Question Zen 4 builders thread

Page 52 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Timmah!

Golden Member
Jul 24, 2010
1,396
604
136
My daughter says her new PC is noticeably faster than her previous one. Programs open faster, updates install faster, browsing the web is faster, etc. Both of my kids PCs load games faster than mine, noticeably faster. Updates install significantly faster.

Her Previous system
Intel i7-4790
12GB DDR3
GTX 1660
500GB 2.5" Crucial SSD (OS)
1TB WD 7200RPM HDD (Games)

Her New system
7600X
32GB DDR5 6000
RX6800XT
1TB Solidigm P41 NVME (OS)
2TB Solidigm P41 NVME (GAMES)

My old as hell PC
i7-920 OC to 3.6GHz
24GB DDR3 1333
GTX 1060
500GB Samsung 2.5" SSD (OS)
1TB WD 7200RPM HDD (Games)

Thats a nice system you built for your daughter.

My biggest gripe with my current system is how long can take for my working CAD-related apps to open/close or save stuff. The other day i was working on some interior design, downloaded some nice bed 3D models, but needed to convert them to different materials and then save...and then 3D studio decided for whatever reason it needs like 2-3 minutes to close. I needed to close it and start again, because ofc it would crash if i tried to convert 2 models without doing so inbetween - but i am not even mad about that. What was annoying was the wait just to close that relatively simple scene with single bed model. Since i had 5 of these models, it took me like 30 minutes to do so - when it really had no business taking that long.

I have to wonder if 7950x and DDR5 RAM will help to alleviate these issues.
 

mmaenpaa

Member
Aug 4, 2009
78
136
106
Thats a nice system you built for your daughter.

My biggest gripe with my current system is how long can take for my working CAD-related apps to open/close or save stuff. The other day i was working on some interior design, downloaded some nice bed 3D models, but needed to convert them to different materials and then save...and then 3D studio decided for whatever reason it needs like 2-3 minutes to close. I needed to close it and start again, because ofc it would crash if i tried to convert 2 models without doing so inbetween - but i am not even mad about that. What was annoying was the wait just to close that relatively simple scene with single bed model. Since i had 5 of these models, it took me like 30 minutes to do so - when it really had no business taking that long.

I have to wonder if 7950x and DDR5 RAM will help to alleviate these issues.

If you open task manager before you start closing programs and observe CPU usage & SSD usage you might get an idea.
Modern NVME SSD is many times faster than old SATA SSD. I would also bet that 12 generations progress will have a great impact on ST perfomance (="snappiness") be it Intel 13th or AMD 7XXX versions. I'd say about 4X faster in single core.

Simple comparision of those two CPUs

Core i7-920 vs Ryzen 9 7950X [in 8 benchmarks] (technical.city)

I'd say you would feel the difference :)
 

repoman0

Diamond Member
Jun 17, 2010
4,445
3,282
136
If you are consciously overclocking, you can feel 200 MHz difference.

Hard disagree. That’s less than 4% with Zen 4 CPUs. Nobody can tell that apart in a blind test. 1 GHz .. maybe an enthusiast, developer or graphic artist etc barely notices but certainly not everyone.

When I started this hobby I could buy a 1.8 GHz Opteron 165 and overclock it to 2.8-3 GHz. That’s in the neighborhood of 60% and noticeable, especially because PCs weren’t so fast then that delays were imperceptible. 200 MHz extra on top of 5700 MHz is a silly bonus for a chip already clocked near the ragged edge of stability straight from the factory.

I upgraded from a 5900X to 7950X mostly just for fun. It was a lot more noticeable than the CPU change when I later put a high end PCI-e 4.0 NVMe in to replace an aging gen 3 Samsung one.
 
Last edited:
  • Like
Reactions: Tlh97 and Timmah!

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
I don't know about 200MHz difference, but I definitely noticed an improvement in CPU-bound games (many multiplayer games) going from 5950X to 7700X. My avg FPS on the 5950X became the new min/0.1% low FPS in some cases.

It remains to be seen if 7950X3D or 7800X3D offer enough performance uplift to make an upgrade worth it. If not, there's always a 8700/8800X3D I can safely expect down the road.
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
This actually interests me. Do you people actually feel the difference in overall responsiveness of the system and whatnot? I mean, compared to your previous hardware. Especially those, who upgraded from some older system (not Ryzen 5000 or AlderLake), if there is anyone like that around here. I am aware its faster at Cinebench/rendering,encoding/decoding,compressing and decompressing... but thats not what you do all the time. What about normal computer usage? Browsing web, multitasking... Is the 5,7GHz speed noticeable compared to some 5 years old 4~4,5GHz system? Or is that more dependent on RAM/fast SSD?
From the 3900X to the 5950X I definitely felt the difference. 5950x -> 7950x wasn’t noticeable. Once you start utilizing the chip, however, you could definitely feel the difference.
 
