower compute demand servers. People who need to crunch data go for 16 core Opterons. That's specifically because they're still better at that than APUs.

VI. That would suit AMD's TC 3.0 setup on Richland, 300 MHz turbo would put TC @ 4.0 GHz, now the issue arises, how often will turbo engage? Most of the time like Intel, or only a small portion of the time like the last gen AMD processors? Even then, it still doesn't mitigate the fact that the 6800k has a turbo of 4.4 GHz.

VII. 3% on the CPU only calculations. They are off as much as 19% in the CPU + GPU possibility calculations. Additionally, you assumed a 4.0 GHz clock, you've now not only lost an extra 3% clock speed, however, you're also 5% shy of what the improvement actually looks to be so far. Which puts your estimates closer to 8% off. Not trying to be a downer, but reality is where we live, and that has to be accounted for.

VIII. No, AMD is beginning to develop for 14nm XM, which is not a FinFET. It could be, though you'd have to design a FinFET to go on 14nm XM substrate (14nm FE with 20nm BEOL). 14nm XM could also be FD-SOI...as STMicro already has some designs for booster tech on that technology as well. Consider that 14nm XM was co-developed with STMicro, they are prepared for whatever it needs to be.

I am not sure why GF has such issues with 28nm booster tech for FD-SOI, personally I think it has everything to do with the delays on 28nm bulk, because you cannot implement BEOL booster tech for FD-SOI if you don't have bulk going yet.

Either way, AMD is being hamstrung by GF over the process technologies they have access to...which is why Kaveri is being made at TSMC, not GF. I expect if they get future FD-SOI tech up and running at acceptable levels, we might see FX successor @ GF. Though currently, that doesn't look likely to happen until GF sort out their issues.

FYI: the only foundry currently using a FinFET is Intel's foundry, and their FinFET is on bulk substrate. That's also why they have had yield issues with the last few CPUs and high overclock variance. Because a FinFET is very complex, and doing it on bulk requires additional mask layers to prevent leakage, making even more issues. Though, I hope Intel doesn't see the light...because FinFET is arguably dead before it got off the ground. Maybe when they chase something past 7nm it will take off...(if FinFET can indeed get to 5nm as some have speculated).

(i) I already commented the S|A article mistakes on S|A forums.

AMD did never announce frequencies. If you read my BSN* article, which you didn't, you will discover how I got the table with different combination of frequency pairs and why selected one pair over the others.

(ii) This doesn't even consider the point.

(iii) Here you miss that Richland was made on mature process, whereas KAveri is first product in the new node. A more rigth comparison is Kaveri vs Trinity.

(iv) Which is the efficiency of GCN compared to VLIW4?

(v) Another evasive that avoids the point. It also misses points (i) and (ii). About Berlin APU, AMD claims "Berlin is cool, and it uses a new Steamroller core from us and delivers tremendous compute and power efficiency" and "When you have a huge amount of compute in a single-socket part, this is ideal for workloads where performance per watt per dollar and compute density per dollar are paramount." AMD also advertise the new Berlin APU as offering 7.8 times more raw performance than a 16 core Opteron for the same power consumption

(vi) As said there is no enough info for discussing this.

(vii) 4.0GHz minus the 5% safety factor imply that it is like assuming 3.8GHz in the benchmarks. Final silicon is 3.7GHz. therefore, I am off by 3%... as said before.

(viii) AMD is moving to 20nm and then FinFETs.

You didn't address any of my points sufficiently, and provided no concrete facts. Only your offbase speculation and marketing slides...which seem to abound in this thread lately.

Your argument that Kaveri is not 900MHz due to bulk has been shown to be without basis. What AMD says in roadmaps, interviews, slides... contradicts what you say, but this is not anything new, true?

Finally, I note that you don't answer my questions (e.g. (iv)). You never do.

 

[griptwister]

this is the fm2+ sniper
http://www.newegg.com/Product/Product.aspx?Item=N82E16813128653
the cpu socket is a giveaway.

What I was trying to say was, Intel Got the cool budget sniper board too. Which means, the product is nothing special and there is no sign of AMD releasing a high end APU for the FM2+ socket.

xD I know the socket is a dead giveaway.

 

[juanrga]

noob2222, de5_Roy,

The article was clear about Temash being replaced. It failed about what would be the replacement.

the article is not being questioned here, since it only dealt with rumors and the writer's personal opinions.
"temash being replaced" is a blanket statement. that succession is natural in the course of lifecycle. but that was not the claim you made. please stop refering to blanket statements. instead, present the proof of your claim of temash getting replaced by an arm soc.

^^^

The article was clear about Temash being replaced. It failed about what would be the replacement.

 

[8350rocks]

GCN is drastically more power efficient than VLIW4, look at how much the TDP got better on the GPUs from HD 6XXX to HD 7XXX. Additionally, for the same thermal envelope they increased performance by a massive amount.

