AMD announce EPYC

[4K8KW10]

I think you guys are forgetting the costs of 7nm wafers. You think that if the chip is smaller it costs less for AMD to buy from TSMC. The costs of 7nm chip is high, especially when AMD goes me first with GPU and CPU this time, so they are not sparing any expenses.
So in the end if they do 16 core mainstream, their ASP will not improve. And again, you guys so easy to spoil, we had mainstream 8 cores for only 1.5 years, and now you are dreaming of cheap 16 cores. You can only go so far in mainstream with core count war.
Intel 10nm? Are we still on that joke? By the time they bring 10nm+ to the market AMD will have mature 7nm+ chips or even something further.
Software developers by your logic @drunkenmaster will cater dual or 4 cores, since there are plenty of them in craptops. Moving to 16cores will not change their mindset.
And its amazing, that all of you forget such chip like Threadripper, you know, 16 cores last year, 32 cores this year. I think AMD is safe in the core count war. I for one don't want any additional software and a reboot to go and play a game which refuses to start due to too many cores in the system, and we have quite a few of those.

All in all, guys, give AMD a frikkin break to earn some money to pump back into R&D, and to balance their sheets. It has been only couple years back since AMD was considered as dead in the water, and now you are pushing them to maintain chumpchange margins while pleasing 2 enthusiasts who cannot be bothered to go proper HEDT?
Remember, even though Intel is broken at the moment, nVidia is in full swing, AMD need every penny of R&D to get themselves in track and keep themselves afloat once Jim Keller and Raja gets Intel on track.

Oh, how not nice
First of all, Intel is not broken, actually they post financial records each quarter.
Look at it from another point - more market share for AMD with tempting products means less sales for Intel and ultimately less threat for AMD. Which is an indirect profit for them, again.
Also, the proper HEDT costs a lot of money and has up to 32 cores / 64 threads.
Why should the MSDT remain at 1/4 of the cores count?!

We have had 8-core Ryzens for 1 1/2 year but before then, Intel had the $1000 8-core i7s for 3-4 years already.
Before then, we had been sitting on sad quad cores for a decade.
Not fair to wait another decade to go up from the 8-core?!

 

[SiDeards73]

Oh, how not nice
First of all, Intel is not broken, actually they post financial records each quarter.
Look at it from another point - more market share for AMD with tempting products means less sales for Intel and ultimately less threat for AMD. Which is an indirect profit for them, again.
Also, the proper HEDT costs a lot of money and has up to 32 cores / 64 threads.
Why should the MSDT remain at 1/4 of the cores count?!

We have had 8-core Ryzens for 1 1/2 year but before then, Intel had the $1000 8-core i7s for 3-4 years already.
Before then, we had been sitting on sad quad cores for a decade.
Not fair to wait another decade to go up from the 8-core?!

Software needs to catch up with Hardware though, im still often sat watching of the cores on my 1700 doing nothing while playing some games...

Once the Software catches up with more core usage then yeah, go for more cores, if your primarily gaming right now, anything above 8c/16t is overkill, most would argue that 6c/12t is ample and 8c/16t is overkill actually.

 

[muziqaz]

Oh, how not nice
First of all, Intel is not broken, actually they post financial records each quarter.
Look at it from another point - more market share for AMD with tempting products means less sales for Intel and ultimately less threat for AMD. Which is an indirect profit for them, again.
Also, the proper HEDT costs a lot of money and has up to 32 cores / 64 threads.
Why should the MSDT remain at 1/4 of the cores count?!

We have had 8-core Ryzens for 1 1/2 year but before then, Intel had the $1000 8-core i7s for 3-4 years already.
Before then, we had been sitting on sad quad cores for a decade.
Not fair to wait another decade to go up from the 8-core?!

Don't let Intels record profits fool you

There is a reason HEDT 32 cores costs more, because it costs more to put the chips together, what do you think happens when you get 16 core "mainstream" chip? It will cost more as well, because you are doubling a core count. How hard is it to realise that this is not a simple 2 core addition, you are asking AMD to price their very expensive brand new chips very low in order to jump to 16 cores for 0.0000001% of enthusiast market. Where is the logic in that? Its not financially viable, and it will not bring more market share, because we, the stupid ones, who want to buy top of the range chips are minority and do not dictate AMD financial sheets.
Core count is not an answer, not with M$ coding magicians.
And as I said AMD has 12/16 core Threadrippers already in the market. They are serving their purpose perfectly as multipurpose workhorses.
You sound like you are being extremely limited with 2700X at this right moment

 

[Panos]

I think you guys are forgetting the costs of 7nm wafers. You think that if the chip is smaller it costs less for AMD to buy from TSMC. The costs of 7nm chip is high, especially when AMD goes me first with GPU and CPU this time, so they are not sparing any expenses.
So in the end if they do 16 core mainstream, their ASP will not improve. And again, you guys so easy to spoil, we had mainstream 8 cores for only 1.5 years, and now you are dreaming of cheap 16 cores. You can only go so far in mainstream with core count war.

