AMD Zen Rumours Point to Earlier Than Expected Release

Darakian said:
Mmmm perhaps. I can see where you're coming from, but 4 cores (or 2 maybe) plus HBM might be a very enticing chip for those without a dedicated gpu.
Click to expand...

You will hit the TDP limit on the chip if you are using enough gpu cores if not then you would have a solution outperformed by simple APU because you could never saturate the bandwidth advantage without hitting ceilings faster (the amount of ram would drive the features also limit the resolution, not using it in this way it becomes expensive L2 cache).

It could work on certain hardware but in the x86 space there are cheaper solutions and AMD does not have the means to just throw money at a HBM APU solution.

I
 
gamerk2 said:
You're not going to have any form of HBM on die, because of the insane power the chip would draw (nevermind the cost to produce said chip). You also won't see HBM as system RAM for the same reason.

HBM for APUs is a pipe dream.
Click to expand...

You would have to be a fool to believe HBM wont make it to an APU. It is the solution to alleviate the biggest negative APUs have.
 
gamerk2 said:
You're not going to have any form of HBM on die, because of the insane power the chip would draw (nevermind the cost to produce said chip). You also won't see HBM as system RAM for the same reason.

HBM for APUs is a pipe dream.
Click to expand...

'Insane power draw'? Are you using Fiji as an example? Because HBM isn't why Fiji consumes as much power as it does.

According to Hynix's Hot Chips '14 HBM presentation, comparing DDR3, DDR4, GDDR5 and HBM using ratios of mW/Gbps/pin; DDR4 uses 37% less power than DDR3, GGDR5 uses 13% less than DDR4, and HBM1 uses 42% less than GDDR5.

At similar bandwidth, HBM would consume far less power than DDR4, while using one HBM2 stack would give an APU 2-4GB of memory with more bandwidth than most low-end/low-mid GPUs.
 
APUs are such a big waste of money, a dGPU + x4 860k beats it to the ground (for same money as the 7870k), i have no idea why would someone try to buy an APU unless space reasons. APU concept was retarded as fuck i have no idea how a company like AMD thought they deserve their own market space.
 
Bear in mind that HBM isn't particularly good for the CPU side of a APU. HBM, while very high bandwidth, is also very high latency which is downright bad for feeding a CPU. You're not going to need 8GB of HBM for an integrated GPU, 1-2GB would be plenty, and you're still going to need low latency DDR4 on the MB to satisfy the CPU latency requirements.
 
a3venom said:
APUs are such a big waste of money, a dGPU + x4 860k beats it to the ground (for same money as the 7870k), i have no idea why would someone try to buy an APU unless space reasons. APU concept was retarded as fuck i have no idea how a company like AMD thought they deserve their own market space.
Click to expand...

90%+ of users will do fine with an APU. This is the reason why all of Intel's mainstream processors are APUs.
 
a3venom said:
APUs are such a big waste of money, a dGPU + x4 860k beats it to the ground (for same money as the 7870k), i have no idea why would someone try to buy an APU unless space reasons. APU concept was retarded as fuck i have no idea how a company like AMD thought they deserve their own market space.
Click to expand...

I dunno, I think an intel i7 6700K APU blows both of those totally out of the water.

Realistically, 90+% of consumer CPUs sold today are APUs. Does that make sense? Maybe not to us enthusiasts, but it saves the CPU companies money so they're going to keep doing it. Intel's insane markup (they don't cost any more to make) on their pure CPU line might also be partly to blame here too.
 
APU = AMD Processor w/ Graphics , it is not a technical term, it is a name of their product like Apple calls their phone "iPhone", so realistically all of the samsungs sold today are iPhones?
So if Intel made an APU they would be sued , that will be the best thing for AMD. Don't know what ya'all are saying.
What i am saying is, the APUs they advertise as being OK for light gaming can cost the same money as a X4 860k + r7 250 which will rape it in gaming performance (even if you put the best ram that money can buy with the APU). So let us pay more $ so we can have both on one chip?
 
a3venom said:
APU = AMD Processor w/ Graphics , it is not a technical term, it is a name of their product like Apple calls their phone "iPhone", so realistically all of the samsungs sold today are iPhones?
So if Intel made an APU they would be sued , that will be the best thing for AMD. Don't know what ya'all are saying.
What i am saying is, the APUs they advertise as being OK for light gaming can cost the same money as a X4 860k + r7 250 which will rape it in gaming performance (even if you put the best ram that money can buy with the APU). So let us pay more $ so we can have both on one chip?
Click to expand...

