Question Zen 4 builders thread

[Markfw]

I know its early, but as someone pointed out, I better get my ram now before the scalpers are all over it.

To go with a 7950x, what do you think of this ram:

https://www.newegg.com/g-skill-32gb-288-pin-ddr5-sdram/p/N82E16820374369?Item=N82E16820374369

(edited for my best recommendation, 12/5/2023)

 

[igor_kavinski]

It takes time and knowledge to tweak every setting to get there.

That might be a headache for some

 

[CakeMonster]

FYI, ASUS may be in trouble. Would advise keeping an eye on the news over the next few weeks. Rumor is > one person let go, forums shut down, etc. Possible widespread layoffs and shutdowns. Hope that members of the press are looking into it.

Hmm, interesting. I've been annoyed that the BIOS thread that I had bookmarked was down. Also, kind of surprising that the BIOS releases has slowed in a time that is typically so busy. And, a small thing, they removed the dates on BIOSes from the BIOS download pages, that just added confusion for me.

 

[eek2121]

Hmm, interesting. I've been annoyed that the BIOS thread that I had bookmarked was down. Also, kind of surprising that the BIOS releases has slowed in a time that is typically so busy. And, a small thing, they removed the dates on BIOSes from the BIOS download pages, that just added confusion for me.

We will see how it shakes out. FWIW Shamino posted new betas in an OCN thread.

 

[lopri]

4 sticks at 6000MT/s and CL36 work just fine, too. Different memory ICs no less. (Hynix+Samsung) Now I am confused.

 

[lopri]

Granted I did countless CMOS clear but that was because I was pushing things around. If all I wanted was 6000/36-36-36, no CMOS clear necessary

 

[igor_kavinski]

4 sticks at 6000MT/s and CL36 work just fine, too. Different memory ICs no less. (Hynix+Samsung) Now I am confused.

That should not be possible. You must have gotten a golden sample CPU if all RAM stress tests pass.

 

[lopri]

Maybe it's my sk... lol kidding. I am still testing it out. But initial impression is good. Temp goes significantly higher than with 2 sticks, both for the CPU and memory and I have a fan blowing over the memory sticks for now. It's my first time with DDR5 and so far I am not liking the heat profile as well as OEMs excessive overvolting in their EXPO/XMP profiles. EXPO/XMP profiles are crappy anyway.

1 hour running:


Ending:


AIDA64 (64GB):


Pi 1M:


It does 1T command rate and that was a surprise. I do not see anywhere in BIOS an option to change command rate, is this normal? Also, I can't seem to lower tRFC below 960 with 4 sticks, which kinda sucks.

Edit: Does anyone know what "North Bridge Clock" refers to in AIDA Cache & Memory Benchmark?

 

[igor_kavinski]

I have a fan blowing over the memory sticks for now.

Am I right that the XMP Hynix DIMMs are hotter than the EXPO Samsung DIMMs?

 

[lopri]

No. With 4 sticks they are about the same so far. With 2 sticks, Hynix is cooler because it required less voltage. But the difference was not big and Samsung did tighter tRCD.

Edit: Actually I am not sure. I will have to double-check. I know that I ran Hynix with 1.26V and Samsung with 1.30~1.35V, though (when testing the kits separately). It may well be the case Hynix runs hotter with same voltage. I will find out.

 

[lopri]

I did not have a chance to do a controlled comparison, but judging by the screenshots I took,

Hynix: 1.26V, 6000/32-38-38, 3.5W load
Samsung: 1.35V, 6000/32-34-34, 5.5W load

Hard to compare temperatures because I do not remember when I started using a fan over the DIMMs. But I did that because I saw temperature encroaching 70C without air flow.

 

[DrMrLordX]

Hard to compare temperatures because I do not remember when I started using a fan over the DIMMs. But I did that because I saw temperature encroaching 70C without air flow.

One of the reasons why I do not like the current airflow designs of modern glass-sided cases. Side case fans were perfect for that sort of cooling.

 

[biostud]

One of the reasons why I do not like the current airflow designs of modern glass-sided cases. Side case fans were perfect for that sort of cooling.

I'm actually wondering if I should choose a case with a side mesh with two 140mm fans or glass, for my next build.

 

[VirtualLarry]

Are you a human?

www.newegg.com

What do you guys think about the B650 Tomahawk from MSI? I'm kind of vaguely a fan of the Tomahawk series of boards. They're fairly spartan on slot amenities compared to an Asus Bx50-F board (my current board), but I like the back-panel I/O layout, and the BIOS flashback button is IMHO helpful.

It's not all that much more expensive than the B550 Tomahawk board, either.

What am I realistically giving up with a B650 instead of an X670(E)? PCI-E 5.0 on the GPU slot? My take on that is, unless there's some new mega-texture-streaming engine that plays back live video streams over the internet in Ray-Traced 3D or something (Metaverse?), that there won't be a real need for PCI-E 5.0 for some time yet, at least for traditional video games. (If the PCI-E 4.0 x4 interface of the RX 6400 is any indication.)

