AMD EPYC 7V13 64 core 128 thread $699

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Bert

Well-Known Member
Mar 31, 2018
868
414
63
45
Last edited:

gb00s

Well-Known Member
Jul 25, 2018
1,206
605
113
Poland
I don't see a lot of great deals on AMD SP3 systems, especially the second gen ones that support Milan and Rome but not Naples. Still, 64 core CPU for $699 is a smokin' hot deal

Link AMD EPYC Milan 7V13 64-Core 2.45GHz ES Processor 100-000000425 *NO VENDOR LOCKED | eBay

Listing says ES but they have the model number printed so they're like Intel QS CPUs, right?
Had one at hand and could not get it working on Gigabyte MZ-AR0 v1 and v3 with newest bios enabling MILAN on v1. Maybe CPU related. Couldn't tell as I had no further MILAN board at hand recently. Not from this specific seller, but there was a similar offer on eBay recently.
 
  • Like
Reactions: Patrick

nutsnax

Active Member
Nov 6, 2014
262
102
43
113

chrgrose

Active Member
Jul 18, 2018
116
59
28
I was thinking these were better deal since you can overclock them:


Looks like they are sold out.
I assume only the QS can overclock? What boards does the overclock work on? Any dual socket boards?
 

DaveLTX

Active Member
Dec 5, 2021
170
40
28
Careful with these, the chips say 7T83 and 7V13 most often wouldn't work on ALL boards except with the earliest firmware.

Im not sure about 7V13 but 7T83 certainly goes cheap (as well as some other rome ones) because they don't work on most boards
 

Advin

Member
Jun 10, 2023
50
18
8
Careful with these, the chips say 7T83 and 7V13 most often wouldn't work on ALL boards except with the earliest firmware.

Im not sure about 7V13 but 7T83 certainly goes cheap (as well as some other rome ones) because they don't work on most boards
I think this specific seller is selling a late revision. ES is kind of a gamble in general since earlier revisions generally only work on early firmware, but this listing seems to be okay.

I bought one and it boots fine on MZ32-AR0 with a late BIOS and have heard of others on the forum booting it fine on AsRockRack boards. I think this one and seller is OK!

Will send pictures tomorrow.
 

Advin

Member
Jun 10, 2023
50
18
8
@Patrick I think an interesting idea for an article on STH could maybe be about engineering sample, qualification samples, and OEM CPUs for AMD. There’s a ton of documentation and comparisons on the older Intel Xeon ES stuff, but very rarely AMD.

There’s a ton of Genoa ES/QS out there as well, super cheap way to get into a modern DDR5 enterprise platform.

I also see a ton of OEM models available in extremely large amounts on eBay like 7B13, 7J13, 7V13, 7R13, etc (hovering around $1-1.2k for 64C/128T Milan) and now even Genoa 9B14 OEM.
 
  • Like
Reactions: Patrick and izx

izx

Active Member
Jan 17, 2016
133
64
28
40
I can confirm this batch of 7V13s work fine on an AsrockRack ROMED8-2T running the latest 3.80 BIOS.

TDP is 240W, and base (2.65-2.7) and boost (3.7) clocks match 7713.
 

Bert

Well-Known Member
Mar 31, 2018
868
414
63
45
@Patrick I think an interesting idea for an article on STH could maybe be about engineering sample, qualification samples, and OEM CPUs for AMD. There’s a ton of documentation and comparisons on the older Intel Xeon ES stuff, but very rarely AMD.

There’s a ton of Genoa ES/QS out there as well, super cheap way to get into a modern DDR5 enterprise platform.

I also see a ton of OEM models available in extremely large amounts on eBay like 7B13, 7J13, 7V13, 7R13, etc (hovering around $1-1.2k for 64C/128T Milan) and now even Genoa 9B14 OEM.
That is controversial. ES/QS chips are not for resale and them getting into public hands is against the end user agreements. They are borderline stolen items. STH should not publicly promote their usage.

I am not saying we should not use ES chips in our private conversation. In any case, I am pretty sure Patrick understand these things better than I do.
 

Advin

Member
Jun 10, 2023
50
18
8
That is controversial. ES/QS chips are not for resale and them getting into public hands is against the end user agreements. They are borderline stolen items. STH should not publicly promote their usage.

I am not saying we should not use ES chips in our private conversation. In any case, I am pretty sure Patrick understand these things better than I do.
I mean, I've seen massive channels like LTT promoting or frequently using ES/QS Xeon processors a while back. It doesn't necessarily have to show it in a good light, I would mostly just be curious about where they come from, how they are available in these quantities (typically from China), and how they are different.

But yeah, I think it could be a controversial topic so not entirely sure. Just an idea or topic I thought could be cool to look into at one point.
 

nutsnax

Active Member
Nov 6, 2014
262
102
43
113
That is controversial. ES/QS chips are not for resale and them getting into public hands is against the end user agreements. They are borderline stolen items. STH should not publicly promote their usage.

I am not saying we should not use ES chips in our private conversation. In any case, I am pretty sure Patrick understand these things better than I do.
All of this is my opinion of course.... begin rant:

You'll notice that many of these QS/ES chips have scuffs, scratches, missing caps etc. That's because they're mostly slammed into a dumpster at some point. Are there proper disposal clauses written into those end user agreements with the companies that they partner with? How would a partner even properly dispose of them? Smash every one of them with a hammer? How does AMD/Intel even enforce that? Is hurling them into a dumpster not good enough?

From the partner's standpoint they disposed of the equipment (by launching it into the dumpster) and they are no longer bound by that agreement.