The 7nm waffer costs 2.5x more than the 14nm waffer. However each waffer produces twice as many chips.
AMD tiny design reduces chances for whole chips going down the drain, like it happens with Intel behemoths.
In addition it will be less losses, as the chips now are more simplified (no IO etc) to produce and improving further yields.

A dual chip TR (2950X/1950X/1920X/2920X/1900X) had cost around $170 to produce (materials cost). A 7601 had double the material cost. Yet both sold with huge margins over BOM.
Ryzen 3 was already sold to profit, using the same material as the Ryzen 5 and Ryzen 7. (half the TR4 cost).

Even if AMD has double the production costs now, still going to have a very healthy profit margin on EPYC and TR4 even if jacking up the prices still going to be reasonable.
Price is something AMD is targeting and mentioned it yesterday numerous times.

On low end aka Ryzen 3/5/7 we have no idea actually what the CPU going to be.

 

[LePhuronn]

@4K8KW10 hang on, hang on.

How often do we see AMD's detractors attack and laugh at their perceived "just throw MOAR COREZ" approach, citing Intel's superior per-core performance as the thing that truly matters? Now that Intel have been pushed into a core count increase with their mainstream products, wouldn't it be far better for AMD to now compensate for their per-core performance deficit whilst the software and gaming vendors catch up with getting their kit properly multi-core? Establish a 6-8 core baseline as our new mainstream level, get software vendors working to that baseline, let AMD uptick their per-core performance and get Intel to sort out their chip design and process woes.

Or are you saying that you'd prefer AMD to keep throwing "inferior" cores at software which doesn't support it and continually feed the trolls?

 

[humbug]

@Muzbaz i agree but i think 16 core mainstream is down to AMDs design, its a consequence of AMDs design.
These new chips are made up of four 16 core dies so consequently you end up with a 16 core CPU when you only use one of them.
The Uncore, the chip in the middle is on the old 14nm, its cheap to make and for mainstream will be half the size, you don't need 128 PCIe lanes, all in all the new 16 core mainstream chips will occupy a little over the same die area with the biggest potion being 14nm and cheap to make, the Uncore is probably made at GloFo.

I don't think AMD will lose any margins 7nm + 14nm 16 core vs 12nm 8 core.

I also think 16 core mainstream will make a lot of Intel's high cost HEDT chips look stupid, and deny Intel those high margin sales, which IMO at this point is a good thing, Intel needs to suffer a little bit to pull their heads out of their backside, but more importantly the kudos AMD gets for beating Intel is invaluable.

 

[humbug]

@4K8KW10 hang on, hang on.

How often do we see AMD's detractors attack and laugh at their perceived "just throw MOAR COREZ" approach, citing Intel's superior per-core performance as the thing that truly matters? Now that Intel have been pushed into a core count increase with their mainstream products, wouldn't it be far better for AMD to now compensate for their per-core performance deficit whilst the software and gaming vendors catch up with getting their kit properly multi-core? Establish a 6-8 core baseline as our new mainstream level, get software vendors working to that baseline, let AMD uptick their per-core performance and get Intel to sort out their chip design and process woes.

Or are you saying that you'd prefer AMD to keep throwing "inferior" cores at software which doesn't support it and continually feed the trolls?

I think the point here is its looking like AMD will have better pre-core performance and "MOAR COREZ". the cake and they get to eat it.

 

[muziqaz]

The 7nm waffer costs 2.5x more than the 14nm waffer. However each waffer produces twice as many chips.
AMD tiny design reduces chances for whole chips going down the drain, like it happens with Intel behemoths.
In addition it will be less losses, as the chips now are more simplified (no IO etc) to produce and improving further yields.

A dual chip TR (2950X/1950X/1920X/2920X/1900X) had cost around $170 to produce (materials cost). A 7601 had double the material cost. Yet both sold with huge margins over BOM.
Ryzen 3 was already sold to profit, using the same material as the Ryzen 5 and Ryzen 7. (half the TR4 cost).