GCN is faster, better at compute, and less power hungry than VLIW4. These are not disputed facts.

 

[de5_Roy]

What I was trying to say was, Intel Got the cool budget sniper board too. Which means, the product is nothing special and there is no sign of AMD releasing a high end APU for the FM2+ socket.

xD I know the socket is a dead giveaway.

ah, now i got it. but gigabyte giving socket fm2+ sniper treatment shows gigabyte's hope on kaveri.
meanwhile, i overlooked this fm2 board. been a while i checked for apu motherboards:
http://www.newegg.com/Product/Product.aspx?Item=N82E16813128566
that's up4. 😱 and it lead me to the second gigabyte fm2+ premium board:
GIGABYTE GA-F2A88X-UP4, $110
http://www.newegg.com/Product/Product.aspx?Item=N82E16813128655
intel's g1 sniper($170, down from $190) costs a lot more than fm2+. and z87 is yet to get a up4 board. i mean, with amd you get similar features at a lower cost. the trade off is cpu performance.

 

[noob2222]

noob2222, de5_Roy,

The article was clear about Temash being replaced. It failed about what would be the replacement.

the article is not being questioned here, since it only dealt with rumors and the writer's personal opinions.
"temash being replaced" is a blanket statement. that succession is natural in the course of lifecycle. but that was not the claim you made. please stop refering to blanket statements. instead, present the proof of your claim of temash getting replaced by an arm soc.

^^^

The article was clear about Temash being replaced. It failed about what would be the replacement.

^^^

The only thing you need to do is admit you were wrong and apologize for being an arrogant a@@ about it.

 

[de5_Roy]

^^^

The article was clear about Temash being replaced. It failed about what would be the replacement.

no it was not, and we won't be discussing the techreport article anymore since you admit that the article doesn't mention the type of replacement (it didn't fail, it just did not discuss replacement). stop trying to divert to the article and instead provide proof to your claim of temash being replaced by an arm soc. You were very clear and certain about amd replacing temash with an arm soc, which does not fit with the article. so the source of your claim must be elsewhere.

 

[juanrga]

Because this go around, Sony is allegedly making a profit per console sold. And if the BOM is only ~$20 cheaper then the retail price, then it follows that each component is being brought at basically break-even value. AMD probably makes the most profit per part, but I can't imagine its in excess of $10 or so.

In other words: Great for market share, not so much for profit. Especially if this locks out production of consumer chips with higher margins.

http://www.zdnet.com/amd-pulls-in-100-for-every-playstation-4-sold-7000023423/

http://www.theinquirer.net/inquirer/news/2308282/amd-makes-more-profit-than-sony-on-the-ps4

Another of your 'predictions' being off by a factor of ten. :sarcasm:

 

[noob2222]

Because this go around, Sony is allegedly making a profit per console sold. And if the BOM is only ~$20 cheaper then the retail price, then it follows that each component is being brought at basically break-even value. AMD probably makes the most profit per part, but I can't imagine its in excess of $10 or so.

In other words: Great for market share, not so much for profit. Especially if this locks out production of consumer chips with higher margins.

http://www.zdnet.com/amd-pulls-in-100-for-every-playstation-4-sold-7000023423/

http://www.theinquirer.net/inquirer/news/2308282/amd-makes-more-profit-than-sony-on-the-ps4

Another of your 'predictions' being off by a factor of ten. :sarcasm:

LOL.

while AMD is likely making around $100 per console sold.
...
The PS4's custom built AMD processor, which combines an eight-core Jaguar architecture CPU with a 1.84TFLOPS Radeon GPU, is apparently the console's most expensive component, costing around $100 all in.

so ... in order to turn $100 profit the cpu doesn't cost AMD anything to manufacture ... Then again, GK was talking about profit, not gross sales.

It could be costing AMD $90 each from TSMC, so that means that TSMC is making $90 for every ps4 sold.

 

[juanrga]

GCN is drastically more power efficient than VLIW4, look at how much the TDP got better on the GPUs from HD 6XXX to HD 7XXX. Additionally, for the same thermal envelope they increased performance by a massive amount.

GCN is faster, better at compute, and less power hungry than VLIW4. These are not disputed facts.

You're dodging the question. I repeat it again:

(iv) Which is the efficiency of GCN compared to VLIW4?

 

[juanrga]

Because this go around, Sony is allegedly making a profit per console sold. And if the BOM is only ~$20 cheaper then the retail price, then it follows that each component is being brought at basically break-even value. AMD probably makes the most profit per part, but I can't imagine its in excess of $10 or so.

