AMD EPYC 7000 Series Details Leaked, Including Product Specifications and Clocks

[R-T-B]

A set of details on AMD's upcoming Naples platform's family of EPYC CPUs has leaked on the site videocardz.com, claiming that the top product, the EPYC 7601, will feature a turbo clock of no less than 3.2 GHz (base clocks are more moderate at around 2.2 GHz for the EPYC 7601), with a core count of 32 Zen units, and 64 threads. These clocks are pretty high for a 32-core CPU, and will probably be only for a handful of cores at a time. The EPYC 7601 is also a hot chip according to this leak, in both thermals and price. It costs over $4000 USD, and has a TDP of 180 Watts (not bad at all for a 32-core part, though!)

The rest of the lineup is detailed as well, but as this is a leak that is blatantly admitting to violating an NDA, it goes without saying this could be nothing more than a fabrication. Take it with your usual dose of healthy skepticism.

For the full details of the leak (including the rest of the expected lineup), you can view the source link below.



View at TechPowerUp Main Site

 

[DeathtoGnomes]

180 watts seems kinda low if thats the target to allow headroom to OC, why not start at 220w like some older chips?

Bet if they down-clocked to 1GHz they could get it down to 65w.

 

[R-T-B]

180 watts seems kinda low if thats the target to allow headroom to OC, why not start at 220w like some older chips?

Bet if they down-clocked to 1GHz they could get it down to 65w.

These are server parts. I really doubt OCing is on the menu.

 

[AsRock]

Due to what happened with the 480 a while back it has to be seen to believed.


But dam those chips looks AWESOME! .

 

[DeathtoGnomes]

These are server parts. I really doubt OCing is on the menu.

I can still dream!

 

[sergionography]

Bet if they down-clocked to 1GHz they could get it down to 65w.

Thats simply not an economical option. Why sell 32cores at 1ghz when that requires a 4chip module if you can sell 16cores at 2ghz requiring a 2 chip module while performing the same if not better. 1ghz is waay off from the performance/power/cost sweet spot

 

[Steevo]

Thats simply not an economical option. Why sell 32cores at 1ghz when that requires a 4chip module if you can sell 16cores at 2ghz requiring a 2 chip module while performing the same if not better. 1ghz is waay off from the performance/power/cost sweet spot

Thin client virtualization servers for users, a user that runs only data entry or the like will perform fine with windows 7 on two 1Ghz cores and a couple gigs of RAM. Lower power means smaller servers with the same or more users.

 

[Prima.Vera]

"EPYC"

 

[eidairaman1]

What socket? I'm pretty lost on what sockets these and "thread ripper" are going to use...

 

[seronx]

What socket? I'm pretty lost on what sockets these and "thread ripper" are going to use...

Same pin count different socket interface. SP3 is LGA and SP4 is BGA.

 

[Caring1]

Same pin count different socket interface. SP3 is LGA and SP4 is BGA.

LGA and PGA?

 

[R-T-B]

LGA and PGA?

Being it's a server, I really doubt he means PGA. He most likely means what he said: BGA. I personally am unsure if it's BGA or LGA but I know it probably won't be PGA.

 

[destroyah3034]

Both Threadripper and Epyc will be on a LGA type of socket. Both should be on socket about the same size, but the pin layout will be slightly different between the two because of the additional amount of pci-e lanes and additional i/o included on Epyc.

 

[HopelesslyFaithful]

my 1650v3 can do 160/220 watts depending on the type of load and thats 6 cores vs 32. I would love to see this full unlocked and pushing 400 watts assuming a MB can deliver that kind of juice.

 

[DeathtoGnomes]

Thats simply not an economical option. Why sell 32cores at 1ghz when that requires a 4chip module if you can sell 16cores at 2ghz requiring a 2 chip module while performing the same if not better. 1ghz is waay off from the performance/power/cost sweet spot

 

[HopelesslyFaithful]

when I did testing on intel IB the sweet spot was about 2GHz on a 4 core in terms of performance per watt. This depends on process and arc but generally for 4 cores like zen is. 2GHz is probably fairly close to the most efficient spot.

some one would have to test that but i would be surpised if 1GHz was truely the most efficient clock. I would wager 1.5-2.5 GHz is the ideal range but i havent tested one of these so i am going off history of CPUs on whats most efficient. There is a large fixed power draw in CPUs where lower mhz does not mean the most efficient.

 

[infrared]

Power efficiency is the name of the game with these, hence the lower than enthusiast grade clocks. Anyway, as R-T-B said, take this info with a good dose of skepticism.

Edit: does anyone know if EPYC is an acronym?

 

[sinnedone]

I take it that sarcasm infused post didn't quite work out.

 

[DeathtoGnomes]

I take it that sarcasm infused post didn't quite work out.

when I did testing on intel IB the sweet spot was about 2GHz on a 4 core in terms of performance per watt. This depends on process and arc but generally for 4 cores like zen is. 2GHz is probably fairly close to the most efficient spot.

some one would have to test that but i would be surpised if 1GHz was truely the most efficient clock. I would wager 1.5-2.5 GHz is the ideal range but i havent tested one of these so i am going off history of CPUs on whats most efficient. There is a large fixed power draw in CPUs where lower mhz does not mean the most efficient.

my comment refers to the fact that high core count cpu's usually have to reduce clocks to keep power requirements down and ... heat. sarcasm lost.

 

[Jism]

These are server parts. I really doubt OCing is on the menu.

I see some threads going on in AMD section of xtremesystems.org, where people are OC'ing server chips. There are motherboards capable of doing OC's. I'm sure that motherboard vendors of servers as well allow some kind of oc'ing.

 

[HTC]

I see some threads going on in AMD section of xtremesystems.org, where people are OC'ing server chips. There are motherboards capable of doing OC's. I'm sure that motherboard vendors of servers as well allow some kind of oc'ing.

I seriously doubt people using these chips for servers will even consider OCing as they ought to value stability far more then CPU speed.

That does not mean that there wont be people who will buy these CPUs and attempt to OC them: i'm sure there will be those that will do so, but these people will most likely not use these chips as servers.

 

[TheoneandonlyMrK]

I seriously doubt people using these chips for servers will even consider OCing as they ought to value stability far more then CPU speed.

That does not mean that there wont be people who will buy these CPUs and attempt to OC them: i'm sure there will be those that will do so, but these people will most likely not use these chips as servers.

I think it's interesting to watch since it's likely to lead to consumer products.

 

[HopelesslyFaithful]

I seriously doubt people using these chips for servers will even consider OCing as they ought to value stability far more then CPU speed.

That does not mean that there wont be people who will buy these CPUs and attempt to OC them: i'm sure there will be those that will do so, but these people will most likely not use these chips as servers.

I would overclock these to 3.8GHz for my NAS/server. I see little reason not to as long as its stable for my usage. Zip/encoding/plex/and so on.

 

[ADHDGAMING]

I seriously doubt people using these chips for servers will even consider OCing as they ought to value stability far more then CPU speed.

That does not mean that there wont be people who will buy these CPUs and attempt to OC them: i'm sure there will be those that will do so, but these people will most likely not use these chips as servers.

Yeeeeah Professional Server guys are not going to overclock this if this something they are using for enterprise. You will have a small demographic that will OC these. As a whole though, these will not be overclocked

 

[DeathtoGnomes]

I want one anyway.