Or, you could pair that r7 250 with a 78xx series APU and get even better performance. ;)
For only $60 more, I think it's a fair price. Though, at this point it'd probably be better to wait or get a better dGPU, like a 280x or 290.

APU vs [whatever Intel calls them] is mostly semantics, though the technology used is very different, the result is the mostly the same.
 
Nobu said:
Or, you could pair that r7 250 with a 78xx series APU and get even better performance. ;)

APU vs [whatever Intel calls them] is mostly semantics, though the technology used is very different, the result is the mostly the same.
Click to expand...

No Sir, that would be the worst thing to do. EVER.
r7 250 + 7870k will cost you more than a x4 860k + r7 260 and the latter will give you performance numbers like 30 - 40% more.
So again pay way more for way less is the concept behind these APUs.
 
The 260 can run in dual-graphics mode too, btw. (well, supposedly. Hard to find any proof) Not the 260x, though.
Edit: Anyway, like I said, probably better to wait at this point.
 
Last edited:
Nobu said:
The 260 can run in dual-graphics mode too, btw. (well, supposedly. Hard to find any proof) Not the 260x, though.
Edit: Anyway, like I said, probably better to wait at this point.
Click to expand...

Nope, 260 wont work. (according to AMD's site)
APUs are a scam if anything, and waiting for new APUs is waiting for new scams. I think no one on this forum will justify paying more money for less performance when the comparison is so easy.
So APUs are an option for people when you can't fit a GPU in that case and every single person on the planet who bought a top end APU wasted their money unless they got it at a good deal.
 
I said wait, not specifically for APUs (though, that's what I'm interested in). They have some interesting products coming out in the dGPU and CPU space too, and if nothing else they should be lowering the price of their current line-up.

Would be interesting to see an A10-7xxx series with a r7 GPU running a game that uses Vulkan, imo.
 
I so hope that Zen is a win for AMD. My home used to be brimming with AMD product, but sadly has slowly dwindled to just a R9 290 and a HD 7850. I guess technically my SSD is "AMD" as well, but it made by OCZ.
 
Refresh my memory a bit here... When Bulldozer was released and universally panned as a big flop, about how far off in % was it compared to what Intel was offering at the time?

And when Zen is released, about how far off in % are we all anticipating it will be compared to what Intel has at the time?
 
jwcalla said:
Refresh my memory a bit here... When Bulldozer was released and universally panned as a big flop, about how far off in % was it compared to what Intel was offering at the time?

And when Zen is released, about how far off in % are we all anticipating it will be compared to what Intel has at the time?
Click to expand...

at stock clocks, the FX8150 was about 50% slower than the Sandy Bridge i7 2600K in Single Threaded performance, they were somewhat Tied in some heavy Multi-threaded applications for obvious reasons, like 7-Zip and x264 encoding, but the rest the 2600K was about 20-40% faster. to the point it was really competitive with the i5 2500K instead.

also the gaming performance intel offerings was a good 50% faster. mostly because in that time games ran in 2 cores mostly..
 
gamerk2 said:
You're not going to have any form of HBM on die, because of the insane power the chip would draw (nevermind the cost to produce said chip). You also won't see HBM as system RAM for the same reason.

HBM for APUs is a pipe dream.
Click to expand...

I was thinking more along the lines of on the motherboard like the L3 cache from the K6-3 days.
 
HBM wont happen due to power draw?
Would the Fury-X disprove that? Yeah it has high draw but its possible is my point.
 
JustReason said:
You would have to be a fool to believe HBM wont make it to an APU. It is the solution to alleviate the biggest negative APUs have.
Click to expand...

You can't cost justify it though. No one will purchase a $500 APU. HBM is simply too expensive and power hungry to use as on-die cache.

JMccovery said:
'Insane power draw'? Are you using Fiji as an example? Because HBM isn't why Fiji consumes as much power as it does.