In other words: Great for market share, not so much for profit. Especially if this locks out production of consumer chips with higher margins.

http://www.zdnet.com/amd-pulls-in-100-for-every-playstation-4-sold-7000023423/

http://www.theinquirer.net/inquirer/news/2308282/amd-makes-more-profit-than-sony-on-the-ps4

Another of your 'predictions' being off by a factor of ten. :sarcasm:

LOL.

while AMD is likely making around $100 per console sold.
...
The PS4's custom built AMD processor, which combines an eight-core Jaguar architecture CPU with a 1.84TFLOPS Radeon GPU, is apparently the console's most expensive component, costing around $100 all in.

so ... in order to turn $100 profit the cpu doesn't cost AMD anything to manufacture ... Then again, GK was talking about profit, not gross sales.

It could be costing AMD $90 each from TSMC, so that means that TSMC is making $90 for every ps4 sold.

From the second link:

AMD makes more profit than Sony on the PS4
Sony makes $18, AMD makes close to $100

JAPANESE ELECTRONICS FIRM Sony reportedly is making an $18 profit on its Playstation 4 (PS4), while AMD is likely making around $100 per console sold.

 

[Cazalan]

This is also an interesting reading:

http://hezik.nl/arm-ceo-we-can-convince-amd-of-giving-up-x86/

Wow that's a shocker. The CEO of an IP company boasting about how their IP is better.

He goes on about Intel competition and their "monopoly". Well on the ARM front they have 3 bigger competitors. Samsung/Qualcomm/Apple are all BIGGER than Intel. Instead of David vs Goliath, it's David vs Goliath x3. In spite of the perpetual litigation the 3 are still in bed with each other. Samsung buys Qualcomm parts, Apple buys Qualcomm and Samsung parts. The fact that Samsung still buys Qualcomm parts made at competitors fabs tells you the bulk of the profits are being made elsewhere. I'll just stop there because the more you dissect the ARM competition and where they're actually making money, the worse it looks (for AMD).

 

[8350rocks]

GCN is drastically more power efficient than VLIW4, look at how much the TDP got better on the GPUs from HD 6XXX to HD 7XXX. Additionally, for the same thermal envelope they increased performance by a massive amount.

GCN is faster, better at compute, and less power hungry than VLIW4. These are not disputed facts.

You're dodging the question. I repeat it again:

(iv) Which is the efficiency of GCN compared to VLIW4?

How am I dodging the question? I clearly, concisely, explicitly answered it very succinctly.

I don't think I can be more precise with my answer.

 

[griptwister]

What I was trying to say was, Intel Got the cool budget sniper board too. Which means, the product is nothing special and there is no sign of AMD releasing a high end APU for the FM2+ socket.

xD I know the socket is a dead giveaway.

ah, now i got it. but gigabyte giving socket fm2+ sniper treatment shows gigabyte's hope on kaveri.
meanwhile, i overlooked this fm2 board. been a while i checked for apu motherboards:
http://www.newegg.com/Product/Product.aspx?Item=N82E16813128566
that's up4. 😱 and it lead me to the second gigabyte fm2+ premium board:
GIGABYTE GA-F2A88X-UP4, $110
http://www.newegg.com/Product/Product.aspx?Item=N82E16813128655
intel's g1 sniper($170, down from $190) costs a lot more than fm2+. and z87 is yet to get a up4 board. i mean, with amd you get similar features at a lower cost. the trade off is cpu performance.

I just hope we see some decent mATX boards at a better price point. I'd like to have a mini-power PC for when I'm on the road...

Also, everyone see the requirements for this game? These games just keep getting heavier and heavier!!!

http://store.steampowered.com/app/250420

 

[noob2222]

Because this go around, Sony is allegedly making a profit per console sold. And if the BOM is only ~$20 cheaper then the retail price, then it follows that each component is being brought at basically break-even value. AMD probably makes the most profit per part, but I can't imagine its in excess of $10 or so.

In other words: Great for market share, not so much for profit. Especially if this locks out production of consumer chips with higher margins.

http://www.zdnet.com/amd-pulls-in-100-for-every-playstation-4-sold-7000023423/

http://www.theinquirer.net/inquirer/news/2308282/amd-makes-more-profit-than-sony-on-the-ps4

Another of your 'predictions' being off by a factor of ten. :sarcasm:

LOL.

while AMD is likely making around $100 per console sold.
...
The PS4's custom built AMD processor, which combines an eight-core Jaguar architecture CPU with a 1.84TFLOPS Radeon GPU, is apparently the console's most expensive component, costing around $100 all in.

so ... in order to turn $100 profit the cpu doesn't cost AMD anything to manufacture ... Then again, GK was talking about profit, not gross sales.

It could be costing AMD $90 each from TSMC, so that means that TSMC is making $90 for every ps4 sold.

