Question The new threadrippers 7980X and 7970X

[aigomorla]

That is, overclocking doesn't void the warranty, but may make customers' claims more easily contested by AMD. But keep in mind that AMD advertises overclocking support by Threadripper.

Myself I don't really understand how you can support and advertise something that will void the warranty.

Because its very difficult to prove you overclocked properly, unless were talking about EXPO and CPU Boosting done by bios naturally.

As i said again its up to how AMD wants to handle this.
They aren't even going on public statement about it in regards to it.
To me, it feels like they are waiting to see how many people will complain about it.

Personally im from the older generation where, if you break it, you pay for it.
I have lost a few high end videocards because of water damage from water cooling gear mishaps, that i have all paid for and never attempted an RMA, unless the card itself was broken.

So if i break something from overclocking, i will pay for it, if i did something stupid like pump a unimaginable amount of vcore enough to power the deathstar's main laser cannon.

But if im doing a modest boost, and the cpu dies, and i get denied a RMA because i fused a resister on my modest boost, well, id like to see how durable these cpu's are before i even buy one then, just to cover myself.

Again its all up to AMD on how they enforce it, if they do enforce it.
But logic states, you would not put a tamper seal on something, unless you wanted to catch the user tampering it so you can reject a request.
Its like Samsung putting wet seals on there IP65 phones, even tho they have IP65 raiting.
And i have seen samsung deny RMA on water damaged phones, even tho they list the phone as water resistant.

 

[iamgenius]

Okay.....Then the best thing to do is to wait and see how things will fold out and decide based on that. >>1000 USD cpus better be durable enough to handle miniature mistakes in overclocking.

 

[StefanR5R]

As long as you keep Voltages in check — which I don't know how difficult this is considering the shenanigans of BIOSes from Ryzen mainboard makers —, you should be good. Temperatures are another point to watch, obviously, but I presume that built-in thermal protections are quite thorough nowadays.

 

[Shmee]

Sigh I want to see HEDT options for gaming...

 

[StefanR5R]

From Gavin Bonshor's HP Z6 G5 A workstation review:

Alas I couldn't find y-cruncher v0.8.2.9523 anywhere for download, so I took v0.8.3.9532 instead (statically linked Linux version from http://numberworld.org/y-cruncher/, the dynamically linked version did not work for me).

First discrepancy: There is no 5M test in the version which I downloaded. These are the only supported sizes:

Benchmark Pi:

    {program} bench {size} [-o {output path}]
                           [-od:{output digits}]
                           [-TD:{tasks}]
                           [-PF:{framework}]
                           [-MA:{allocator}]

    y-cruncher bench 25m
    y-cruncher bench 1b -TD:8
    y-cruncher bench 1b -TD:1 -PF:none
    y-cruncher bench 1b -od:0

The valid values for {size} are:
    25m, 50m, 100m, 250m, 500m,
    1b, 2.5b, 5b, 10b, 25b, 50b, 100b, 250b, 500b,
    1t, 2.5t,
    1M, 2M, 4M, 8M, 16M, 32M, 64M, 128M, 256M, 512M,
    1G, 2G, 4G, 8G, 16G, 32G, 64G, 128G

The 4M and 8M tests are very quick, so I tried the 5b test.
Machine is an EPYC 9554P (64 cores), set to 400 W cTDP, with all 12 RAM channels populated. Result:

Total Computation Time: 30.629 seconds ( 0.510 minutes )
Start-to-End Wall Time: 36.628 seconds ( 0.610 minutes )​

So the next thing which I am uncertain about is whether Gavin Bonshor reported computation time, or start-to-end time.

Full output from my test:

$ time "y-cruncher v0.8.3.9532-static/y-cruncher" bench 5b
y-cruncher v0.8.3 Build 9532

Detecting Environment...

Hardware Features:
(*) Indicates it is used explicitly by y-cruncher.