Was thinking maybe Jan/Feb. Would I be better off waiting for the Zen 4 X3D processors, will they be out by then? I want a 7950X3D for DC work, the extra cache will help primegrid.

My biggest hassle, will probably be making certain that whatever PSU that I get for this future rig, has TWO EPS12V connectors on it.

 

[DrMrLordX]

I'm actually wondering if I should choose a case with a side mesh with two 140mm fans or glass, for my next build.

Good luck finding cases with side fans. They're hard to find nowadays. Especially nice 280mm fans.

 

[Markfw]

Are you a human?

www.newegg.com

What do you guys think about the B650 Tomahawk from MSI? I'm kind of vaguely a fan of the Tomahawk series of boards. They're fairly spartan on slot amenities compared to an Asus Bx50-F board (my current board), but I like the back-panel I/O layout, and the BIOS flashback button is IMHO helpful.

It's not all that much more expensive than the B550 Tomahawk board, either.

What am I realistically giving up with a B650 instead of an X670(E)? PCI-E 5.0 on the GPU slot? My take on that is, unless there's some new mega-texture-streaming engine that plays back live video streams over the internet in Ray-Traced 3D or something (Metaverse?), that there won't be a real need for PCI-E 5.0 for some time yet, at least for traditional video games. (If the PCI-E 4.0 x4 interface of the RX 6400 is any indication.)

Was thinking maybe Jan/Feb. Would I be better off waiting for the Zen 4 X3D processors, will they be out by then? I want a 7950X3D for DC work, the extra cache will help primegrid.

My biggest hassle, will probably be making certain that whatever PSU that I get for this future rig, has TWO EPS12V connectors on it.

Most 850 watt and above, and at least one 750 have 2 eps12 connectors on it. I just got one for a build for a friend,

 

[biostud]

Good luck finding cases with side fans. They're hard to find nowadays. Especially nice 280mm fans.

This beauty:
North — Fractal Design (fractal-design.com)

 

[DrMrLordX]

This beauty:

Hmm. Not bad. It might work for my next build but I dunno. Will have to look at it some more to see how well I can utilize my external MO-RA3 420 with it. I would rather have a 280mm fan bracket, but can't always get what you want.

edit: can the fan bracket be shifted upwards? The pic showing the fan bracket has the fans aimed below the DIMMs.

edit edit: Looks like the rear has nowhere to route WC tubing for external rads. Sad face.

 

[AdamK47]

I decided to implement and test curve optimizer the right way. My previous attempts at it were to either set an all core value or a per CCD value. It's fine for all core loads. Not so good for lighter loads where single to a few cores get hit. Reboots and instability abound.

I run CoreCycler with Prime95 AVX512 large FFT with two threads per core for 6 minutes each. It finds errors faster vs. the default configuration.

The first round I set all cores to -10. After a few iterations of CoreCycler I found 3 cores that errored. Moved those to -5 and moved the rest to -15. Ran a few more iterations. The ones that failed I moved down by 5 and the ones that pass I moved up by 5. Did this for several cycles/iterations.

Three days later and I'm still testing. Two cores made it up to -35. The rest reside somewhere between that and -5. Going to run other types of tests after I complete a full 24 hours of CoreCycler error free.

 

[alexruiz]

Are you a human?

www.newegg.com

What do you guys think about the B650 Tomahawk from MSI? I'm kind of vaguely a fan of the Tomahawk series of boards. They're fairly spartan on slot amenities compared to an Asus Bx50-F board (my current board), but I like the back-panel I/O layout, and the BIOS flashback button is IMHO helpful.

It's not all that much more expensive than the B550 Tomahawk board, either.

What am I realistically giving up with a B650 instead of an X670(E)? PCI-E 5.0 on the GPU slot? My take on that is, unless there's some new mega-texture-streaming engine that plays back live video streams over the internet in Ray-Traced 3D or something (Metaverse?), that there won't be a real need for PCI-E 5.0 for some time yet, at least for traditional video games. (If the PCI-E 4.0 x4 interface of the RX 6400 is any indication.)

Was thinking maybe Jan/Feb. Would I be better off waiting for the Zen 4 X3D processors, will they be out by then? I want a 7950X3D for DC work, the extra cache will help primegrid.

My biggest hassle, will probably be making certain that whatever PSU that I get for this future rig, has TWO EPS12V connectors on it.

All AM5 boards have the BIOS flashback button, so big difference here.
I did see that the Tomahawk has S/PDIF out for audio. Not all B650 boards have it, so if you need, it is there.
This one also has a newer audio chip than most B650 boards. A lot of boards are using the older Realtek ALC897.
Something where I think MSI excels is in the number of fan headers, his one has CPU + 7.
I would say go ahead

 

[Det0x]

I decided to implement and test curve optimizer the right way. My previous attempts at it were to either set an all core value or a per CCD value. It's fine for all core loads. Not so good for lighters loads where single to a few cores get hit. Reboots and instability abound.

