It is official. AMD announced and demonstrated Heavy Metal 32c/64t Threadripper 2. 7nm on the way.

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

TheGiant

Senior member
Jun 12, 2017
748
353
106
Well with the current RAM prices we will see 32c/64t with 32GB RAM? o really?

to achieve acceptable single core/low core count performance it will draw 250W+ imo

why not to push 16cores higher than this useless nonsense of 32c/64t desktop machine (my opinion, not talking about that skl-x furnace)

when I bought Athlon XP 1700+ throughbred it was like 35W CPU with high oc potential, but you could use it as power efficient and high perf cpu with like 70W
when I got c2d e6300 and oced it with default voltage from 1,86GHZ to 3.16GHz it took like 50-60W with top of the top performance
now we are looking at more cores with less performance on the desktop with more power- but not just more power- like 150-300W (doesnt matter if intel or AMD )

reminds me of something (car fuel and emission measuring metric)
 
Last edited:

Abwx

Lifer
Apr 2, 2011
10,854
3,298
136
why not to push 16cores higher than this useless nonsense of 32c/64t desktop machine (my opinion, not talking about that skl-x furnace)

32 cores at frequency F consume no more than 16 cores at frequency 1.3-1.35F while providing substancially higher throughput, on a perf/watt POV less cores is not relevant.
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
32 cores at frequency F consume no more than 16 cores at frequency 1.3-1.35F while providing substancially higher throughput, on a perf/watt POV less cores is not relevant.

well, if I buy a bus it provides much better transfer capacity pre l/diesel that my personal car too...but I wont buy a bus...
 
  • Like
Reactions: CHADBOGA

TheGiant

Senior member
Jun 12, 2017
748
353
106
AMD doesn't have access to manufacturing process which would allow that.
that is obvious from the tech details and reviews...but its just marketing war imo

whats is your opinion on power from such 32c/64t at 4GHz (with realistic voltage) ?
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
that is obvious from the tech details and reviews...but its just marketing war imo

whats is your opinion on power from such 32c/64t at 4GHz (with realistic voltage) ?

Beyond the realms of possible.
32C/64T EPYC OC'd to 3.8GHz / 1.28V draws excess 420W in CB15, so you do the math.

I wouldn't be surprised if AMD has similar power limit thing going on with TR2, as they do with other Pinnacle Ridge CPUs (the actual power limit is >= 35% higher than the TDP).
250W is very low for a 32C/64T CPU, so I wouldn't be surprised if that'd be the case.
 
  • Like
Reactions: Drazick

TheGiant

Senior member
Jun 12, 2017
748
353
106
well that is more than I thought (maybe wrong conclusion that adding moar coarz doesn't add power so quickly)

so that going to be 500+ watts since the voltage curve is quite steep at that frequency...that is unusable
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
unusable for what exactly?
you are really ok with saying the desktop use has 500W+ CPU? just the cpu?

which accidentally has so little desktop use above 5+-GHz 8700K as allrounder best atm CPU (followed by the oced ryzen 2xxx from AMD)

I have a 4c/4c 4,5GHz 0,025Vvoltage i5-6600K (which consumes around non avx 35W of power) for excellent single/low threaded performance

and 14C/28T Broadwell xeon withc 256GB ECC for CFD calcs- I cannot imagine heating the room with this kind of power for desktop usage
 

Abwx

Lifer
Apr 2, 2011
10,854
3,298
136
Beyond the realms of possible.
32C/64T EPYC OC'd to 3.8GHz / 1.28V draws excess 420W in CB15, so you do the math.

.

The maths say that Epyc use the previous process while TR2 benefit from the more efficient 12nm, particularly above 3.5GHz, it shouldnt be far from 250W even at 3.8GHz, moreover on CB15.

Ryzen-7-2700X-OC-Consommation.png


https://www.overclockingmadeinfrance.com/test-amd-ryzen-7-2700x/15/
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
you are really ok with saying the desktop use has 500W+ CPU? just the cpu?

which accidentally has so little desktop use above 5+-GHz 8700K as allrounder best atm CPU (followed by the oced ryzen 2xxx from AMD)

I have a 4c/4c 4,5GHz 0,025Vvoltage i5-6600K (which consumes around non avx 35W of power) for excellent single/low threaded performance

and 14C/28T Broadwell xeon withc 256GB ECC for CFD calcs- I cannot imagine heating the room with this kind of power for desktop usage

Turning 250W stock to "500W+" seems a bit excessive. As long as your workload scales up to 64 threads it could use north of 250W and still be the most efficient CPU.

If you want to talk about power efficiency and heat load to a room, you have to look at the actual energy in Joules (or kJ) consumed for a given task. TechReport did that for the original Threadripper:
taskenergy-scatter.png

Disclaimer: they didn't compile this for all of their benchmark suite, just 1 example.

