Ryzen 4 Reviews Are Out.

Since I only upgrade my CPU / Mottherboard like once in 5 years, and video card every other generation, or sometimes every new gen,

Anyways, at this point probably best waiting for the upcoming X3D processor coming in Jan / Feb, and then we will have more motherboard options, and plenty my DDR5 options.
 
Since I only upgrade my CPU / Mottherboard like once in 5 years, and video card every other generation, or sometimes every new gen,

Anyways, at this point probably best waiting for the upcoming X3D processor coming in Jan / Feb, and then we will have more motherboard options, and plenty my DDR5 options.
I'm starting to get the feeling those might come very late to the party or have serious issues with heat management.
 
Is Zen 4 Too Hot for a Box Cooler? Wraith Spire + 7600X


While I am no fan of HUB, it's nice to see the processors can run on box coolers. Looks like that answered my questions about workstation grade cooling and their new processor lineup.
 
Awesome. Can you show us perf scaling in games with the 7700X stock and overclocked? I’m assuming running at 5.7GHz you’re way ahead of your 12900K?
Well to be honest,since I play PC Games at 4K there is no performance scaling. I even ran a 12900K at 5700Mhz and still got around the same performance.
Now 720p running show of the tomb raider on lowest settings where it is all CPU there is a difference.

The 7700X should run up to 5550Mhz out of the box on stock settings ,unless you have a Motherboard that limits boosting.
To be clear also, I find no PC Gaming difference at 4K from 7700X vs 12900K.I have not tested the 7700X yet on lower resolutions but if/when I do I can try out some performance scaling.
I can say that 7700X is the most satisfying CPU I have ever owned to date,runs awesome ,cool and no weird things happening with my PC.

Back to what you asked,I can run low resolutions on some games and post results from CPU clock speed alone.
 
https://www.newegg.com/mushkin-enhanced-32gb-288-pin-ddr5-sdram/p/N82E16820226971?quicklink=true
This kit should be Hynix and clock up to 6000Mhz. Hynix is Hynix, spend as little as possible.
That's awesome.

I just picked up a G.Skill 6000 MHz / CL30 kit from Micro Center since it's EXPO ready and I suck at memory OC:

https://www.gskill.com/specification/165/390/1661410135/F5-6000J3038F16GX2-TZ5NR-Specification


The price was steep at $290 but I reserved it yesterday and today the price went up to $420. So at least I got it at $290. ;) Must be part of MC's profit-where-they-can pricing (as people buy full systems there). Gonna sell my free 5600 MHz / CL36 32GB kit.
 
Like I said before, your 9700 is still a stout gaming performer. If you want X3D, wait for X3D

Yeah looking at some benchmarks and gaming at 4k or close to that res, it's still 95% video card than it is CPU. I have a RTX-3080, and good for now.

Most likely I may upgrade once I see the new X3D processor next year, or wait for the 14th gen Intel.
 
Seems like EXPO RAM profiles screw up something about ACPI. I don't have sleep mode enabled when selecting the EXPO profile, comes right back setting RAM back to auto. It also boots much, much faster just using the JEDEC profile, at least as fast as my AM4 build (a second or two). Not that it's as bad as some people say with the OC profile but it does take 10-15 seconds before anything shows up on screen. I'm just going to run DDR5-4800 for a few days to make extra sure the new build is stable, then play around with setting the timings, frequency and voltages manually to see if that makes it work.
 
Seems like EXPO RAM profiles screw up something about ACPI. I don't have sleep mode enabled when selecting the EXPO profile, comes right back setting RAM back to auto. It also boots much, much faster just using the JEDEC profile, at least as fast as my AM4 build (a second or two). Not that it's as bad as some people say with the OC profile but it does take 10-15 seconds before anything shows up on screen. I'm just going to run DDR5-4800 for a few days to make extra sure the new build is stable, then play around with setting the timings, frequency and voltages manually to see if that makes it work.
Man, that sucks. Early adopter tax. :( What board is it?
 
Man, that sucks. Early adopter tax. :( What board is it?

X670E Taichi, but I found posts on amd.com from users of MSI and Gigabyte boards as well. I'm guessing it will be an AGESA level fix.

Luckily I can't actually tell the difference between 4800 and 5600MHz and am just here to play with new computing hardware.
 