CPU Vendor:
    AMD         = Yes
    Intel       = No

OS Features:
  * 64-bit      = Yes
  * OS AVX      = Yes
  * OS AVX512   = Yes

Hardware Features:
    MMX         = Yes
  * x64         = Yes
  * ABM         = Yes
    RDRAND      = Yes
    RDSEED      = Yes
  * BMI1        = Yes
  * BMI2        = Yes
  * ADX         = Yes
    PREFETCHW   = Yes
    PREFETCHWT1 = No
    RDPID       = Yes
    SERIALIZE   = No
    CLDEMOTE    = No

SIMD: 128-bit
  * SSE         = Yes
  * SSE2        = Yes
  * SSE3        = Yes
  * SSSE3       = Yes
    SSE4a       = Yes
  * SSE4.1      = Yes
  * SSE4.2      = Yes
    PCLMULQDQ   = Yes
    AES         = Yes
    SHA         = Yes
    GFNI        = Yes
    SM3         = No

SIMD: 256-bit
  * AVX             = Yes
  * FMA4            = No
    XOP             = No
  * FMA3            = Yes
  * AVX2            = Yes
    VAES            = Yes
    VPCLMULQDQ      = Yes
  * AVX-GFNI        = Yes
    AVX-VNNI        = No
    AVX-VNNI_INT8   = No
    AVX-IFMA        = No
    AVX-NE-CONVERT  = No
    AVX-VNNI_INT16  = No
    SHA512          = No
    SM4             = No

SIMD: 512-bit
  * AVX512-F                = Yes
    AVX512-CD               = Yes
    AVX512-PF               = No
    AVX512-ER               = No
  * AVX512-VL               = Yes
  * AVX512-BW               = Yes
  * AVX512-DQ               = Yes
    AVX512-VPOPCNTDQ        = Yes
    AVX512-4VNNIW           = No
    AVX512-4FMAPS           = No
  * AVX512-IFMA             = Yes
  * AVX512-VBMI             = Yes
    AVX512-VNNI             = Yes
  * AVX512-VBMI2            = Yes
    AVX512-BITALG           = Yes
    AVX512-VPCLMULQDQ       = Yes
    AVX512-VAES             = Yes
  * AVX512-GFNI             = Yes
    AVX512-BF16             = Yes
    AVX512-VP2INTERSECT     = No
    AVX512-FP16             = No

APX + AVX10:
    APX-F           = No
    AVX10           = No
    AVX10/128       = No
    AVX10/256       = No
    AVX10/512       = No

AMX:
    AMX-TILE        = No
    AMX-INT8        = No
    AMX-BF16        = No
    AMX-FP16        = No
    AMX-COMPLEX     = No


Auto-Selecting: 22-ZN4 ~ Kizuna

/home/xyz/Y-Cruncher/y-cruncher v0.8.3.9532-static/Binaries/22-ZN4 ~ Kizuna


Launching y-cruncher...
================================================================



Insufficient permissions to set thread priority. Please retry as root.

Further messages for this warning will be suppressed.

Checking processor/OS features...

Required Features:
    x64, ABM, BMI1, BMI2, ADX,
    SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
    AVX, FMA3, AVX2,
    AVX512-(F/CD/VL/BW/DQ/IFMA/VBMI/GFNI)



Parsing Core -> Handle Mappings...
    Cores:  0-127 

Parsing NUMA -> Core Mappings...
    Node  0:  0-127 



This process does not have "CAP_IPC_LOCK". Page locking will not be possible.
Please run y-cruncher with elevation to enable page locking.


Constant:   Pi
Algorithm:  Chudnovsky (1988)

Decimal Digits:       5,000,000,000
Hexadecimal Digits:   Disabled

Computation Mode:     Ram Only
Multi-Threading:      Push Pool  ->  128 / ?  (randomization on)

Start Time: Sun Dec 17 17:27:26 2023

Working Memory...  22.5 GiB  (spread: ?)                                       
Twiddle Tables...  10.9 MiB  (spread: ?)                                       

Begin Computation:

Series CommonP2B3...  352,568,359 terms  (Expansion Factor = 2.847)
Time:    26.131 seconds  ( 0.436 minutes )                                     
Large Division...
Time:    1.227 seconds  ( 0.020 minutes )                                      
InvSqrt(10005)...
Time:    0.830 seconds  ( 0.014 minutes )                                      
Large Multiply...
Time:    0.528 seconds  ( 0.009 minutes )                                      

Pi:  28.717 seconds  ( 0.479 minutes )

Base Converting:
Time:    1.912 seconds  ( 0.032 minutes )                                      
Writing Decimal Digits:
Time:    4.135 seconds  ( 0.069 minutes )                                      