For those of us who run a lot of rigs (e.g. Distributed computing, render farm, etc) there's a reason why we've been buying up Ryzen 1700s and Threadrippers. I don't preorder parts due to having been burned one too many times so I'll wait until launch reviews, but I suspect Threadripper 2 could be a good addition to a DC farm. I'll certainly be waiting to see how it performs for the tasks I care about.
 

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
I have a 4c/4c 4,5GHz 0,025Vvoltage i5-6600K (which consumes around non avx 35W of power) for excellent single/low threaded performance
I also have a 6600K and it definitely consumes more than 35W under non avx loads. Even at 4Ghz it goes past 50W in CB15, so your chip is using almost 2x your power estimate when properly loaded.

I cannot imagine heating the room with this kind of power for desktop usage
Sure you can imagine, if you get back to realistic goals that is - like using the CPU at stock TDP and with high throughput loads in mind, you know, like that Xeon is doing.
 
Last edited:
  • Like
Reactions: krumme

ShowsOn

Junior Member
Aug 24, 2007
2
0
61
What sort of clocks could AMD get if they make a 16 core Threadripper 2 part?

Surely they could do better than 3.4 base / 4.0 boost using Zen+?

I suspect the 24 and 32 core parts will be restricted to ~3.5 boost to ensure backwards compatibility with the older boards. But in a good board they will boost to 4.0+.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
At 4.0Ghz on the 32 core part, you're looking at north of 320watts (using the information from the above chart). I suspect that the 32 core parts are going to probably cap at around 3.8-3.9Ghz with relaxed bios protections (but not aggressive manual overclocking). Now, the 24 core part looks the most interesting. It should consume a measurable amount less power than the 32 core part at a given clock speed. That being the case, I would not be surprised to see it boost into the 4.1Ghz range on quality boards with modest tweaking of the bios. I wouldn't be surprised to see it pushing stock boost numbers in the 3.9-4.0Ghz range.
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
I also have a 6600K and it definitely consumes more than 35W under non avx loads. Even at 4Ghz it goes past 50W in CB15, so your chip is using almost 2x your power estimate when properly loaded.


Sure you can imagine, if you get back to realistic goals that is - like using the CPU at stock TDP and with high throughput loads in mind, you know, like that Xeon is doing.

I am not ok that we are telling that desktop use needs 16cores and 250W. with 32 cores even more. This is not the future of PC computing I hope.

My experience is, if you give people space, they will use it. So if we create a bubble that 32 cores are needed, welcome unlimited amount of bloatware just to keep your system used....

I am really looking for a good improved 8 core design that can be underclocked/undervolted to 30W if needed and on the other hand oced to 5GHz to have good single thread/low latency gaming performance. Current AMD and Intel responses are just pure redacted

Profanity is NOT allowed
in the tech areas.

AT Mod Usandthem
 
Last edited by a moderator:

TheGiant

Senior member
Jun 12, 2017
748
353
106
Turning 250W stock to "500W+" seems a bit excessive. As long as your workload scales up to 64 threads it could use north of 250W and still be the most efficient CPU.

.

Imo its not excessive. Its very real, just look how much ryzen 8C consumes at 3,7,4,4,2 GHz .

I still value 6-8C at current per core performance of i7-8700K at 5GHz more than any intel or AMD offerings with moar coarz for desktop use
 

mattiasnyc

Senior member
Mar 30, 2017
356
337
136
I am not ok that we are telling that desktop use needs 16cores and 250W. with 32 cores even more. This is not the future of PC computing I hope.

My experience is, if you give people space, they will use it. So if we create a bubble that 32 cores are needed, welcome unlimited amount of bloatware just to keep your system used....

You're ignoring people using those CPUs in workstations, where actual work is being done on those machines and where the workload actually scales (non-linearly, but still) across all cores. More speed doesn't always help as much as more cores (scaling again).

I am really looking for a good improved 8 core design that can be underclocked/undervolted to 30W if needed and on the other hand oced to 5GHz to have good single thread/low latency gaming performance. Current AMD and Intel responses are just pure

This I agree with.
 
Last edited by a moderator:

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Why can't the 8700K be underclocked to 30W? I know it's a 6 core, but it stands to reason the 8 core version is going to be similar.
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
AM4 Pinnacle Ridge CPUs have >= 35% higher power limit than the advertised TDP.
We'll see if that will be the case with 2nd gen. TR as well.

1950X already drops it frequency below it's base in some cases, usually when the memory has been pushed above the maximum official speeds.
And that's at 180W power limit and 16 cores.

At 3.7GHz / 1.150V I've measured 220W+ consumption for the CPU itself.

12nm LP does provide higher frequency at ISO power than the 14nm LPP used on gen. 1 TR, however the difference is marginal (< 4% on average).

