7 ms·

Here’s a summary. Ryzen 5 9600X: 6 Zen 5 cores, 32 MB L3 cache, 65W TDP, available Aug 8 Ryzen 7 9700X: 8 Zen 5 cores, 32 MB L3 cache, 65W TDP, available Aug

Here’s a summary. Ryzen 5 9600X: 6 Zen 5 cores, 32 MB L3 cache, 65W TDP, available Aug 8 Ryzen 7 9700X: 8 Zen 5 cores, 32 MB L3 cache, 65W TDP, available Aug 8 Ryzen 9 9900X: 12 Zen 5 cores, 64 MB L3 cache, 120W TDP, available Aug 15 Ryzen 9 9950X: 16 Zen 5 cores, 64 MB L3 cache, 170W TDP, available Aug 15 All models support dual-channel DDR5, the speed is DDR5-5600 when 2 modules are installed, DDR5-3600 when 4 modules are installed. They all support ECC memory, with the caveat that it requires motherboard support.

jsheard 2y ago
> the speed is DDR5-5600 when 2 modules are installed That's the conservative JEDEC-compliant speed, but if you want to go for maximum performance they are saying to aim for DDR5-6400 this time, up from DDR5-6000 on the previous generation. https://www.techpowerup.com/324023/ddr5-6400-confirmed-as-sweetspot-speed-of-ryzen-9000-zen-5-desktop-processors https://www.techpowerup.com/324023/ddr5-6400-confirmed-as-sw... That speed is likely only attainable with 2 modules installed, but with 96GB (2x48GB) kits available now that's not much of a limitation for most people.
- bravetraveler 2y ago
  With enough time sunk into it, these numbers are more... suggestions, anyway. My system has a 7950X3D and four sticks at DDR5-6600. The module density, board trace topology, and luck-of-the-draw on IMC/SOC can matter a fair bit. Tuning voltages/timings can make up differences. Also remember the infinity fabric clocks and such being somewhat paired
  - martiuk 2y ago
    Is this with the EXPO profiles, or are they running at the JEDEC transfer rate?
    - bravetraveler 2y ago
      This is indeed with EXPO! There were great strides made with firmware revisions in the last year or so. I do have to make a point to boost the supporting IMC/SOC voltages before enabling that, or I can get a bit wedged. The profile voltages (and clocks/timings) for the sticks themselves work fine - just the board/CPU itself needs some care.
  - Numerlor 2y ago
    Is that running at 1:1 uclk/mclk?
    - bravetraveler 2y ago
      Glad you asked! I just checked and tried, no/not yet. It's at 1:2 or 2:1 or whatever makes sense, my firmware shows memclk/2 - haven't had my caffeine yet. The memory can train/the system POSTs, but the OS fails to load. With some more fiddling this may be feasible - I'm not sure. Excited to try! Additionally: the IF frequency is 2100Mhz, trying to go much higher doesn't work out. If I could get 2200 it'd factor nicely into the memory frequency too, but alas - chip isn't up to it.
      - Numerlor 2y ago
        you could try 6400 on memory synced if 6600 boots with 4 dimms which is a very good result. 6600 is into the territory of needing luck with IO die and then also tuning all the voltages. For the infinity fabric if you're trying to push it checking for performance regressions is a good idea as it delays and retransmits as a form of pseudo ecc when it detects an error. Same with memory if you're not using all custom timings as motherboards can set some very much not good timings if left on auto and sitting on the edge of stability
        bravetraveler 2y ago
        I appreciate the tips, will give it a try :) When speccing out this system/building it, I chased memory frequency as we can tell. Seeing 1:1 is potentially within reach, perhaps dropping down to 6400 is worthwhile. Thanks again, will give it a whirl! While I have chosen the EXPO profile, history has shown boards can't reliably read the (sub)timings... so I do it/apply those :P I try not to mess with those much, I don't really know the relationships between them.
        Numerlor 2y ago
        For the most part it's usually best to just go lowest that's stable, there are some "rules" online but most of them are just random things someone came up with. There are a couple JEDEC ones but the memory controller will adjust to work with those or it'd just not work. A 1:2 sync usually starts being worth it above ~7200MTs but I don't know how viable that'd be with 4 dimms, at the very least it'd need tuning the various termination impedances
  - oakpond 2y ago
    May I suggest to also run memtest if you're gonna tune these kind of things. Your computer may appear to work normally after changing the settings, but memtest can reveal unnoticeable memory errors.
    - bravetraveler 2y ago