Start Time: Sun Dec 17 17:27:26 2023
End Time:   Sun Dec 17 17:28:03 2023

Total Computation Time:     30.629 seconds  ( 0.510 minutes )
Start-to-End Wall Time:     36.628 seconds  ( 0.610 minutes )

CPU Utilization:        10673.16 %  +   22.79 % kernel overhead
Multi-core Efficiency:     83.38 %  +    0.18 % kernel overhead

Last Decimal Digits:  Pi
4384678622 1397184596 0181195416 0748430457 5386741865  :  4,999,999,950
0914971996 1298184401 9216126684 9425834935 5440797257  :  5,000,000,000

Spot Check: Good through 5,000,000,000

Version:              0.8.3.9532 (Linux/22-ZN4 ~ Kizuna)
Processor(s):         AMD EPYC 9554P 64-Core Processor 
Topology:             128 threads / 64 cores / 1 socket / 1 NUMA node
Usable Memory:        405,230,354,432 ( 377 GiB)
CPU Base Frequency:   3,099,990,080 Hz

Validation File: Pi - 20231217-172803.txt


real    0m36.986s
user    54m46.962s
sys     0m9.339s

PS, here is AnandTech's data collection of 2024 CPU bench results in the so-called "(5-1b) y-cruncher 0.8.2.9523: MT (5M Pi)" test:
https://www.anandtech.com/bench/CPU-2024/3387
Maybe somebody can reproduce this on a computer which matches one of these entries.

 

[StefanR5R]

As far as I can tell from the photos, the HP Z6 G5 A has got a slim dual-tower CPU cooler with a 120 mm fan between the two fin stacks, helped by the rear fan which seems to be a 140 mm fan. For 350 W heat to dissipate during multithreaded loads, this seems to be a weak cooling solution to me.

Besides, the 96-core Threadripper would certainly be able to stretch its legs at power targets way above the stock TDP. Around 500 W, it should still have very reasonable power efficiency for a workstation CPU.

 

[aigomorla]

Okay.....Then the best thing to do is to wait and see how things will fold out and decide based on that. >>1000 USD cpus better be durable enough to handle miniature mistakes in overclocking.

Im more waiting to see if the CPU doesn't melt though the socket on EXPO like the X3D did.
I would really eat at me if i couldn't RMA the CPU because EXPO wrecked my CPU, (which btw were sort of mostly stuck on ASUS for the TRX50 unless you want a really bad scaled down board.)
And im not really 100% confident in ASUS + AMD combo's.

As long as you keep Voltages in check — which I don't know how difficult this is considering the shenanigans of BIOSes from Ryzen mainboard makers —, you should be good. Temperatures are another point to watch, obviously, but I presume that built-in thermal protections are quite thorough nowadays.

Its not the overclock im worried about, its the EXPO boosting im more concerned about.
Unless they did a revision on the CPU socket, i am worried it will be another repeat of the X3D.

Hence why im just waiting to see if we see any issues or problems before i decide on my selection.

Sigh I want to see HEDT options for gaming...

Me too, before HEDT and High End Gaming, didn't have such a big margin.
But the 7800X3D and 14900K completely blows these HEDT cpu's out of the water in PC gaming performance.
I want a X3D or 14th Gen cpu with just 64 PCI-E 5.0 lanes, and that would make me super happy.

 

[iamgenius]

Im more waiting to see if the CPU doesn't melt though the socket on EXPO like the X3D did.
I would really eat at me if i couldn't RMA the CPU because EXPO wrecked my CPU, (which btw were sort of mostly stuck on ASUS for the TRX50 unless you want a really bad scaled down board.)
And im not really 100% confident in ASUS + AMD combo's.



Its not the overclock im worried about, its the EXPO boosting im more concerned about.
Unless they did a revision on the CPU socket, i am worried it will be another repeat of the X3D.

Hence why im just waiting to see if we see any issues or problems before i decide on my selection.



Me too, before HEDT and High End Gaming, didn't have such a big margin.
But the 7800X3D and 14900K completely blows these HEDT cpu's out of the water in PC gaming performance.
I want a X3D or 14th Gen cpu with just 64 PCI-E 5.0 lanes, and that would make me super happy.

I wasn't aware of this X3D EXPO issue. I can see they fixed it with a BIOS update. So it was a mistake on AMD's part or board makers part? Did they compensate for the burned cpus? This sounds more like a serious bug which was fixed after being reported. If EXPO or XMP or PBO end up causing a threadrippers to burn, then we shouldn't take the blame. They are all features offered by manufacturers.