According to Hynix's Hot Chips '14 HBM presentation, comparing DDR3, DDR4, GDDR5 and HBM using ratios of mW/Gbps/pin; DDR4 uses 37% less power than DDR3, GGDR5 uses 13% less than DDR4, and HBM1 uses 42% less than GDDR5.

At similar bandwidth, HBM would consume far less power than DDR4, while using one HBM2 stack would give an APU 2-4GB of memory with more bandwidth than most low-end/low-mid GPUs.
Click to expand...

And several times more then the TTL logic circuits that currently make up CPU cache, which happens to be the largest and most power hungry part of modern CPUs.
 
jwcalla said:
Refresh my memory a bit here... When Bulldozer was released and universally panned as a big flop, about how far off in % was it compared to what Intel was offering at the time?

And when Zen is released, about how far off in % are we all anticipating it will be compared to what Intel has at the time?
Click to expand...

What is your point here? Why don't you list the difference in R&D budget between Intel and AMD for that time period as well also the manufacturing process difference between both as well then maybe you consider not making such smart ass comments in the future...


Can I remind you as well that Intel pulled stupid stuff as compiler shenanigans dumping chips for really low prices outright buying of OEM as DELL from going anywhere else then Intel even if they did not have the stock to supply them.
 
Pieter3dnow said:
What is your point here? Why don't you list the difference in R&D budget between Intel and AMD for that time period as well also the manufacturing process difference between both as well then maybe you consider not making such smart ass comments in the future...


Can I remind you as well that Intel pulled stupid stuff as compiler shenanigans dumping chips for really low prices outright buying of OEM as DELL from going anywhere else then Intel even if they did not have the stock to supply them.
Click to expand...

Hey man, when AMD was releasing good chips they still had limited R&D, that's not an excuse to release a shit tier product.
This was probably a bad decision made by a handful of top level engineers/execs and whole company had to pay the price. They could've made something good(or even viable) in the same amount of money they wasted on the FX. Looks like Zen is the right direction but its a pretty long wait though.

Intel did miserable shit for business practices but they didn't hop on the multi core zero IPC bandwagon thankfully.
 
Araxie said:
at stock clocks, the FX8150 was about 50% slower than the Sandy Bridge i7 2600K in Single Threaded performance, they were somewhat Tied in some heavy Multi-threaded applications for obvious reasons, like 7-Zip and x264 encoding, but the rest the 2600K was about 20-40% faster. to the point it was really competitive with the i5 2500K instead.

also the gaming performance intel offerings was a good 50% faster. mostly because in that time games ran in 2 cores mostly..
Click to expand...

the only time gaming benchmarks are 50% better then an FX8150 is when the resolution was 1024x768. as the resolution and graphic settings increase the gains intel have mostly dissolved.

Also the I5 matches the I7 in single threaded performance so in the cases where an i7 stomps a FX8150 so does the I5 or even i3. In my general use (gaming, surfing the web, movie encoding, photo editing) AMD has never made me feel obsolete at least in "seat of the pants computing"

can you see a difference between 120fps and 90fps? i cant. Then again i dont use 144hz monitors either.
 
Quix said:
I dunno, I think an intel i7 6700K APU blows both of those totally out of the water.

Realistically, 90+% of consumer CPUs sold today are APUs. Does that make sense? Maybe not to us enthusiasts, but it saves the CPU companies money so they're going to keep doing it. Intel's insane markup (they don't cost any more to make) on their pure CPU line might also be partly to blame here too.
Click to expand...

If you were offered an I7-6700k $450 or a $63 AMD 860k with a $349 geforce 970GTX OC which would you take? No brainer there, id take the AMD.

A lot of you have a lot more disposable income then I do, but i see the i7-6700k as a giant waste of money that doesnt give me any "tangible" increase over an overclocked i7-920. We could argue about this, but my use case is "mild gaming and surfing the web". Games I play mostly play fine on and AMD 6000+

My sons a10-5800k with a 970gtx plays everything i have absolutely flawless and an 860k is 20% faster... meanwhile its $63 shipped from amazon.
 