Even if AMD has double the production costs now, still going to have a very healthy profit margin on EPYC and TR4 even if jacking up the prices still going to be reasonable.
Price is something AMD is targeting and mentioned it yesterday numerous times.

On low end aka Ryzen 3/5/7 we have no idea actually what the CPU going to be.

And how many TR chips they sold? Don't include EPYC, as those come with validation baggage and that's completely different scale of economy. For what I care they can go as many cores as they like in enterprise, as there is software ecosystem wanting for more everyday. Now you are asking AMD to cut all those margins in order to satisfy few select enthusiasts who want to find out if their 16 core chip can run 8k youtube video on their 4k monitor with less CPU utilisation.

@Muzbaz i agree but i think 16 core mainstream is down to AMDs design, its a consequence of AMDs design.
These new chips are made up of four 16 core dies so consequently you end up with a 16 core CPU when you only use one of them.
The Uncore, the chip in the middle is on the old 14nm, its cheap to make and for mainstream will be half the size, you don't need 128 PCIe lanes, all in all the new 16 core mainstream chips will occupy a little over the same die area with the biggest potion being 14nm and cheap to make, the Uncore is probably made at GloFo.

I don't think AMD will lose any margins 7nm + 14nm 16 core vs 12nm 8 core.

I also think 16 core mainstream will make a lot of Intel's high cost HEDT chips look stupid, and deny Intel those high margin sales, which IMO at this point is a good thing, Intel needs to suffer a little bit to pull their heads out of their backside, but more importantly the kudos AMD gets for beating Intel is invaluable.

Threadripper already makes Intel high end look stupid 2700x makes 9900k look stupid. AMD managed to make the whole intel look stupid.
From the picture of Rome cpu it seems to be 8x8core dies.

Either way, in my opinion, going 16 core 32 thread to mainstream is completely not needed at this very moment, because ecosystem is not there, and soon you gonna hit the marketing wall with 32 threads for your gaming PC. We all know how frikkin lazy game devs are, its not even funny. The main problem I see in all of this, if AMD indeed release 16 core Ryzen, my silly arse will go out and buy it even if I don't really need all this threadiness

 

[LePhuronn]

I think the point here is its looking like AMD will have better pre-core performance and "MOAR COREZ". the cake and they get to eat it.

Oh I know. The numbers for EPYC cast a favourable light on Ryzen 3000 getting a serious performance boost. I just think it's more logical at this point in time to get the per-core performance up with the 8 core config we have already to shut the Intel shills up, and then start boosting core counts again with Ryzen 4000. As a result, I don't get 4K8KW10's belief that staying with 8 cores is a bad thing.

 

[Panos]

Now you are asking AMD to cut all those margins in order to satisfy few select enthusiasts who want to find out if their 16 core chip can run 8k youtube video on their 4k monitor with less CPU utilisation.

No I do not ask for AMD to cut margins. But showing that compared to Intel, even at +30%-40% higher cost per BOM to make a CPU it won't break our wallets.
True the new CPUs might be more expensive but also depends how the Ryzen 5/7 is manufactured.

If it is split IO using chiplets on mainstream also, then I wouldn't be surprised getting all the failed chiplets on Ryzen 7 and even Threadripper series, and going on mainstream with 2 failed chiplets.
Using the 2/3/4 core CCX/chiplet (x2) keeping the 5/6/7 for Threadripper.

After all the Zen 2 designed is to cut costs and salvage as much as possible, if you see the design.

In addition the new IO design (where every core has access to all the ram), allows a full blown TR4 CPU to compete with EPYC and I don't believe AMD wants to do that.

 

[SiDeards73]

How much more was Zen+ over Zen1? £20? or so for 1700x vs 2700x? i am happy to pay another £30 ontop of that for Zen2 tbh, im fairly certain the performance will warrant it and anyhow it will still be almost half the price of the 9900k haha

 

[muziqaz]

No I do not ask for AMD to cut margins. But showing that compared to Intel, even at +30%-40% higher cost per BOM to make a CPU it won't break our wallets.
True the new CPUs might be more expensive but also depends how the Ryzen 5/7 is manufactured.

If it is split IO using chiplets on mainstream also, then I wouldn't be surprised getting all the failed chiplets on Ryzen 7 and even Threadripper series, and going on mainstream with 2 failed chiplets.
Using the 2/3/4 core CCX/chiplet (x2) keeping the 5/6/7 for Threadripper.

