AMD X399 ！！！！！

[french toast]

We are not getting 1800x speeds for 180w, not a chance in hell.
Unless....That B2 stepping is better than i think?

 

[dnavas]

We are not getting 1800x speeds for 180w, not a chance in hell.
Unless....That B2 stepping is better than i think?

In the list of things to fix, moving the frequency ~300Mhz to the right, while impressive and even awe-inspiring, would not be the first item on my list. I bet 3.6Ghz *was* the goal, but I just don't buy it. Current ES samples, which are rumored to already be on the next stepping, are only at 3.1. Either we're not getting 3.6, or we're not getting B2. I'll be really happy with 3.2-3. Memory latency, on the otherhand, and memory overclock stability -- those things seem substantially more important to me. YMMV, but I expect half of that rumor was baloney. At least it isn't as ridiculous as "overclocks to 5G on air."

 

[french toast]

In the list of things to fix, moving the frequency ~300Mhz to the right, while impressive and even awe-inspiring, would not be the first item on my list. I bet 3.6Ghz *was* the goal, but I just don't buy it. Current ES samples, which are rumored to already be on the next stepping, are only at 3.1. Either we're not getting 3.6, or we're not getting B2. I'll be really happy with 3.2-3. Memory latency, on the otherhand, and memory overclock stability -- those things seem substantially more important to me. YMMV, but I expect half of that rumor was baloney. At least it isn't as ridiculous as "overclocks to 5G on air."

Yea probably 3.2 ghz, 3.5 turbo, no idea about ST or xfr, personally i think that leak is fake.

 

[OrangeKhrush]

My guess is 2900-3000 range at 3.4 all core

 

[IEC]

1800X is 3.6GHz base, 4.1GHz XFR/precision boost, and 3.7GHz all-core turbo @ 95W TDP.

Slap together 2x 1800X = ~190W. It's not out of the realm of possibility, especially since you are going to have double the die area to dissipate the heat, meaning something like a Noctua D15 should be adequate.

X399 boards are going to need some beefy and efficient VRMs...

 

[imported_jjj]

1800X is 3.6GHz base, 4.1GHz XFR/precision boost, and 3.7GHz all-core turbo @ 95W TDP.

Slap together 2x 1800X = ~190W. It's not out of the realm of possibility, especially since you are going to have double the die area to dissipate the heat, meaning something like a Noctua D15 should be adequate.

X399 boards are going to need some beefy and efficient VRMs...

I kinda forgot lol but isn't the all cores XFR 3.8GHz?
XFR however is above the listed TDP as it's used "when there is better cooling". So the listed TDP is for base clocks 3/6/4GHz without XFR and XFR just expands the TDP. in practice.
Pushing it to 1800X clocks is a bit of a stretch but almost doable if they tune it a bit. I don't quite expect those clocks but they could clock it slightly lower and add a larger XFR.
They don't really need to push too hard from day one though so 3.2-3.4 base would be fine but ST turbo at 4GHz with 4 cores would be nice.

 

[dnavas]

I kinda forgot lol but isn't the all cores XFR 3.8GHz?

3.7

but ST turbo at 4GHz with 4 cores would be nice.

I agree. I don't understand why ST turbo isn't active with one core per CCX in the first place. Seems like there would be less coordination involved :shrug:

 

[IEC]

3.7

I agree. I don't understand why ST turbo isn't active with one core per CCX in the first place. Seems like there would be less coordination involved :shrug:

AMD's precision boost speed is for up to 2 cores.

 

[Atari2600]

AMD's precision boost speed is for up to 2 cores.

Which you'd assume would be 4 for a 16C variant...?

 

[IEC]

Which you'd assume would be 4 for a 16C variant...?

It's a possibility. I wouldn't rule it out, but I wouldn't count on it either.

 

[Teizo]

So everyone is excited about the HEDT platform again? Cool. Just a few months ago before Ryzen launched everyone was saying how overbuilt and overblown HEDT chips were and how no one wants them anymore and it is a dying market. Forgive me if it's hard to keep up.

That said, it will be interesting to see how these chips measure to Skylake-X. The IPC won't be as good, so how the clock speeds and pricing compared to Skylake-X will be interesting. I wonder just how aggressive Intel's shareholders will let them be with their pricing.