{NG}Fidel said:
HBM wont happen due to power draw?
Would the Fury-X disprove that? Yeah it has high draw but its possible is my point.
Click to expand...

I expect we'll see it in the server parts, but in consumer APUs I think it's hard to say.
HBM worked on the fury because it lowered the power consumption of the memory and memory bus by moving on chip. A cpu/apu has an external memory system so unless HBM + HBM controller comes in at a lower TDP than a DDR4 controller then it's going to mean a rise in power draw on the chip.
As far as I know there are no public numbers for any of this.
 
Pieter3dnow said:
What is your point here? Why don't you list the difference in R&D budget between Intel and AMD for that time period as well also the manufacturing process difference between both as well then maybe you consider not making such smart ass comments in the future...
Click to expand...

I guess my point is that there seems to be some cognitive dissonance. People won't think twice about throwing the back of their hand to their forehead when thinking about the catastrophe that was Bulldozer, but at the same time hold out some kind of high unicorn hopes for Zen. But I don't really understand why... When you compare the relative performance back then (Bulldozer vs Sandy Bridge) to what is being predicted for Zen vs. whatever Intel will have, it doesn't seem like it's much different.

It's great that Zen will be much better than what AMD has right now. But if the thing is benchmarked on release and comes up way short of what Intel has, it'll be deemed just as much a failure as Bulldozer was.

It seems like it would have been a great chip maybe two years ago.

Maybe I'm wrong.
 
jwcalla said:
I guess my point is that there seems to be some cognitive dissonance. People won't think twice about throwing the back of their hand to their forehead when thinking about the catastrophe that was Bulldozer, but at the same time hold out some kind of high unicorn hopes for Zen. But I don't really understand why...
Click to expand...
Because AMD has made good products before and there's no reason to assume one mistake means they can't make another good chip.
 
jwcalla said:
I guess my point is that there seems to be some cognitive dissonance. People won't think twice about throwing the back of their hand to their forehead when thinking about the catastrophe that was Bulldozer, but at the same time hold out some kind of high unicorn hopes for Zen. But I don't really understand why... When you compare the relative performance back then (Bulldozer vs Sandy Bridge) to what is being predicted for Zen vs. whatever Intel will have, it doesn't seem like it's much different.

It's great that Zen will be much better than what AMD has right now. But if the thing is benchmarked on release and comes up way short of what Intel has, it'll be deemed just as much a failure as Bulldozer was.

It seems like it would have been a great chip maybe two years ago.

Maybe I'm wrong.
Click to expand...

Nobody will hold it as a failure or really expect it to hold up to intel's current offering at that time.
People are expecting Perf equal to Sandy / Ivy on this and if it matches haswell that will be a wet dream for most. If AMD succeeds in this nobody can call it a failure.
I think Bulldozer was considered a flop because of it's really bad IPC not overall multicore benched performance. 8 slow cores getting beaten by fast dual cores + HT today says a lot.
 
jwcalla said:
I guess my point is that there seems to be some cognitive dissonance. People won't think twice about throwing the back of their hand to their forehead when thinking about the catastrophe that was Bulldozer, but at the same time hold out some kind of high unicorn hopes for Zen. But I don't really understand why... When you compare the relative performance back then (Bulldozer vs Sandy Bridge) to what is being predicted for Zen vs. whatever Intel will have, it doesn't seem like it's much different.

It's great that Zen will be much better than what AMD has right now. But if the thing is benchmarked on release and comes up way short of what Intel has, it'll be deemed just as much a failure as Bulldozer was.

It seems like it would have been a great chip maybe two years ago.

Maybe I'm wrong.
Click to expand...
Maybe they aren't the same people? ;) At least, I wasn't on forums crying about how bulldozer should have been much better than it was, and how I was ripped off... :/
 
jwcalla said:
I guess my point is that there seems to be some cognitive dissonance. People won't think twice about throwing the back of their hand to their forehead when thinking about the catastrophe that was Bulldozer, but at the same time hold out some kind of high unicorn hopes for Zen. But I don't really understand why... When you compare the relative performance back then (Bulldozer vs Sandy Bridge) to what is being predicted for Zen vs. whatever Intel will have, it doesn't seem like it's much different.