After all the Zen 2 designed is to cut costs and salvage as much as possible, if you see the design.

In addition the new IO design (where every core has access to all the ram), allows a full blown TR4 CPU to compete with EPYC and I don't believe AMD wants to do that.

Yes, their design is wise to save as much money as possible and maximise the yields, but that does not mean AMD should just straight away to negate those savings to step on their own products' heels just to satisfy couple of enthusiasts. Afterall AMD has TR with 32 cores, 24 cores, 16 cores and 12 cores, why would they release another 16 core cpu for much cheaper?

 

[SiDeards73]

Yes, their design is wise to save as much money as possible and maximise the yields, but that does not mean AMD should just straight away to negate those savings to step on their own products' heels just to satisfy couple of enthusiasts. Afterall AMD has TR with 32 cores, 24 cores, 16 cores and 12 cores, why would they release another 16 core cpu for much cheaper?

I agree, i cant see AMD putting a 16c/32t chip into AM4, it would cause bad karma with all the people that specifically bought into TR4 platform.. thats an Intel type move. No i think they will stick to 8c/16 as top end AM4, and keep TR where it is as well... They have a really healthy product stack right now, if you can offer big improvements in clockspeed and IPC gain per generation that is better than simply lobbing more cores on, unless of course you can do both!

 

[LePhuronn]

Ryzen will go 16 cores max when Threadripper goes 24 cores minimum.

 

[4K8KW10]

Software needs to catch up with Hardware though

This is impossible and never happens neither in practice, nor in theory.

@4K8KW10 hang on, hang on.

How often do we see AMD's detractors attack and laugh at their perceived "just throw MOAR COREZ" approach, citing Intel's superior per-core performance as the thing that truly matters? Now that Intel have been pushed into a core count increase with their mainstream products, wouldn't it be far better for AMD to now compensate for their per-core performance deficit whilst the software and gaming vendors catch up with getting their kit properly multi-core? Establish a 6-8 core baseline as our new mainstream level, get software vendors working to that baseline, let AMD uptick their per-core performance and get Intel to sort out their chip design and process woes.

Or are you saying that you'd prefer AMD to keep throwing "inferior" cores at software which doesn't support it and continually feed the trolls?

People vote with their wallets - this is why the FX-8350 with its 8-core configuration is the bestseller now and not some ultra-fast dual-core Pentium.
Better to have a wider processor for many simultaneous tasks, and not narrow processor that works at the highest possible frequency.

 

[drunkenmaster]

Software developers by your logic @drunkenmaster will cater dual or 4 cores, since there are plenty of them in craptops. Moving to 16cores will not change their mindset.
And its amazing, that all of you forget such chip like Threadripper, you know, 16 cores last year, 32 cores this year. I think AMD is safe in the core count war. I for one don't want any additional software and a reboot to go and play a game which refuses to start due to too many cores in the system, and we have quite a few of those.

16 core mainstream means likely 8 core minimum, it likely means 8 core APUs and it likely means quad core becoming the bare minimum in laptops as well. THat moves the baseline from dual core, to quad core. That's how that works. They still sold single cores along with dual cores, when they moved to quad cores, single cores went the way of the dodo, when they all went to 8 cores.... wait, Intel didn't until just now but while still selling 8th gen with dual cores in large supply.

More cores at the top end means more cores at the bottom end, that's the entire point. Game devs can't target the highest core chips sold because it will cut off 99% of potential game sales, which is the reason they target the lowest chips being sold. When the lowest chips improve, so do the targets game devs design for, this is both the way it's always been and the only way it can work. This worked on single core also, when the chips had X performance at the low end then that's what they targetted, as single core performance rose significantly and both AMD and Intel raised the performance of the lowest end chips so went up demands from games.

 

[4K8KW10]

16 core mainstream means likely 8 core minimum, it likely means 8 core APUs and it likely means quad core becoming the bare minimum in laptops as well. THat moves the baseline from dual core, to quad core. That's how that works. They still sold single cores along with dual cores, when they moved to quad cores, single cores went the way of the dodo, when they all went to 8 cores.... wait, Intel didn't until just now but while still selling 8th gen with dual cores in large supply.