Seems like EXPO RAM profiles screw up something about ACPI. I don't have sleep mode enabled when selecting the EXPO profile, comes right back setting RAM back to auto. It also boots much, much faster just using the JEDEC profile, at least as fast as my AM4 build (a second or two). Not that it's as bad as some people say with the OC profile but it does take 10-15 seconds before anything shows up on screen. I'm just going to run DDR5-4800 for a few days to make extra sure the new build is stable, then play around with setting the timings, frequency and voltages manually to see if that makes it work.
I noticed something similar on my steel legend. When coming back in the morning it is sitting there idle and awake. I thought it had something to do with using the windows install from my 9900k, because sleep and turbo never worked on that system. Setting it to 5ghz broke sleep and turbo settings on its board and I gave up trying to get it to work as forcing the issue caused blue screens. I had heavily messed with windows power settings a while back trying to get it to work for the last time. I was planning on doing a clean install as soon as I get a chance. Just a pain with all of my work software that is licensed.

Cold boot with expo at 6000mhz is definitely slower, I wouldn’t say it is 10-15 seconds to post though. My old 1366 Xeon and Ryzen 1700 were worse about cold boot times if I remember clearly.

I’m going to keep mine at 6000 though. High 70s latency vs mid 60s after tweaking timings is worth shutting the thing down if I need to. Plus if I am caught up with work. I don’t want the machine going to sleep and slack showing me offline lol.
 
I just remembered what the issue was with my 9900k. It needed 1.28 volts to run 5ghz all core and adaptive voltage needed to be on for turbo and sleep. It was damn hard to dial in the voltage for adaptive as it liked to overshoot significantly and reducing the voltage to counter act that caused it to undershoot and crash. I remember trying + .05 volts and it yielding 1.52 volts in CPU-Z…. Damn wonky voltage tables…
 
I confess I am a resolution racist.I post 4K PC gaming videos online and I always get asked questions like this especially every time new PC gear is released.I hate 1080p
Can you do Gameplays on lower resolution? You limited by the GPU. It's interesting to see how well it's scale performance with frequency uplift

So I clapped back with how about AMD 7700X 5700Mhz all core at 1.30v and 6200Mhz ram and 2200Mhz Infinity fabric with recording straight from desktop and HWInfo64 on to game.

I bet I get NO RESPONSE from the 1080p whor..............wait I actually do not care what resolution I play at but prefer 4K and with the new prices of RTX 4090 and possible AMD Radeon ,I must just buy a phone to play games L/K
:D

 
I confess I am a resolution racist.I post 4K PC gaming videos online and I always get asked questions like this especially every time new PC gear is released.I hate 1080p


So I clapped back with how about AMD 7700X 5700Mhz all core at 1.30v and 6200Mhz ram and 2200Mhz Infinity fabric with recording straight from desktop and HWInfo64 on to game.

I bet I get NO RESPONSE from the 1080p whor..............wait I actually do not care what resolution I play at but prefer 4K and with the new prices of RTX 4090 and possible AMD Radeon ,I must just buy a phone to play games L/K
:D



I’m a 4K gamer as well so I doubt the 7700X will make a huge difference, but I’d be curious to see you run an in game benchmark at 1080p & 1440p ultra settlings. Stock vs. 5500 MHz and 5700 MHz.

There’s a shadow of the Tomb raider demo on steam you can use. You could also use the CP2077 built in benchmark if you have it. Plenty of games here:

https://xsreviews.co.uk/news/games-with-built-in-benchmarks-how-to-benchmark-your-pc/
 
I’m a 4K gamer as well so I doubt the 7700X will make a huge difference, but I’d be curious to see you run an in game benchmark at 1080p & 1440p ultra settlings. Stock vs. 5500 MHz and 5700 MHz.

There’s a shadow of the Tomb raider demo on steam you can use. You could also use the CP2077 built in benchmark if you have it. Plenty of games here:

https://xsreviews.co.uk/news/games-with-built-in-benchmarks-how-to-benchmark-your-pc/
Yeah it wont make much of a difference now, other than edge cases, but it isn’t a fixed target.
 