It's great that Zen will be much better than what AMD has right now. But if the thing is benchmarked on release and comes up way short of what Intel has, it'll be deemed just as much a failure as Bulldozer was.

It seems like it would have been a great chip maybe two years ago.

Maybe I'm wrong.
Click to expand...

It depends on where you are thinking processor power lies if you stick to x86 then Intel is not really something noteworthy to begin with, they beat AMD by money, process technology and marketing it is not really a contest.

But if it was easy then we would have more x86 processors and we don't now you know why Intel is so good...
 
a3venom said:
People are expecting Perf equal to Sandy / Ivy on this and if it matches haswell that will be a wet dream for most. If AMD succeeds in this nobody can call it a failure.
Click to expand...

They can't? Why not? If Zen releases in late 2016 / early 2017 and is comparable to Sandy or Ivy Bridge that puts them on the same level as 5 year-old processors.

The number of people interested in buying a Sandy / Ivy Bridge processor in 2017 is going to be close to 0.

I'm just not sure how that's going to fly better than the Bulldozer launch.

Zen needs to be at least better than Kaby Lake at launch, and be out in retail by mid-2017, or it's DOA in terms of generating profit. That's my view of it. I mean... it's nothing personal or anything like that... this is just basic business here.
 
I like my 6700k build and the fact that it just works. (As long as I overclock correctly, anyways. :D ) However, I still think AMD processors carry more excitement and I would love to build an FX 9590 machine with crossfire R9 Furies. However, no need and besides, I already spent enough money in the last 3 months anyways.

I can still enjoy my FX8350 and MSI 970 Gaming board at work though. The machine is very fast even at just 4.2GHz and I do not game much on it anyways. An 8 core multitasking monster with a 500GB SSD, 240GB SSD for VHD's and 32GB of ram. :) Perhaps Zen will be good but, I wish they had done the FX 1090 Chipset with a steamroller 10 core processor anyways.

jwcalla said:
They can't? Why not? If Zen releases in late 2016 / early 2017 and is comparable to Sandy or Ivy Bridge that puts them on the same level as 5 year-old processors.

The number of people interested in buying a Sandy / Ivy Bridge processor in 2017 is going to be close to 0.

I'm just not sure how that's going to fly better than the Bulldozer launch.

Zen needs to be at least better than Kaby Lake at launch, and be out in retail by mid-2017, or it's DOA in terms of generating profit. That's my view of it. I mean... it's nothing personal or anything like that... this is just basic business here.
Click to expand...

Zen will be whatever it is but with the boards released in March, we may get some leaked benches sooner than we expect. (I hope so, anyways.)
 
buttons said:
If you were offered an I7-6700k $450 or a $63 AMD 860k with a $349 geforce 970GTX OC which would you take? No brainer there, id take the AMD.

A lot of you have a lot more disposable income then I do, but i see the i7-6700k as a giant waste of money that doesnt give me any "tangible" increase over an overclocked i7-920. We could argue about this, but my use case is "mild gaming and surfing the web". Games I play mostly play fine on and AMD 6000+

My sons a10-5800k with a 970gtx plays everything i have absolutely flawless and an 860k is 20% faster... meanwhile its $63 shipped from amazon.
Click to expand...

what about the G3258? its also 63$ on amazon, and will do everything a 860k will do and better.

jwcalla said:
I guess my point is that there seems to be some cognitive dissonance. People won't think twice about throwing the back of their hand to their forehead when thinking about the catastrophe that was Bulldozer, but at the same time hold out some kind of high unicorn hopes for Zen. But I don't really understand why... When you compare the relative performance back then (Bulldozer vs Sandy Bridge) to what is being predicted for Zen vs. whatever Intel will have, it doesn't seem like it's much different.

It's great that Zen will be much better than what AMD has right now. But if the thing is benchmarked on release and comes up way short of what Intel has, it'll be deemed just as much a failure as Bulldozer was.

It seems like it would have been a great chip maybe two years ago.

Maybe I'm wrong.
Click to expand...