More cores at the top end means more cores at the bottom end, that's the entire point. Game devs can't target the highest core chips sold because it will cut off 99% of potential game sales, which is the reason they target the lowest chips being sold. When the lowest chips improve, so do the targets game devs design for, this is both the way it's always been and the only way it can work. This worked on single core also, when the chips had X performance at the low end then that's what they targetted, as single core performance rose significantly and both AMD and Intel raised the performance of the lowest end chips so went up demands from games.

Software and studios follow hardware by the lowest common denominator. They will utilise 8 cores effectively when most gamers have 8 cores, not when most gamers have 4 cores. While most gamers have 4 cores they'll design for 4 cores, as dual core has been around heavily up to and including the current generations from Intel dual core is what they target.

You release 8/12/16 core options in mainstream and get Intel to match that is when devs will start targetting 8 cores baseline. Software follows hardware, hardware can't simply increase once games catch up because... that's not how it works. Push the CPU power available, kill dual and quad cores and games move forward.

We need the baseline performance to keep increasing if we want game devs to start spending more cpu power on physics, on AI and on other things. Almost every game made to date basically needs to run on a dual core i3, meaning the base game physics, AI and engine all have to fit inside that, then you add some cpu heavy effects that can bleed onto other cores but it's not much. When a dev goes oh, the lowest anyone has is an 8 core you can now say hell, I can put 3 cores purely for AI, and 2 more purely for tracking physical objects so lets have more items that can be interactable, destructible, etc and still have cores left over for other crucial things.

Driving cores and/or clock speed is crucial for the progression of gaming.

The bare minimum in laptops and office computers on a global level is extremely critical because nowadays the vast majority of these sold are cheap and sluggish, very irresponsive systems which make the users experience a torture.
The only way to go Forward is to discontinue once and for all all quad-core processors and below, make the hexa-core the absolute minimum, and twelve-sixteen-core configurations the mainstream.
I see this as the only solution to improve our comfort when working or doing anything on a PC.

 

[Journey]

It's not that hard to figure out is it?

Retail pricing of the R5 1600/2600 is USD $140-160 it's a six core CPU, using an 8 core die. Moving to a 16-core CPU isn't that big of a push at the highest end of the product stack for the enthusiast space. £499 again, like the 1800X was on release but this time for a 16c/32t CPU with 20+4 PCI-E lanes, and dual channel RAM, with a performance of that to outdo the competition. Again don't forget they sold the TR4 1900X which was only an 8-core CPU on the HEDT platform, giving access to those features that AM4 doesn't have. They could happily start the HEDT platform at 16c then go to 24/32/40/48 etc, allowing scope to make it impossible to chose Intel be it for number of cores, speed, or price.

I'd be highly surprised not to see a 6/8/12/16 in the standard desktop space, with a view to abandoning 4 cores to all but the very entry level devices.

 

[kitfit1]

It's not that hard to figure out is it?

Retail pricing of the R5 1600/2600 is USD $140-160 it's a six core CPU, using an 8 core die. Moving to a 16-core CPU isn't that big of a push at the highest end of the product stack for the enthusiast space. £499 again, like the 1800X was on release but this time for a 16c/32t CPU with 20+4 PCI-E lanes, and dual channel RAM, with a performance of that to outdo the competition. Again don't forget they sold the TR4 1900X which was only an 8-core CPU on the HEDT platform, giving access to those features that AM4 doesn't have. They could happily start the HEDT platform at 16c then go to 24/32/40/48 etc, allowing scope to make it impossible to chose Intel be it for number of cores, speed, or price.

I'd be highly surprised not to see a 6/8/12/16 in the standard desktop space, with a view to abandoning 4 cores to all but the very entry level devices.

Yep. While i was at work today i was thinking a lot about "Horizon" last night. To be honest, if Ryzen 2 is as good as AMD claims, they would be stupid even to sell 6 core it makes much more sense in my mind to just start with 8 in the first place.

 

[Journey]

Yep. While i was at work today i was thinking a lot about "Horizon" last night. To be honest, if Ryzen 2 is as good as AMD claims, they would be stupid even to sell 6 core it makes much more sense in my mind to just start with 8 in the first place.

I think 6 cores is a great starting point, and will allow them to use partially faulty dies (potentially), and if you consider they could be bringing them in at $99 and hopefully with a nice APU attached it would greatly disrupt the OEM market, and the low end desktop, more so than the 2200G has so far. It would be competing against Intel's i7 8700 or the i5 9600 but at a huge cost advantage and overall be better in power, speed, and security, a quadruple whammy.