      Good suggestion, though I have - Prime95, memtest, many long workloads :) With DDR5 and really, most modern hardware, having some more robustness (that pseudo-ECC whatever, clock stretching)... it might be worth considering performance as part of your validation
      - oakpond 2y ago
        Sounds good. I didn't mean to make a suggestion personally to you by the way. Sorry if it seemed that way. English is not my native language. My remark was more targeted towards the reader of the thread who gets curious and may start experimenting.
        bravetraveler 2y ago
        Totally fair :) I err towards defensive, no harm done
    - noverify 2y ago
      memtest is a pretty basic test and is not very reliable at triggering errors in memory pushed just past of what it is capable of. It's fine to run it as a smoke test before booting into the operating system (to make sure it won't crash right away), but after that you're gonna need something like TestMem5. I keep a Windows install just for this.
      - oakpond 2y ago
        Cool, didn't know about TestMem5 until now. Thanks!
  - cherryteastain 2y ago
    What's the first word latency/CAS latency? I think EXPO 6000CL30 kits can achieve 10ns, so if you have to run 6600CL40 or something it may be worse.
    - bravetraveler 2y ago
      It's CAS 32 - one doesn't have to go sky high latency to get high frequency, the binning is truly incredible these days. Just a boring Corsair Vengeance kit with OOTB settings as far as they're concerned. The CPU needed help
      - cherryteastain 2y ago
        Nice, which ICs does the RAM kit have? I don't think your experience is usual by any means with uclk:fclk 1:1 though.
        bravetraveler 2y ago
        Can't say I've been brave enough to tear them apart :D I took the lazy way out and just sorted by specs on the listing. I actually expressly didn't want to overclock the memory, just get something that ran nicely as-is. I'm actually not at 1:1... stable. Perhaps wires got crossed? In another post I mentioned it POSTs but that's about it :( I may be able to get something there, but can't promise or over-represent it. The memory is at 6600 but I have some divisor or whatever at memclk/2. This is super stable
        Numerlor 2y ago
        6000 at cl32 is most likely going to be Hynix A die, the cas latency could be better if they push like 1.8V into the memory but it's not a particularly important timing for performance
- rbanffy 2y ago
  Still, 4 modules should have more memory bandwidth than 2, even if the modules are themselves slower.
  - jsheard 2y ago
    That would be the case if the CPU had four memory channels, but consumer platforms only have two. When you install four modules the bandwidth is shared between each pair, so on paper you end up with the same bandwidth as if you used two modules at the same clock, but running four modules puts additional strain on the CPUs IMC and demands better motherboard signal integrity so the clock speed usually has to be reduced to get it stable.
    - rbanffy 2y ago
      Yikes!
    - Numerlor 2y ago
      AMD is also very limited by its infinity fabric, with one ccd CPUs capping out bandwidth at something around 5600MTs with memory frequency above that mostly just improving latencies
- Const-me 2y ago
  Personally I don’t want maximum performance. I want maximum reliability, and minimum amount of time spent configuring the hardware. If I were assembling a new computer tomorrow, I would probably get the 65W Ryzen 7 9700X with two 5600MT/s ECC unbuffered memory modules.
cudder 2y ago
> DDR5-5600 when 2 modules are installed, DDR5-3600 when 4 modules are installed Is this a new thing? I don't rememebr four sticks being slower than two on previous generations. Then again I haven't been paying too close attention these past few years.
- pantalaimon 2y ago
  It's a DDR5 thing - four sticks have absolutely been slower on the Zen 4 too. I've tried 4x32 GiB on a 7950X and I would get exactly that speed after several minutes (!) of POST. Anything faster would not be stable, I ended up sending back half of it and settled with 64 GiB total, POST still takes 30s.
  - redditor98654 2y ago
    I have a similar setup and with 4x32 sticks the most I have been able get it 4800 with EXPO enabled. Is that your experience too?
- Tuna-Fish 2y ago
  New with DDR5, and it's the reason no-one should consider upgrading memory by adding additional sticks on DDR5 platform -- instead, if you need more memory, you should replace your modules. The additional two slots are basically a newbie trap. Next gen will do away with this because it moves to CAMM2 and there will only be a single connector on consumer motherboards.