I confess I am a resolution racist.I post 4K PC gaming videos online and I always get asked questions like this especially every time new PC gear is released.I hate 1080p


So I clapped back with how about AMD 7700X 5700Mhz all core at 1.30v and 6200Mhz ram and 2200Mhz Infinity fabric with recording straight from desktop and HWInfo64 on to game.

I bet I get NO RESPONSE from the 1080p whor..............wait I actually do not care what resolution I play at but prefer 4K and with the new prices of RTX 4090 and possible AMD Radeon ,I must just buy a phone to play games L/K
:D


Did you test for performance regressions when adjusting the Fabric clock?
I set mine to 2200 and checked AIDA64 latency and it seemed to go up. Doesn't necessarily mean anything since windows could have been doing something in the background and the test gives a different score every time.
I know the memory controller is decoupled from the Fabric clock now, but don't know if there is some sort of divider being used that could gimp memory performance.
I was able to get mine higher than 2200 at stock voltage. I can't remember if it was 2300 or 2400, but any higher required a cmos reset.
 
Did you test for performance regressions when adjusting the Fabric clock?
I set mine to 2200 and checked AIDA64 latency and it seemed to go up. Doesn't necessarily mean anything since windows could have been doing something in the background and the test gives a different score every time.
I know the memory controller is decoupled from the Fabric clock now, but don't know if there is some sort of divider being used that could gimp memory performance.
I was able to get mine higher than 2200 at stock voltage. I can't remember if it was 2300 or 2400, but any higher required a cmos reset.
Well I have not actually done much Ram test or FCLK yet but default voltage on the 5600 Ram seems fine at 6200 and FCLK also seems fine with no reduced performance at 2200.
21617.jpg


I’m a 4K gamer as well so I doubt the 7700X will make a huge difference, but I’d be curious to see you run an in game benchmark at 1080p & 1440p ultra settlings. Stock vs. 5500 MHz and 5700 MHz.

There’s a shadow of the Tomb raider demo on steam you can use. You could also use the CP2077 built in benchmark if you have it. Plenty of games here:

https://xsreviews.co.uk/news/games-with-built-in-benchmarks-how-to-benchmark-your-pc/

I own over 1000+ games.I have all the good ones lol. I have not tested this yet on the 7700X and I am not tryiing to avoid you asking me to test ,I will test it and post results but have tested this over the years and there is no real difference in FPS per game from CPU clock speed (lets say with newer CPU CPU clock range 4800Mhz-5700Mhz on Ultra RT settings on lowest setting CPU bound clock speed makes a differance at 720/1080p).
As equinox654 said there may be a few edge cases.

This screen is with a 12900K at 1080 in cyberpunk.FPS around the same
4900Mhz P-Core E-Cores 4000 DDR4 Gear1 CL15 4000Mhz vs 5400Mhz P-Cores E-Cores 3700 Gear2 DDR4 CL17 4000Mhz
 
Well I have not actually done much Ram test or FCLK yet but default voltage on the 5600 Ram seems fine at 6200 and FCLK also seems fine with no reduced performance at 2200.
View attachment 516095




I own over 1000+ games.I have all the good ones lol. I have not tested this yet on the 7700X and I am not tryiing to avoid you asking me to test ,I will test it and post results but have tested this over the years and there is no real difference in FPS per game from CPU clock speed (lets say with newer CPU CPU clock range 4800Mhz-5700Mhz on Ultra RT settings on lowest setting CPU bound clock speed makes a differance at 720/1080p).
As equinox654 said there may be a few edge cases.

This screen is with a 12900K at 1080 in cyberpunk.FPS around the same
4900Mhz P-Core E-Cores 4000 DDR4 Gear1 CL15 4000Mhz vs 5400Mhz P-Cores E-Cores 3700 Gear2 DDR4 CL17 4000Mhz
Would you mind doing a comparison in Faster Than Light at 1080p and 4K please?
 
Well I have not actually done much Ram test or FCLK yet but default voltage on the 5600 Ram seems fine at 6200 and FCLK also seems fine with no reduced performance at 2200.
View attachment 516095

1664904514684.png

This is at 6000MHZ 36-36-36-96-132
I have seen the latency go as low as 68ns. After I posted earlier I saw a guide at Igor's Lab. It showed a few percentage improvement with 2166 over 2000 FCLK.