I run CoreCycler with Prime95 AVX512 large FFT with two threads per core for 6 minutes each. It finds errors faster vs. the default configuration.

The first round I set all cores to -10. After a few iterations of CoreCycler I found 3 cores that errored. Moved those to -5 and moved the rest to -15. Ran a few more iterations. The ones that failed I moved down by 5 and the ones that pass I moved up by 5. Did this for several cycles/iterations.

Three days later and I'm still testing. Two cores made it up to -35. The rest reside somewhere between that and -5. Going to run other types of tests after I complete a full 24 hours of CoreCycler error free.

Should set runtime @ 9min for y-cruncher avx512 (takes alittle under 9mins to run a complete y-cruncher cycle for each core)

This is what i ended up with when i have Asus mid-boost-it enabled : (worst/best core needed -6 to be stable in corecycler y-cruncher avx512 at +100mhz boostclock

 

[lopri]

No comment

 

[lopri]

BTW, Samsung is DIMM 0/2, and Hynix is 1/3 on HWiNFO. Both were fed 1.275V. Between the proximity with each other and a fan blowing over, there is no meaningful difference in temperature under load in this setup. Nevertheless, we can infer from the load power consumption figure (Samsung 3W, Hynix ~2.4W) that Samsung chips run likely hotter under a similar configuration.

Edit: I mixed them up in my original typing. Samsung is DIMM 0/2, and Hynix is DIMM 1/3

 

[VirtualLarry]

Are you a human?

www.newegg.com

This board also catches my eye, it's an ASRock B650 LiveMixer. No wifi, but other than that, it has three PCI-E 4.0 x16 (physical, last two are x4), three x4 NVMe (one PCI-E 5.0, two PCI-E 4.0), front USB 3.0 and Type-C, plenty of USB in back, though most are USB 2.0 (?). Bios flashback.

Edit: Oh, if PCI-E_3 slot is filled, M2_C is disabled, and vice-versa. So you get either/or 3x PCI-E 4.0 slots (and 2x NVMe), or 3x NVMe (and 2x PCI-E 4.0 slots).

Edit: Only two traditional SATA ports.


$209 right now, too.

 

[In2Photos]

Are you a human?

www.newegg.com

This board also catches my eye, it's an ASRock B650 LiveMixer. No wifi, but other than that, it has three PCI-E 4.0 x16 (physical, last two are x4), three x4 NVMe (one PCI-E 5.0, two PCI-E 4.0), front USB 3.0 and Type-C, plenty of USB in back, though most are USB 2.0 (?). Bios flashback.

Edit: Oh, if PCI-E_3 slot is filled, M2_C is disabled, and vice-versa. So you get either/or 3x PCI-E 4.0 slots (and 2x NVMe), or 3x NVMe (and 2x PCI-E 4.0 slots).

Edit: Only two traditional SATA ports.


$209 right now, too.

I went with the Gigabyte board for my daughter's build which has a few more USB 3 ports on the back over the ASRock. Also has WiFi, 3 PCIe x16 slots though 2 run at x1, 3 m.2, and 4 SATA. $229 but has a $10 rebate.
https://www.newegg.com/p/N82E168131...m_mmc=snc-social-_-sr-_-13-145-412-_-12132022

 

[lopri]

Are you a human?

www.newegg.com

What do you guys think about the B650 Tomahawk from MSI? I'm kind of vaguely a fan of the Tomahawk series of boards. They're fairly spartan on slot amenities compared to an Asus Bx50-F board (my current board), but I like the back-panel I/O layout, and the BIOS flashback button is IMHO helpful.

It's not all that much more expensive than the B550 Tomahawk board, either.

What am I realistically giving up with a B650 instead of an X670(E)? PCI-E 5.0 on the GPU slot? My take on that is, unless there's some new mega-texture-streaming engine that plays back live video streams over the internet in Ray-Traced 3D or something (Metaverse?), that there won't be a real need for PCI-E 5.0 for some time yet, at least for traditional video games. (If the PCI-E 4.0 x4 interface of the RX 6400 is any indication.)

Was thinking maybe Jan/Feb. Would I be better off waiting for the Zen 4 X3D processors, will they be out by then? I want a 7950X3D for DC work, the extra cache will help primegrid.

My biggest hassle, will probably be making certain that whatever PSU that I get for this future rig, has TWO EPS12V connectors on it.

I do think B650 is a better choice than X670 in general unless you do need extra connectivity X670 offers, but MSI is an exception here because their X670-P board is just so well spec'ed. Compared to the B650 Tomahawk, it's got:

8 layer PCB (vs 6)
More PCIE slots and flexibility
4 x M.2 slots (vs 3)
2 x USB Type-C ports (vs 1, but the Tomahawk does come with 2 more Type-A ports)

Especially 8 layer PCB part is relevant for memory overclocking and with the rest of feature set it's worth $30 premium, imo. I seriously considered X670-P before I ultimately decided on more PCIE 5.0 future-proofing (right).