  - flemhans 2y ago
    I'm buying a server with 12 slots, should I go for 4x128 gb instead of 8x64 then?
    - Tepix 2y ago
      No, you want to utilize the extra memory bandwidth of the EPYC server CPUs. Use all 12 channels! The desktop CPUs only have two memory channels so you don't get a performance benefit by using four memory modules.
magicalhippo 2y ago
While these CPUs look great, I'm not digging the lack of PCIe lanes on the AM5 platform. Still, might finally be time to upgrade my desktop.
- cherryteastain 2y ago
  Are you coming from a HEDT platform? AM5 has 28 PCIe 5.0 lanes as opposed to 24 in AM4 and 20 in Intel Z790.
  - magicalhippo 2y ago
    Yeah, was hoping for a bit more. But at least now there are AM5 motherboards with three PCIe x16 slots, so looking a lot more tempting.
GordonS 2y ago
The relatively high TDPs are putting me off Ryzen 9 a bit, even if you are getting 12-16 cores.
- opencl 2y ago
  You can always just change the power limits in the BIOS settings.
- SushiHippie 2y ago
  You always have the option to limit the TDP via the BIOS with nearly no performance hit (at least according to the benchmarks I made, and I also feel no difference when using my PC) (Reposting a comment I made two weeks ago https://news.ycombinator.com/item?id=41043961 https://news.ycombinator.com/item?id=41043961) In my BIOS I can activate a "ECO Mode", which lets me decide if I want to run my 7950x on full 170W TDP, 105W TDP or 60W TDP. I benchmarked it, the difference between 170 and 105 is basically zero, and the difference to 60W is just a few percent of a performance hit, but way worth it, as it's ~0.3€/kWh over here.
  - kvemkon 2y ago
    > the difference between 170 and 105 is basically zero From tests I've read, if you would compile something or render or encode/transcode (on CPU) or similar, there is a measurable effect. It will take only little bit longer but consume noticeably less energy. The other positive effect: reduced noise from fan or even no noticeable noise at all. > 170W TDP, 105W TDP or 60W TDP I'm curious why it is not possible to adjust these values in only few watt steps.
    - SushiHippie 2y ago
      > From tests I've read, if you would compile something or render or encode/transcode (on CPU) or similar, there is a measurable effect. It will take only little bit longer but consume noticeably less energy. One of my "benchmarks" included compiling a large rust project, I don't have the exact numbers, but in tge end it was such a tiny fraction (something like 5 seconds difference, with a compilation time of 2 minutes). Which is definitely not worth the ~70% increased power consumption. > I'm curious why it is not possible to adjust these values in only few watt steps. Don't know probably to keep it simple (and maybe nor to Fry your CPU if you swap your CPU to one which is rated for e.g. 65W TDP). This "ECO Mode" has three settings: "OFF" which means max TDP (in the case of my CPU 170W) "105W", "ON" which is 65W. It also says in the description that "105W" mode only exists for CPUs with 170W TDP. Also if you look at the other CPUs from AMD you'll see that they only sell CPUs with these TDP numbers and additionally 120W (the x3d variants), that's probably why they chose theese numbers.
TacticalCoder 2y ago
> They all support ECC memory, with the caveat that it requires motherboard support. Is that different than the 7000 series? Basically: is this "ECC support all across the models" a 9000 thing or a DDR5 one? P.S: you didn't list the 9700X in your summary (the one I'm eyeing on) EDIT: ah no I see: you did list it but you made a typo and wrote 9600X twice.
- Const-me 2y ago
  Thanks, corrected the typo
nfriedly 2y ago
Typo: I think that second one should be 9700X, not 9600X again.
Tepix 2y ago
Finally the TDPs are going down instead of up! This is long overdue.
- layer8 2y ago
  TDP isn't a standardized measure, it has been keeping changing what exactly it means. I'm not saying they aren't improving now, but we'll have to see how it behaves in practice.
TacticalCoder 2y ago
> Ryzen 7 9700X: 8 Zen 5 cores, 32 MB L3 cache, 65W TDP, available Aug 8 Hmmm max TDP listed at 65W that's lower than my 7700X which listed max TDP at 105W I think. I know people have been running the 7700X with big fat fanless CPU heatsinks. So I take it that even if these TDP ratings are a bit of a lie, it's still not much and it should be possible to run a 9700X without any fan on the CPU heatsink?
112233 2y ago
Thanks. So nice to see official ECC support come back to the non-PRO parts.