Possession is nine-tenths of the law in this case. If someone fishes them out of a dumpster or a landfill after a partner, threw them away then I'm sorry, but not only are those agreements out the window but any rights AMD or Intel had to "their" chips also go out the window. It is not "their" chip anymore, it belongs to the landfill dumpster diver I don't care how high-tech it is.

....and punishing partners because they didn't properly dispose of these chips is great look for the company. Do I, as a partner, get a discount for properly incinerating or crushing your chips while protecting your multi-billion-dollar margins? How about a coupon off my next purchase at least?

IMO Intel/AMD's "rights" here are in the dumpster the moment that box of chips hits the dumpster and good luck trying to sue to enforce this (not gonna happen and then there's the whole ".....but you threw it in the trash" thing).

end rant.
 

Advin

Member
Jun 10, 2023
50
18
8
chrome_YvsOyDfizv.jpgchrome_4RhxSwzXY3.png

CPU temps reported in OCCT appear to be wrong; it looks normal in IPMI and other programs.

Using the latest BIOS M18_R34 flashed on REV 1.0 Gigabyte MZ32-AR0 board.

2.45 GHz base clock, meaning that this is most likely an EPYC 7763 OEM version, but OCCT is reporting 240W TDP by default (it could probably be increased to 280W in BIOS, but I'll leave it at 240W).

All 16 DIMMs are showing up, so memory channels are intact.
 

VMman

Active Member
Jun 26, 2013
128
47
28
I can confirm this batch of 7V13s work fine on an AsrockRack ROMED8-2T running the latest 3.80 BIOS.

TDP is 240W, and base (2.65-2.7) and boost (3.7) clocks match 7713.
Any idea if these play nice with Supermicro boards?
 

izx

Active Member
Jan 17, 2016
133
64
28
40
CPU temps reported in OCCT appear to be wrong; it looks normal in IPMI and other programs.
Thanks for reporting this. I see weird behavior with on-CPU temp sensors too.

Only the Tctl/Tdie temp reporting via OCCT (ie HWInfo engine) appears to be accurate based on when thermal throttling occurs (~90-93 C) and hand feel of heatsink/exhaust air.

The CCD/CCX and other on-die Temps appear to remain oddly low. Needs some investigation.

Since you can't run HWInfo and OCCT together, I tried HWMonitor and that reports the idle package temp at ~85C. Shooting to near 110 when running OCCT. Looks like an offset isn't getting interpreted correctly by HWMonitor.

Whatever sensor IPMI is measuring for me (on the ROMED8 board) also remains low, meaning fan control is messed up.

Going to try HWInfo sensors with another stress program like Prime95 or Linpack to get a better idea.
 

wildpig1234

Well-Known Member
Aug 22, 2016
2,257
503
113
49
How are these compared to 7702? I got 7702 several months ago for the same price of $700
 

Advin

Member
Jun 10, 2023
50
18
8
How are these compared to 7702? I got 7702 several months ago for the same price of $700
I think this CPU is in between the EPYC 7713 and 7763 in performance because of 240W TDP.
I would look at reviews for EPYC 7713, usually they contain comparison to EPYC 7702. These ES will perform probably 10-15% better than 7702 but it will depend on your workload and obviously you won't get any official support from AMD/etc.
 

GhettoSuperstar

New Member
Mar 12, 2020
27
14
3
I think this CPU is in between the EPYC 7713 and 7763 in performance because of 240W TDP.
I would look at reviews for EPYC 7713, usually they contain comparison to EPYC 7702. These ES will perform probably 10-15% better than 7702 but it will depend on your workload and obviously you won't get any official support from AMD/etc.
With that said, I have a Epyc 7713 with the MZ32-AR0 Rev. 1 with latest bios. How are you cooling it? Also, can you run cinebench multi-core test and see what you max boost freq is in HWinfo for all cores enabled. Just looking for comparison.

Just a note, you can see my build here (servethehome).

With all 64-core test, I can boost to 3.0ghz all core. Max. Boost Clock is up to 3.675GHz but it only works with half the cores enabled. Base Clock is 2.0GHz. Default TDP is 225W but you can override it bios. You have to cool the VRM heatsink to get higher TDP. Mines max's out at 250w. Also your finding with HWinfo with the temps is the same that I get. I think the epyc processors are just built that way. As a guidence, with my overkill watercooling setup, it will hit 90-95C with multi-core test in cinebench R23. I don't worry because the water temp only goes up a degree. The board will ramp all fans that is connected the motherboard just in case. And I have 16 fans in push pull config with two 480 radiators.
 

Advin

Member
Jun 10, 2023
50
18
8
With that said, I have a Epyc 7713 with the MZ32-AR0 Rev. 1 with latest bios. How are you cooling it? Also, can you run cinebench multi-core test and see what you max boost freq is in HWinfo for all cores enabled. Just looking for comparison.

Just a note, you can see my build here (servethehome).

With all 64-core test, I can boost to 3.0ghz all core. Max. Boost Clock is up to 3.675GHz but it only works with half the cores enabled. Base Clock is 2.0GHz. Default TDP is 225W but you can override it bios. You have to cool the VRM heatsink to get higher TDP. Mines max's out at 250w. Also your finding with HWinfo with the temps is the same that I get. I think the epyc processors are just built that way. As a guidence, with my overkill watercooling setup, it will hit 90-95C with multi-core test in cinebench R23. I don't worry because the water temp only goes up a degree. The board will ramp all fans that is connected the motherboard just in case. And I have 16 fans in push pull config with two 480 radiators.
I am just using a 1U server with a passive SP3 heatsink with some 11K+ RPM fans blowing at it
I think IPMI and lm-sensors (Linux) should report the correct CPU temps

Will also check max boost freq in a min