  • Like
Reactions: Timmah!

Chaotic42

Lifer
Jun 15, 2001
33,929
1,097
126
So I was curious if using 960GB of Optane as a swap device would make having 64GB of memory at least somewhat bearable:

128GB @ 3600MT/s:
Code:
Coll:   217,000,000
Start:  2023-01-02 02:53:12.055788
Load:   0:00:00.000006
---Calc:   0:04:21.223420
---Trim:   0:05:22.942517
Write:  0:00:00.001014
Total:  0:09:44.166957
Maximum Memory Used (GB): 123.32

64GB @ 5600MT/s with Optane Swap:
Code:
Coll:   217,000,000
Start:  2023-01-05 12:38:17.727767
Load:   0:00:00.000158
---Calc:   0:14:42.951998
---Trim:   1:07:36.651591
Write:  0:00:00.008853
Total:  1:22:19.612600
Maximum Memory Used (GB): 54.69
Configured Memory Speed: 5600 MT/s
Volatile Size: 32 GB
Configured Memory Speed: 5600 MT/s
Volatile Size: 32 GB

Survey says..... Negative :p
 

lopri

Elite Member
Jul 27, 2002
13,209
594
126
My goodness that is almost absurd. That is like 88% of theoretical maximum out of DDR5-6600!

One question though, does anyone know what IF's (Infinity Fabric) bandwidth is? I always thought IF is a bottleneck on this platform, be that for memory or for PCIE. And the reason why I often see 64GB/s for a single CCD is also because 2:3 = fclk:mclk gets you 2/3 of theoretical max of DDR5-6000 (96GB/s).

But apparently that is not the case and I gotta correct my assumption.
 
Last edited:

lopri

Elite Member
Jul 27, 2002
13,209
594
126
Thats a nice system you built for your daughter.

My biggest gripe with my current system is how long can take for my working CAD-related apps to open/close or save stuff. The other day i was working on some interior design, downloaded some nice bed 3D models, but needed to convert them to different materials and then save...and then 3D studio decided for whatever reason it needs like 2-3 minutes to close. I needed to close it and start again, because ofc it would crash if i tried to convert 2 models without doing so inbetween - but i am not even mad about that. What was annoying was the wait just to close that relatively simple scene with single bed model. Since i had 5 of these models, it took me like 30 minutes to do so - when it really had no business taking that long.

I have to wonder if 7950x and DDR5 RAM will help to alleviate these issues.

Back in the days people used to make ramdisks to speed up I/O. Not sure if it is worthwhile these days when PCIE 4.0 NVME gives nearly 7.0GB/s.

Hard disagree. That’s less than 4% with Zen 4 CPUs. Nobody can tell that apart in a blind test. 1 GHz .. maybe an enthusiast, developer or graphic artist etc barely notices but certainly not everyone.

When I started this hobby I could buy a 1.8 GHz Opteron 165 and overclock it to 2.8-3 GHz. That’s in the neighborhood of 60% and noticeable, especially because PCs weren’t so fast then that delays were imperceptible. 200 MHz extra on top of 5700 MHz is a silly bonus for a chip already clocked near the ragged edge of stability straight from the factory.

I upgraded from a 5900X to 7950X mostly just for fun. It was a lot more noticeable than the CPU change when I later put a high end PCI-e 4.0 NVMe in to replace an aging gen 3 Samsung one.
I mean when you are in the middle of clocking. Sure I won't tell the difference of 200 MHz for daily use, but say, the moment I went to Infinity Fabric 1933 MHz from 1800 MHz on my 3700X, it was unmistakable which configuration was more responsive.
 

cellarnoise

Senior member
Mar 22, 2017
709
394
136
Screenshot 3 and 4 should show everything you need: Let me know if your still missing something..
I seem to have many more hours of playing to go. Though I am not shooting for the moon. Just a little better than 30-38-38 expo on ram that does not appear to perform nearly as well as yours, and I am not really looking to overvolt as much on the RAM or cpu/soc. Just want a little tighter timings from 6000 expo 30-38....