jsheard 2y ago

> the speed is DDR5-5600 when 2 modules are installed That's the conservative JEDEC-compliant speed, but if you want to go for maximum performance they are saying to aim for DDR5-6400 this time, up from DDR5-6000 on the previous generation. https://www.techpowerup.com/324023/ddr5-6400-confirmed-as-sweetspot-speed-of-ryzen-9000-zen-5-desktop-processors https://www.techpowerup.com/324023/ddr5-6400-confirmed-as-sw... That speed is likely only attainable with 2 modules installed, but with 96GB (2x48GB) kits available now that's not much of a limitation for most people.

bravetraveler 2y ago
With enough time sunk into it, these numbers are more... suggestions, anyway. My system has a 7950X3D and four sticks at DDR5-6600. The module density, board trace topology, and luck-of-the-draw on IMC/SOC can matter a fair bit. Tuning voltages/timings can make up differences. Also remember the infinity fabric clocks and such being somewhat paired
- martiuk 2y ago
  Is this with the EXPO profiles, or are they running at the JEDEC transfer rate?
  - bravetraveler 2y ago
    This is indeed with EXPO! There were great strides made with firmware revisions in the last year or so. I do have to make a point to boost the supporting IMC/SOC voltages before enabling that, or I can get a bit wedged. The profile voltages (and clocks/timings) for the sticks themselves work fine - just the board/CPU itself needs some care.
- Numerlor 2y ago
  Is that running at 1:1 uclk/mclk?
  - bravetraveler 2y ago
    Glad you asked! I just checked and tried, no/not yet. It's at 1:2 or 2:1 or whatever makes sense, my firmware shows memclk/2 - haven't had my caffeine yet. The memory can train/the system POSTs, but the OS fails to load. With some more fiddling this may be feasible - I'm not sure. Excited to try! Additionally: the IF frequency is 2100Mhz, trying to go much higher doesn't work out. If I could get 2200 it'd factor nicely into the memory frequency too, but alas - chip isn't up to it.
    - Numerlor 2y ago
      you could try 6400 on memory synced if 6600 boots with 4 dimms which is a very good result. 6600 is into the territory of needing luck with IO die and then also tuning all the voltages. For the infinity fabric if you're trying to push it checking for performance regressions is a good idea as it delays and retransmits as a form of pseudo ecc when it detects an error. Same with memory if you're not using all custom timings as motherboards can set some very much not good timings if left on auto and sitting on the edge of stability
      - bravetraveler 2y ago
        I appreciate the tips, will give it a try :) When speccing out this system/building it, I chased memory frequency as we can tell. Seeing 1:1 is potentially within reach, perhaps dropping down to 6400 is worthwhile. Thanks again, will give it a whirl! While I have chosen the EXPO profile, history has shown boards can't reliably read the (sub)timings... so I do it/apply those :P I try not to mess with those much, I don't really know the relationships between them.
        Numerlor 2y ago
        For the most part it's usually best to just go lowest that's stable, there are some "rules" online but most of them are just random things someone came up with. There are a couple JEDEC ones but the memory controller will adjust to work with those or it'd just not work. A 1:2 sync usually starts being worth it above ~7200MTs but I don't know how viable that'd be with 4 dimms, at the very least it'd need tuning the various termination impedances
- oakpond 2y ago
  May I suggest to also run memtest if you're gonna tune these kind of things. Your computer may appear to work normally after changing the settings, but memtest can reveal unnoticeable memory errors.
  - bravetraveler 2y ago