Earlier today I set the FLCK to 2200 and the Ram to 6200mhz with same timings and it seemed to work with similar latency to yours but it randomly rebooted after 10 mins.
Then I tried 6000mhz 34-34-34-72-128 and same thing. Hoping I can adjust some voltages. Just don't know what is safe yet on this platform or even what to raise besides SOC and VDDR.

Also have PBO curve optimizer at -12 with +200mhz on the max clock. Seeing boosts to 5.9 in lightly threaded loads.

Bus overclocking is a thing again too. You have 2 PLL generators on X670 boards. You can bump the speed on the 2nd to say 106 then do max clock negative offsets to keep it from trying to boost to the moon. I booted into windows at 6ghz. Not stable yet though.
You have to use the curve optimizer with a positive offset since you will be running at a higher speed with a lower multiplier. I think PBO scales voltage with multiplier. So, if you are not thermally or current limited you can trick the PBO algorithm to boost higher. Pretty cool for low thread tasks.
 

Attachments

  • 1664904332492.png
    1664904332492.png
    293.6 KB · Views: 0
  • 1664904488500.png
    1664904488500.png
    164.2 KB · Views: 0
This is at 6000MHZ 36-36-36-96-132
I have seen the latency go as low as 68ns. After I posted earlier I saw a guide at Igor's Lab. It showed a few percentage improvement with 2166 over 2000 FCLK.

Earlier today I set the FLCK to 2200 and the Ram to 6200mhz with same timings and it seemed to work with similar latency to yours but it randomly rebooted after 10 mins.
Then I tried 6000mhz 34-34-34-72-128 and same thing. Hoping I can adjust some voltages. Just don't know what is safe yet on this platform or even what to raise besides SOC and VDDR.

Also have PBO curve optimizer at -12 with +200mhz on the max clock. Seeing boosts to 5.9 in lightly threaded loads.

Bus overclocking is a thing again too. You have 2 PLL generators on X670 boards. You can bump the speed on the 2nd to say 106 then do max clock negative offsets to keep it from trying to boost to the moon. I booted into windows at 6ghz. Not stable yet though.
You have to use the curve optimizer with a positive offset since you will be running at a higher speed with a lower multiplier. I think PBO scales voltage with multiplier. So, if you are not thermally or current limited you can trick the PBO algorithm to boost higher. Pretty cool for low thread tasks.
Yes I watched the skatterbench video overclocking to 6Ghz ,was awesome and it makes me want to buy a 7950X.I knew before release that some 7950X would hit 6Ghz.I just picked the 7700X for PC Gaming because that is all I do on the computer. I use the BCLK 1 Mhz clock to round out the bus to even numbers since I bought the AMD 7700X.

With the FCLK like I mentioned I have not really been testing it yet but have done overnight around 12Hr test on the default setting DDR5 6200 and 2200FCLK,I have not tested any Ram timings yet but I will.Aida64 is not consistent with latency and some people put way to much into looking at the numbers.

Thing is the AMD 7700X is so good,runs great and I am satisfied with the settings I am running 100% that I am not even sure I want to do anymore testing. AM5 is just that good and in reality chasing higher CPU or lower Latency does nothing for my PC Gaming as I have tested this crap to death on these CPU's in the last couple years 2600X/3600X/3600XT/3800X/3800XT/5600X/5800X/10850K/12900K

Oh about the safe voltage thing,I would trust the AMD engineers to know how to set there CPU limits on voltage more than anyone on the internet.Plus you have a warranty if anything goes wrong. There is no lose in trying different things
 
Yes I watched the skatterbench video overclocking to 6Ghz ,was awesome and it makes me want to buy a 7950X.I knew before release that some 7950X would hit 6Ghz.I just picked the 7700X for PC Gaming because that is all I do on the computer. I use the BCLK 1 Mhz clock to round out the bus to even numbers since I bought the AMD 7700X.

With the FCLK like I mentioned I have not really been testing it yet but have done overnight around 12Hr test on the default setting DDR5 6200 and 2200FCLK,I have not tested any Ram timings yet but I will.Aida64 is not consistent with latency and some people put way to much into looking at the numbers.