 

[imported_jjj]

So everyone is excited about the HEDT platform again? Cool. Just a few months ago before Ryzen launched everyone was saying how overbuilt and overblown HEDT chips were and how no one wants them anymore and it is a dying market. Forgive me if it's hard to keep up.

That said, it will be interesting to see how these chips measure to Skylake-X. The IPC won't be as good, so how the clock speeds and pricing compared to Skylake-X will be interesting. I wonder just how aggressive Intel's shareholders will let them be with their pricing.

How excited folks are or aren't factors in price.Nobody is excited about a 1k$ octa core but 16 cores at same price or less is another matter.

The IPC conversation is no that simple. It's a 16 cores chip and nobody buys it for ST.
If you think MT perf, AMD has much better SMT and on a per core basis it catches up quite a bit in pretty much anything that's not AVX2. Let's call it MT IPC and that's the relevant metric with many cores.
Ofc remains to be seen how a MCM really performs, what AMD did to make it work well enough.

 

[lopri]

Well, it is interesting as a technical preview of Naples. I seriously doubt the utility of 16 cores on a desktop. (With inevitably lower max frequency)

 

[itsmydamnation]

So everyone is excited about the HEDT platform again? Cool. Just a few months ago before Ryzen launched everyone was saying how overbuilt and overblown HEDT chips were and how no one wants them anymore and it is a dying market. Forgive me if it's hard to keep up.

That said, it will be interesting to see how these chips measure to Skylake-X. The IPC won't be as good, so how the clock speeds and pricing compared to Skylake-X will be interesting. I wonder just how aggressive Intel's shareholders will let them be with their pricing.

Actually im hoping for a 12 core for not much more then the 1800X(obviously not clocked as high) . I dont care about avx-512 and absolute highest IPC/clock isn't that important. Lots of solid threads and lots of dimm slots is what i am after. If that doesn't happen then i wold just get 2x1700's for the job ( home esxi setup).

 

[french toast]

12 core for 799$ would put the cat amongst the pigeons, especially if software and bios are significantly improved in that time.

 

[itsmydamnation]

12 core for 799$ would put the cat amongst the pigeons, especially if software and bios are significantly improved in that time.

Nup, to much, if a 1700X is 399, a 1700 is $329 then and a 1600 is $219 then to me a 12 core non should be ~$630 a 12 Core X ~$700, the 16 core ~$750 the 16 core X $850 and the 16 core flag ship X $999.

All of these prices would give better margin then selling the two dies separate.
This kind of market is where the Zeppelin SOC design starts to really ramp up vs the monolithic die AMD should ram home that advantage to the benefit of us .

 

[french toast]

Nup, to much, if a 1700X is 399, a 1700 is $329 then and a 1600 is $219 then to me a 12 core non should be ~$630 a 12 Core X ~$700, the 16 core ~$750 the 16 core X $850 and the 16 core flag ship X $999.

All of these prices would give better margin then selling the two dies separate.
This kind of market is where the Zeppelin SOC design starts to really ramp up vs the monolithic die AMD should ram home that advantage to the benefit of us .

Too much?? That would obliterate a 1100$ 6900k, remember technology pricing is not anywhere close to linear

 

[imported_jjj]

Too much?? That would obliterate a 1100$ 6900k, remember technology pricing is not anywhere close to linear

What matters is to get the most out of it. Pricing needs to be appealing enough to generate sales.
12 cores with a costlier mobo would be a tough sell vs 8 cores if it's not between 500$ and 700$. I would say 599$ or 619$ would be best for AMD.
Being able to buy both a 16 cores CPU and a mobo with 1k$ would be a big plus so lets say 8 cores at 799$, 999$ and one more SKU above.
If they price it too high it will sell poorly, mobo makers won't care about it.
They could even go lower with 12 cores to shift more folks to the platform and hope that they upgrade in the future to a newer CPU.

 

[itsmydamnation]

Too much?? That would obliterate a 1100$ 6900k, remember technology pricing is not anywhere close to linear

A 1600 is $219 a 12 core Zen is two 1600's together @ 630 that is 43% more dollars per Core then a 1600.
A 1600X is $249 a 12 core X @ 700 that is 41% more dollars per Core then a 1600x so maybe make that 730 instead and it becomes 47% more per core.