    Good suggestion, though I have - Prime95, memtest, many long workloads :) With DDR5 and really, most modern hardware, having some more robustness (that pseudo-ECC whatever, clock stretching)... it might be worth considering performance as part of your validation
    - oakpond 2y ago
      Sounds good. I didn't mean to make a suggestion personally to you by the way. Sorry if it seemed that way. English is not my native language. My remark was more targeted towards the reader of the thread who gets curious and may start experimenting.
      - bravetraveler 2y ago
        Totally fair :) I err towards defensive, no harm done
  - noverify 2y ago
    memtest is a pretty basic test and is not very reliable at triggering errors in memory pushed just past of what it is capable of. It's fine to run it as a smoke test before booting into the operating system (to make sure it won't crash right away), but after that you're gonna need something like TestMem5. I keep a Windows install just for this.
    - oakpond 2y ago
      Cool, didn't know about TestMem5 until now. Thanks!
- cherryteastain 2y ago
  What's the first word latency/CAS latency? I think EXPO 6000CL30 kits can achieve 10ns, so if you have to run 6600CL40 or something it may be worse.
  - bravetraveler 2y ago
    It's CAS 32 - one doesn't have to go sky high latency to get high frequency, the binning is truly incredible these days. Just a boring Corsair Vengeance kit with OOTB settings as far as they're concerned. The CPU needed help
    - cherryteastain 2y ago
      Nice, which ICs does the RAM kit have? I don't think your experience is usual by any means with uclk:fclk 1:1 though.
      - bravetraveler 2y ago
        Can't say I've been brave enough to tear them apart :D I took the lazy way out and just sorted by specs on the listing. I actually expressly didn't want to overclock the memory, just get something that ran nicely as-is. I'm actually not at 1:1... stable. Perhaps wires got crossed? In another post I mentioned it POSTs but that's about it :( I may be able to get something there, but can't promise or over-represent it. The memory is at 6600 but I have some divisor or whatever at memclk/2. This is super stable
      - Numerlor 2y ago
        6000 at cl32 is most likely going to be Hynix A die, the cas latency could be better if they push like 1.8V into the memory but it's not a particularly important timing for performance
rbanffy 2y ago
Still, 4 modules should have more memory bandwidth than 2, even if the modules are themselves slower.
- jsheard 2y ago
  That would be the case if the CPU had four memory channels, but consumer platforms only have two. When you install four modules the bandwidth is shared between each pair, so on paper you end up with the same bandwidth as if you used two modules at the same clock, but running four modules puts additional strain on the CPUs IMC and demands better motherboard signal integrity so the clock speed usually has to be reduced to get it stable.
  - rbanffy 2y ago
    Yikes!
  - Numerlor 2y ago
    AMD is also very limited by its infinity fabric, with one ccd CPUs capping out bandwidth at something around 5600MTs with memory frequency above that mostly just improving latencies
Const-me 2y ago
Personally I don’t want maximum performance. I want maximum reliability, and minimum amount of time spent configuring the hardware. If I were assembling a new computer tomorrow, I would probably get the 65W Ryzen 7 9700X with two 5600MT/s ECC unbuffered memory modules.

cudder 2y ago

> DDR5-5600 when 2 modules are installed, DDR5-3600 when 4 modules are installed Is this a new thing? I don't rememebr four sticks being slower than two on previous generations. Then again I haven't been paying too close attention these past few years.