Your scores are the best I have seen though after a quick search. Congrats!
 

lopri

Elite Member
Jul 27, 2002
13,209
594
126
Finally gotten back to it and now I am done testing my system. Curve Optimizer and holidays threw a curve ball so it took longer than I planned, but after everything is said and done I am satisfied. Hopefully the system will serve me well for the next couple of years.. well, who knows? Let's see what 7800X3D can do :p

I have settled on the followings:
  • Eco Mode 65W
  • Boost override +200 MHz
  • Mid load boost (?) Enabled
  • Per core Curve Optimizer (-20, -14, -18, -18, -20, -20, -16, -16)
There are also some VRM settings I fiddled around, such as LLC. (level 3 for both CPU and SOC, level 2 got me occasional shutdowns) Other than that, most other voltages were lowered as much as possible.

This setting gets me;
  • Max Package power: 90.2 W (never seen it go above)
  • Max Boost Clock: 5750 MHz
  • Avg. Effective Clock during Prime95 (LargeFFT, AVX-512): 5120 MHz. It's not all-core OC because I see minimum effective frequencies below 4000 MHz after a while. I suppose cores take a break here and there. With regular PBO and boost override, average effective clock during Prime95 was 5340 MHz and max PPT rose up to 160W(!). That is 78% power increase for extra 220 MHz. o_O
  • Max temperature: 75C for 1T max benches, 70C for full-blown nT stuff. 75C is also generally the max temp I observe on day-to-day use.
Torture tests passed
  • BIOS Memtest: 8 hrs
  • Prime95 LargeFFT (AVX-512): 8 hrs
  • Y-Cruncher (all tests): 4 hrs
  • 3DMark Time Spy: 2 hrs
Tests were done with all 4 DIMM slots filled, all 4 M.2 sockets used, and 2 out of 3 PCIE slots occupied. On average, the hottest running component in this setup is the X670 chipset @60~80C (ugh)

Performance result

Cinebench 30 min run: Eco Mode 65W 19174 | Stock Boost + PBO 20159
Cinebench R23.jpg

Geekbench 5: Eco Mode 65W (2305 ST, 14777 MT) | Stock Boost + PBO (2263 ST, 15609 MT) ->

Essentially the same performance for single-threaded tasks, and 5% performance loss for the Eco mode. System has been rock solid. ASUS B650E-E gets a seal of approval from me when it comes to handling a 7700X. If it were cheaper by $100, it'd be a no-brainer recommendation.
 
Last edited:

CP5670

Diamond Member
Jun 24, 2004
5,508
586
126
Are there still any USB issues with the AMD platform? The B550/X570 boards seemed to have quite a few reports of USB dropouts and connection issues, particularly with the Reverb G2 VR headset. I'm wondering if either B650E or X670E is better/worse with this.

I was also reading that boot times are longer on DDR5 platforms in general. I did notice that going from Z97 to Z490 in my last upgrade resulted in a big increase in boot times.
 

Dave3000

Golden Member
Jan 10, 2011
1,343
91
91
Are there still any USB issues with the AMD platform? The B550/X570 boards seemed to have quite a few reports of USB dropouts and connection issues, particularly with the Reverb G2 VR headset. I'm wondering if either B650E or X670E is better/worse with this.

I was also reading that boot times are longer on DDR5 platforms in general. I did notice that going from Z97 to Z490 in my last upgrade resulted in a big increase in boot times.

I wonder if my iPhone 12 having USB dropouts with my Asus B550-F Strix motherboard is related to the USB issue that plagues some B550/X570 boards. I have BIOS 2806 installed which is the latest version but still experience the dropout issue when my iPhone 12 is connected through any USB port on my motherboard but not Wi-Fi using through an 3rd party app (Wi-Fi Photo) installed on my iPhone. Another workaround is using a USB-A to USB-C adapter and connecting it through USB-C to my PC but strangely I still experience this issue with a Lightening/USB-C cable. In my next CPU/motherboard upgrade, if sticking with AMD, I'm seriously considering going X670-E if it does not have the USB dropout issues. I heard the X670-E is basically 2 B650-E chipsets, so that means that if the B650-E chipset has the USB dropout issue then the X670-E would have it to? Maybe I should go for X670-E instead of B650-E anyways since I will be using 2 NVMe drives (one for OS/applications and one for games) and one SATA SSD? If the USB issue still exists on X670-E and B650-E, I think I will upgrade to an i7-13700k with a Z790 motherboard for my next CPU/motherboard upgrade. I need a system with reliable USB connections with every USB device for my next CPU/motherboard upgrade.
 
  • Like
Reactions: CP5670