If the 250W power limit is accurate, the frequencies will be EPYC'ish.

Curious, in what situations have you ever seen the 1950X drop below 3.4 GHz?

In that case stock base/boost will depend entirely on what their definition of "250W" and what it encompasses entails.

If they are strict about power limits I could see closer to Epyc clocks. Which would be disappointing, but not unreasonable.

We know the dies themselves should be capable of 2700X+ clocks and are the best of the best, so I'm still optimistic a 4GHz all-core clock could be achieved at ~1.0V (since the better 2700X can get away with around 1.1V for those clocks).

Nope. I run my current 1950x with a slight OC and wattage creeps at or above 250Watts when max'd and the rushed Noctua works just fine :

<snip>

With 250watts worth of heat spread out even more (4 dies vs 2) this should be just fine for stock. VRMs are fine too. I noticed the back of my mobo around the CPU bracket and VRMs around the CPU area where the v-neck cut out usually is in a computer cases gets kind of toasty so I put an exhaust fan in that area. Everything runs just fine and that's with some reference GPUs doing external exhaust and some after-market GPUs dumping toasty heat into the case. All you need is proper air flow and you're just fine. No way in heck i'm ever putting water anywhere near a workstation with soo many pricey components and data.

VRMs are going to be key here. However, I thought the clocks on the 32 core version were finalized? 3.0 base, 3.4 boost + XFR.

Also, no 16 core version announcement? Maybe they figure the 1950X is good enough. Enabling Sense Mi got my Threadripper to boost to 4.2 GHz for long periods of time unless under extreme load. I also have power saving set to 1%, which drops the chip to 2000MHz or so, so the system sips power except when it's needed. The boost algorithm works great with this setup and I can't see things improving much with 12nm. It would be nice if they could get things up to 4.4 GHz though for 16 cores. I would replace my 1950X if they did that. I have no desire for a 32 core chip. The 16 core chip has enough CPU cores to do what I need. I just wish it were a bit faster.
 
  • Like
Reactions: lightmanek

ultimatebob

Lifer
Jul 1, 2001
25,135
2,445
126
you are really ok with saying the desktop use has 500W+ CPU? just the cpu?

which accidentally has so little desktop use above 5+-GHz 8700K as allrounder best atm CPU (followed by the oced ryzen 2xxx from AMD)

I have a 4c/4c 4,5GHz 0,025Vvoltage i5-6600K (which consumes around non avx 35W of power) for excellent single/low threaded performance

and 14C/28T Broadwell xeon withc 256GB ECC for CFD calcs- I cannot imagine heating the room with this kind of power for desktop usage

These processors are NOT meant for normal desktop use. These 32 core monsters are for CAD workstations, 4K video editing rigs, and wealthy jerks who want bragging rights.
 

ultimatebob

Lifer
Jul 1, 2001
25,135
2,445
126
And those that do DC to research cancer and try to cure it. Like me, and I have cancer

Heh, OK. That too. I'd imagine that it would do a fantastic job mining certain cryptocurrencies like Monero or Nimiq as well, since the encryption algorithm they use doesn't work well with a GPU.
 

ultimatebob

Lifer
Jul 1, 2001
25,135
2,445
126
Also, no 16 core version announcement? Maybe they figure the 1950X is good enough. Enabling Sense Mi got my Threadripper to boost to 4.2 GHz for long periods of time unless under extreme load. I also have power saving set to 1%, which drops the chip to 2000MHz or so, so the system sips power except when it's needed. The boost algorithm works great with this setup and I can't see things improving much with 12nm. It would be nice if they could get things up to 4.4 GHz though for 16 cores. I would replace my 1950X if they did that. I have no desire for a 32 core chip. The 16 core chip has enough CPU cores to do what I need. I just wish it were a bit faster.

I would expect AMD to release the 32 core, 16 core, and 8 core model numbers around the same time they announce pricing. Who knows, they might have a 24 core model as well, where only 3 of the 4 die slots in the processor are filled.
 

wahdangun

Golden Member
Feb 3, 2011
1,007
148
106
Heh, OK. That too. I'd imagine that it would do a fantastic job mining certain cryptocurrencies like Monero or Nimiq as well, since the encryption algorithm they use doesn't work well with a GPU.


Yes, they do fantastic job mining, hell even AMD advertising it, and it's even faster than Rx Vega.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Curious, in what situations have you ever seen the 1950X drop below 3.4 GHz?

With everything else at stock, while the MEMCLK is increased e.g. from 2666MHz to 2933MHz (SoC voltage on TR is dynamic).
This causes the CPU to drop slightly below the base frequency even in CB15 nT as the power limit is exhausted faster.

It is not uncommon to see TR performing worse at higher MEMCLKs when they're power bound.