From the second link:

AMD makes more profit than Sony on the PS4
Sony makes $18, AMD makes close to $100

JAPANESE ELECTRONICS FIRM Sony reportedly is making an $18 profit on its Playstation 4 (PS4), while AMD is likely making around $100 per console sold.

Both articles are more or less a copy of each other. they both say that sony is paying amd around $100 for the cpu. Any tool with a small brain in business knows there are costs to make the part.

The processor, which combines an eight-core Jaguar architecture CPU with a 1.84TFlop Radeon GPU comes in, according to IHS, at a round $100.

thats from the other one. pretty much the exact same thing. This term is called GROSS SALES or Gross profit. That means AMD is getting $100 from sony and giving $XX to TSMC. $100 minus $xx = NET PROFIT. The term "making money" generally refers to net profit and not gross sales.

the only way AMD is making $100 is if they are paying TSMC $0.

other than that, the article is pretty amusing.

 

[juanrga]

This is also an interesting reading:

http://hezik.nl/arm-ceo-we-can-convince-amd-of-giving-up-x86/

Wow that's a shocker. The CEO of an IP company boasting about how their IP is better.

He goes on about Intel competition and their "monopoly". Well on the ARM front they have 3 bigger competitors. Samsung/Qualcomm/Apple are all BIGGER than Intel. Instead of David vs Goliath, it's David vs Goliath x3. In spite of the perpetual litigation the 3 are still in bed with each other. Samsung buys Qualcomm parts, Apple buys Qualcomm and Samsung parts. The fact that Samsung still buys Qualcomm parts made at competitors fabs tells you the bulk of the profits are being made elsewhere. I'll just stop there because the more you dissect the ARM competition and where they're actually making money, the worse it looks (for AMD).

Samsung/Qualcomm/Apple are not a monopoly, because don't control ARM. However, x86 is a monopoly controlled by Intel which cross-licenses with AMD and several litigations.

For instance, Nvidia went to purchase a x86 license to make x86 chips and Intel rejected it. That is a monopoly. However, Nvidia went to ARM and got a license to use standard cores and another to design their own curstom cores, independently of how big Samsung/Qualcomm/Apple are.

There are other sides of that monopoly. For instance (correct me if I am wrong), if someone buys AMD, the license to make x86 chips is lost, because it is not transferable.

There are many more legal issues with x86. This range from cache technology to restrictions in the volume of chips that AMD is allowed to do in certain factories. (*)

There are also some interesting legal issues with the custom APUs made for the consoles.

The interesting part of the ARM CEO link is the last paragraph:

the company can choose to develop the processors which integrate ARM and x86, when the developed ARM CPUs is mature, they can gradually decrease the development for x86 processors, or even finally give up X86 to get rid of monopoly and suppression from Intel.

The first stage is ready. Kaveri, Berlin, Beema, and Mullins include a standard ARM core for security features.

The second state is partially in progress. Warsaw is launched for legacy purposes. In the words of AMD, Warsaw is for institutional customers that will be slow on migrating to ARM servers. Berlin is for HSA workloads.


(*) http://www.cpushack.com/2012/09/06/intel-vs-the-world-the-338-patent/

http://www.nytimes.com/2009/11/13/technology/companies/13chip.html?_r=0

 

[blackkstar]

I think some of you are overlooking this.

Take a look at Nvidia. They have wanted to make CPUs forever and the best they can do are ARM units. They would kill for an x86 license. An x86 license means you can offer a product that powers Microsoft Office, modern PC games, and other full featured software suites like Photoshop.

ARM means you get absolutely none of that. You get Angry Birds and Photoshop Elements and Google Documents.

AMD replacing x86 chips with ARM and discontinuing x86 (which is what replacing their line-up implies) means that AMD is either about to throw away their x86 license OR that they're admitting that they can't even come close to competing with Intel and they would rather compete with Qualcomm, Samsung, Apple, Nvidia (in CPUs and GPUs instead of just GPUs), MediaTek, Calxeda, etc.

We are starting to see some design wins for Kabini and Temash now, and you know what? It coincides with the product lifecycle of Windows 8 updating to Windows 8.1.

If anything it's AMD's fault for releasing products in the middle of an OS lifecycle and then waiting until the product is fixed. People are forgetting that the lack of design wins for Temash and Kabini might have had more to do with the fact that Windows 8 was a dud as opposed to the fact that no one wants Temash.

I see a lot of people in this thread that are making conclusions and then looking for the evidence to support those conclusions instead of looking at other possibilities and making conclusions from that.

As an example, note how some people think that AMD is done with dCPU because all AMD is talking about is HSA? They're completely neglecting the fact that there could be no big dCPU Steamroller because it wouldn't work out on 28nm Bulk and SOI is a long ways away.

I'm not saying that that's exactly what's happened, AMD might actually be done with dCPUs. But the fact remains that there are other reasons explaining why there was no October HEDT CPU release this year like what usually happens.