When Bulldozer came out, against Sandy Bridge the FX-8150 was on par with a 2500k in multi threaded programs. Same thing with the FX-8350 and the I5-3570k. Then AMD killed the FX line while Intel released Haswell, and now Skylake. More than likely AMD could have remained more competitive against Haswell, and Skylake if they released Steamroller and Excavator FX cpus. Granted they would still use more power than their Intel counterparts and you wouldn't get all the new chipset features you currently see. Maintaining parity with your competitors mid-range products isn't really gonna get them anywhere.

a3venom said:
Nobody will hold it as a failure or really expect it to hold up to intel's current offering at that time.
People are expecting Perf equal to Sandy / Ivy on this and if it matches haswell that will be a wet dream for most. If AMD succeeds in this nobody can call it a failure.
I think Bulldozer was considered a flop because of it's really bad IPC not overall multicore benched performance. 8 slow cores getting beaten by fast dual cores + HT today says a lot.
Click to expand...

Last I checked the FX 8 cores are still faster than Skylake I3's with all cores are loaded.

jwcalla said:
They can't? Why not? If Zen releases in late 2016 / early 2017 and is comparable to Sandy or Ivy Bridge that puts them on the same level as 5 year-old processors.

The number of people interested in buying a Sandy / Ivy Bridge processor in 2017 is going to be close to 0.

I'm just not sure how that's going to fly better than the Bulldozer launch.

Zen needs to be at least better than Kaby Lake at launch, and be out in retail by mid-2017, or it's DOA in terms of generating profit. That's my view of it. I mean... it's nothing personal or anything like that... this is just basic business here.
Click to expand...

I agree. While it would be super nice if AMD had a cpu that was more competitive than their current offerings, in order to sway buyers away from Intel it needs to offer an advantage over the products that are available at the time, not old products. Meaning Zen will need to compete with Skylake and Kaby Lake in price and performance. I wouldn't mind if Zen was 5-10% slower in single thread, but offered 6 cores for the price of intel's 4. Or perhaps was just a bit cheaper overall.


As for Zen leaked Benchmarks, You can replace Zen with Bristol Ridge aka excavator based desktop APU's. We will likely see some benchmarks of those surface here soon.
 
There are only a few things Zen would have to do to be interesting enough.
1. price/performance
2. at least 8 cores
3. Get close(r) in single thread while doing well in multi thread
4. Zen should be something AMD can build upon the next 3 to 5 years.

I'm not to bothered if single thread would not be close enough, a new platform with new features from ancient AM3+ with decent enough performance.
Still a long way away from launch....
 
I think we need to think about this some.... Uber chip design guy left because design was finished. Jim left! If he left that quickly, and all theyve had is in-house samples, I think its zafe to say that AMD is where they want to be. They started from scratch this time, and Jim has left before ES have even left the building. I think its very safe to say that we will see a fantastic processor from AMD with Zen.
 
Pieter3dnow said:
There are only a few things Zen would have to do to be interesting enough.
1. price/performance
2. at least 8 cores
3. Get close(r) in single thread while doing well in multi thread
4. Zen should be something AMD can build upon the next 3 to 5 years.

I'm not to bothered if single thread would not be close enough, a new platform with new features from ancient AM3+ with decent enough performance.
Still a long way away from launch....
Click to expand...

If they priortize multi thread over single it will be the end of the company, but i think they haven't made that mistake this time.
They promised increased IPC and said nothing about cores, hopefully it will be decent 6 core parts with viable IPC, that will work well for most users and they won't get any hate.
 
a3venom said:
Intel did miserable shit for business practices but they didn't hop on the multi core zero IPC bandwagon thankfully.
Click to expand...

Yes they fucking did, hell they were driving that thing. The Pentium 4 was a giant pile of shit, the only reason we aren't running 8 ghz p4s right now is AMD.
 
I had a P4D 820 iirc, and it was a slow and hot pos. Although, that +1GHz on air I got was pretty damn impressive
 
qwksi1 said:
Fallout 4 would like to speak with you.
Click to expand...

That was funny in case of the pentium.
But the 4.7 Ghz 8 core loses to a dual core in this case of an i3 here. This will change when they make Zen with atleast acceptable lvl of IPC and then add more cores. Fallout scales pretty OK with CPU but people got to realize even games that scale well with cores will start to fall behind without IPC


CPU_01.png
 
You must log in or register to reply here.
Back
Top