Thing is the AMD 7700X is so good,runs great and I am satisfied with the settings I am running 100% that I am not even sure I want to do anymore testing. AM5 is just that good and in reality chasing higher CPU or lower Latency does nothing for my PC Gaming as I have tested this crap to death on these CPU's in the last couple years 2600X/3600X/3600XT/3800X/3800XT/5600X/5800X/10850K/12900K

Oh about the safe voltage thing,I would trust the AMD engineers to know how to set there CPU limits on voltage more than anyone on the internet.Plus you have a warranty if anything goes wrong. There is no lose in trying different things
If I wasn’t doing rendering and simulations. I would have went with the 7700x.

About the voltages. I was just concerned because the soc and vddr_io are auto set to 1.3 when turning on expo which puts them in the red. I actually dropped them to 1.25 after googling and reading about people degrading their fabric on Ryzen 5000 with 1.3 volts.
 
Yes I watched the skatterbench video overclocking to 6Ghz ,was awesome and it makes me want to buy a 7950X.I knew before release that some 7950X would hit 6Ghz.I just picked the 7700X for PC Gaming because that is all I do on the computer. I use the BCLK 1 Mhz clock to round out the bus to even numbers since I bought the AMD 7700X.

With the FCLK like I mentioned I have not really been testing it yet but have done overnight around 12Hr test on the default setting DDR5 6200 and 2200FCLK,I have not tested any Ram timings yet but I will.Aida64 is not consistent with latency and some people put way to much into looking at the numbers.

Thing is the AMD 7700X is so good,runs great and I am satisfied with the settings I am running 100% that I am not even sure I want to do anymore testing. AM5 is just that good and in reality chasing higher CPU or lower Latency does nothing for my PC Gaming as I have tested this crap to death on these CPU's in the last couple years 2600X/3600X/3600XT/3800X/3800XT/5600X/5800X/10850K/12900K

Oh about the safe voltage thing,I would trust the AMD engineers to know how to set there CPU limits on voltage more than anyone on the internet.Plus you have a warranty if anything goes wrong. There is no lose in trying different things

Funny you mention this as I’m currently testing a 2600X with an RTX 3090. Mostly testing the effect of memory latency when CPU bottlenecked. Going from 3200 CL16 to 3533 CL14 I’m seeing on AVG around a 10% increase in performance at 1080p. Obviously those numbers would dwindle at the resolution increases. Still, definitely seems worth it if you have a cheap B-Die kit.
 
Would you mind doing a comparison in Faster Than Light at 1080p and 4K please?
Sorry I do not own the game.
Funny you mention this as I’m currently testing a 2600X with an RTX 3090. Mostly testing the effect of memory latency when CPU bottlenecked. Going from 3200 CL16 to 3533 CL14 I’m seeing on AVG around a 10% increase in performance at 1080p. Obviously those numbers would dwindle at the resolution increases. Still, definitely seems worth it if you have a cheap B-Die kit.
Awesome share your results when complete.I did really like my 2600X and did the same testing years ago.I do not remember getting any difference in performance but I was only using 2600X and 1080Ti.
 
B650 boards announce today. We've already has

This still low key blows my mind. I understand the big boys wanna eek out as much performance as they possibly can for launch benchmarks and whatnot, but isn't there just as much cred amongst the enthusiast crowd for OC headroom? Why not match these seriously undervolted numbers out of the box in the name of efficiency with available stock profiles to OC up to this max performance? Or, at the very least, include these undervolted profiles as additional Eco modes? I just don't get it. There's gotta be some kind of stability issues with this kind of undervolting, right? It just doesn't make sense to me.
 
B650 boards announce today. We've already has

This still low key blows my mind. I understand the big boys wanna eek out as much performance as they possibly can for launch benchmarks and whatnot, but isn't there just as much cred amongst the enthusiast crowd for OC headroom? Why not match these seriously undervolted numbers out of the box in the name of efficiency with available stock profiles to OC up to this max performance? Or, at the very least, include these undervolted profiles as additional Eco modes? I just don't get it. There's gotta be some kind of stability issues with this kind of undervolting, right? It just doesn't make sense to me.

you actually don't even have to go this low. I think with PBO offset etc you can probably tweak it closer to stock like there was another video posted. Drop the power consumption improve temps to 70s etc and retain most of the performance.
 
Back
Top