 

[Atari2600]

Nup, to much, if a 1700X is 399, a 1700 is $329 then and a 1600 is $219 then to me a 12 core non should be ~$630 a 12 Core X ~$700, the 16 core ~$750 the 16 core X $850 and the 16 core flag ship X $999.

Can't see it going that cheap (although I'd be utterly delighted if AMD proved me wrong!!).

X399 is going to be deep into the prosumer market - thus your looking at margins somewhere between Ryzen and Opteron.

I reckon you could double the 8C prices to get the 12C prices, and probably 3.5x the 8C prices to get the 16C prices.

1800X = $499
1900X = $1000 (12C)
2000X = $1750 (16C)

So, the 1900X would be similar price to the Intel 6900, but batter it in (non AVX512) productivity, while the 2000X is similar to the 6950, and would embarrass it in (non AVX512) workloads.

 

[CatMerc]

So everyone is excited about the HEDT platform again? Cool. Just a few months ago before Ryzen launched everyone was saying how overbuilt and overblown HEDT chips were and how no one wants them anymore and it is a dying market. Forgive me if it's hard to keep up.

That said, it will be interesting to see how these chips measure to Skylake-X. The IPC won't be as good, so how the clock speeds and pricing compared to Skylake-X will be interesting. I wonder just how aggressive Intel's shareholders will let them be with their pricing.

I'm fairly certain the argument was that the market is very small, not that it's dying.

It is overbuilt and it is overblown, for 99% for users. But for the remainder, a 1000$ 16 core would be an absolute boon.

 

[itsmydamnation]

Can't see it going that cheap (although I'd be utterly delighted if AMD proved me wrong!!).

X399 is going to be deep into the prosumer market - thus your looking at margins somewhere between Ryzen and Opteron.

I reckon you could double the 8C prices to get the 12C prices, and probably 3.5x the 8C prices to get the 16C prices.

1800X = $499
1900X = $1000 (12C)
2000X = $1750 (16C)

So, the 1900X would be similar price to the Intel 6900, but batter it in (non AVX512) productivity, while the 2000X is similar to the 6950, and would embarrass it in (non AVX512) workloads.

No one is going to buy a 1750 16C processor, just like no one buys the 6950. AMD took the 8 core price point to 329. Why would they follow intel in higher core count? They want to take market share, with the prices i have suggested they make more margin selling the higher core parts by a large amount then selling the two dies separate.

AMD have already said they are going ultra aggressive in data center pricing, they have already gone very aggressive on 6 core and 8 core pricing ( ignoring the "flag ship" 1800X). Why do they all of a sudden not do it for one market?

 

[Atari2600]

No one is going to buy a 1750 16C processor, just like no one buys the 6950.

Your thinking of you and not looking at the wider marketplace.

Prosumers would jump at the chance to buy a 16C processor running at >3.0 GHz for under $2K.

The price of a 16C Xeon running at between 2.6 GHz (base) - 3.6 GHz (max ST turbo) is $2900 (E5-2697v4).

That is what you are comparing to.

 

[Atari2600]

For reference, the fastest 12C Xeon (3.0GHz - 3.5 GHz) is at $2150.


While these prices are off the ark - I don't believe workstations typically are bought in enough bulk to get substantial CPU discounts.

 

[beginner99]

Your thinking of you and not looking at the wider marketplace.

Prosumers would jump at the chance to buy a 16C processor running at >3.0 GHz for under $2K.

The price of a 16C Xeon running at between 2.6 GHz (base) - 3.6 GHz (max ST turbo) is $2900 (E5-2697v4).

That is what you are comparing to.

True but >$1700 is still pretty high if you want to sell actual volume. Intel HEDT CPUs are identical to their Xeon parts. This hypothetical Ryzen chip is not a Naples reject it is a separate design, an MCM of 2x Ryzen so not a complete new design but sure there is some work involved to get it up and running, So you will want to sell enough volume to make that investment worth it. >$1000 is too much for enthusiasts. That is where people buy that make more money from having a faster CPU. My guess is the top 16-core version will slot it at $1300. A bit more than 2x 1800x but not too much to prefer 2x1800x rigs over this. The slowest 16-core will then be $999 and fastest 12-core obviously less probably $849.