Are we going to see more options when it comes to motherboards as time goes on? I want a good a cpu+mobo combination that works and very reliable.

 

[aigomorla]

So it was a mistake on AMD's part or board makers part?

They started pointing finger at each other saying it was ASUS's fault, and ASUS blaming AMD for falsely disclosing real limits.
Later on everyone shafted ASUS because ASUS has the worst customer service in that the customer is always wrong, not to mention they have the worst RMA department, with people logging in and proving ASUS deliberately incited more damage to the board, blamed it on the mail carrier, so they could deny RMA claims.

In short, i don't trust ASUS, but the board options on Threadripper is FAR superior then all the other vendors.
Its the classic statement, of "your damn'd if you do, and damn'd if you don't".

Lastly, I don't call the EXPO fix a real fix.
They had to put big limiters on the cpu then what was originally intended, and i believe i heard multipul avenues say the entire platform required a pin layout change to disrtibute the power for a real fix.

But that would be like what intel did that pissed a lot of us off when they went from lga1150 to LGA1151.
1 stupid pin required an entirely new socket, new boards, and new nightmare of things to address poor power delivery.

So in all i am not fully confident in this FIX, until i see some history on it.
You'll be sad when you lost a 600 dollar cpu along with a 400 dolllar motherboard, but imagine if you scaled that up to a 1700 dollar cpu and a 1200 dollar motherboard.

What face would you put on then?


Oh btw more information on Threadripper Pro has finally come out.
The 7000WX edition with 128 PCI-E 5.0 lanes.

AMD Ryzen Threadripper Pro 7000WX at 96 Cores and Threadripper 7000 HEDT

The new AMD Ryzen Threadripper Pro 7000WX and Threadripper 7000 parts are launched bringing Genoa-era CPUs to the workstation market

www.servethehome.com

@Shmee and no its not meant for gaming.. 😖
I vote we voice out a demand for HEGP... High End Gaming Platform.

They could at least pull it off using PLX chips on the motherboard for the nVME's, or give us cpu's which have a PLX bridge to connect to PLX on motherboards to give us as many extra lanes as needed.

 

[iamgenius]

They started pointing finger at each other saying it was ASUS's fault, and ASUS blaming AMD for falsely disclosing real limits.
Later on everyone shafted ASUS because ASUS has the worst customer service in that the customer is always wrong, not to mention they have the worst RMA department, with people logging in and proving ASUS deliberately incited more damage to the board, blamed it on the mail carrier, so they could deny RMA claims.

In short, i don't trust ASUS, but the board options on Threadripper is FAR superior then all the other vendors.
Its the classic statement, of "your damn'd if you do, and damn'd if you don't".

Lastly, I don't call the EXPO fix a real fix.
They had to put big limiters on the cpu then what was originally intended, and i believe i heard multipul avenues say the entire platform required a pin layout change to disrtibute the power for a real fix.

But that would be like what intel did that pissed a lot of us off when they went from lga1150 to LGA1151.
1 stupid pin required an entirely new socket, new boards, and new nightmare of things to address poor power delivery.

So in all i am not fully confident in this FIX, until i see some history on it.
You'll be sad when you lost a 600 dollar cpu along with a 400 dolllar motherboard, but imagine if you scaled that up to a 1700 dollar cpu and a 1200 dollar motherboard.

What face would you put on then?


Oh btw more information on Threadripper Pro has finally come out.
The 7000WX edition with 128 PCI-E 5.0 lanes.

AMD Ryzen Threadripper Pro 7000WX at 96 Cores and Threadripper 7000 HEDT

The new AMD Ryzen Threadripper Pro 7000WX and Threadripper 7000 parts are launched bringing Genoa-era CPUs to the workstation market

www.servethehome.com

@Shmee and no its not meant for gaming.. 😖
I vote we voice out a demand for HEGP... High End Gaming Platform.

They could at least pull it off using PLX chips on the motherboard for the nVME's, or give us cpu's which have a PLX bridge to connect to PLX on motherboards to give us as many extra lanes as needed.

So people just lost their cpus because of this issue! My God that's too harsh. And if it happens again with ~5000 USD cpus it is going to be a disastor.

Isn't the article you linked an older one? It is the same which we got at first.

Yeah, HEGP sounds good.