pantalaimon 2y ago
It's a DDR5 thing - four sticks have absolutely been slower on the Zen 4 too. I've tried 4x32 GiB on a 7950X and I would get exactly that speed after several minutes (!) of POST. Anything faster would not be stable, I ended up sending back half of it and settled with 64 GiB total, POST still takes 30s.
- redditor98654 2y ago
  I have a similar setup and with 4x32 sticks the most I have been able get it 4800 with EXPO enabled. Is that your experience too?
Tuna-Fish 2y ago
New with DDR5, and it's the reason no-one should consider upgrading memory by adding additional sticks on DDR5 platform -- instead, if you need more memory, you should replace your modules. The additional two slots are basically a newbie trap. Next gen will do away with this because it moves to CAMM2 and there will only be a single connector on consumer motherboards.
- flemhans 2y ago
  I'm buying a server with 12 slots, should I go for 4x128 gb instead of 8x64 then?
  - Tepix 2y ago
    No, you want to utilize the extra memory bandwidth of the EPYC server CPUs. Use all 12 channels! The desktop CPUs only have two memory channels so you don't get a performance benefit by using four memory modules.

magicalhippo 2y ago

While these CPUs look great, I'm not digging the lack of PCIe lanes on the AM5 platform. Still, might finally be time to upgrade my desktop.

cherryteastain 2y ago
Are you coming from a HEDT platform? AM5 has 28 PCIe 5.0 lanes as opposed to 24 in AM4 and 20 in Intel Z790.
- magicalhippo 2y ago
  Yeah, was hoping for a bit more. But at least now there are AM5 motherboards with three PCIe x16 slots, so looking a lot more tempting.

GordonS 2y ago

The relatively high TDPs are putting me off Ryzen 9 a bit, even if you are getting 12-16 cores.

opencl 2y ago
You can always just change the power limits in the BIOS settings.
SushiHippie 2y ago
You always have the option to limit the TDP via the BIOS with nearly no performance hit (at least according to the benchmarks I made, and I also feel no difference when using my PC) (Reposting a comment I made two weeks ago https://news.ycombinator.com/item?id=41043961 https://news.ycombinator.com/item?id=41043961) In my BIOS I can activate a "ECO Mode", which lets me decide if I want to run my 7950x on full 170W TDP, 105W TDP or 60W TDP. I benchmarked it, the difference between 170 and 105 is basically zero, and the difference to 60W is just a few percent of a performance hit, but way worth it, as it's ~0.3€/kWh over here.
- kvemkon 2y ago
  > the difference between 170 and 105 is basically zero From tests I've read, if you would compile something or render or encode/transcode (on CPU) or similar, there is a measurable effect. It will take only little bit longer but consume noticeably less energy. The other positive effect: reduced noise from fan or even no noticeable noise at all. > 170W TDP, 105W TDP or 60W TDP I'm curious why it is not possible to adjust these values in only few watt steps.
  - SushiHippie 2y ago
    > From tests I've read, if you would compile something or render or encode/transcode (on CPU) or similar, there is a measurable effect. It will take only little bit longer but consume noticeably less energy. One of my "benchmarks" included compiling a large rust project, I don't have the exact numbers, but in tge end it was such a tiny fraction (something like 5 seconds difference, with a compilation time of 2 minutes). Which is definitely not worth the ~70% increased power consumption. > I'm curious why it is not possible to adjust these values in only few watt steps. Don't know probably to keep it simple (and maybe nor to Fry your CPU if you swap your CPU to one which is rated for e.g. 65W TDP). This "ECO Mode" has three settings: "OFF" which means max TDP (in the case of my CPU 170W) "105W", "ON" which is 65W. It also says in the description that "105W" mode only exists for CPUs with 170W TDP. Also if you look at the other CPUs from AMD you'll see that they only sell CPUs with these TDP numbers and additionally 120W (the x3d variants), that's probably why they chose theese numbers.

TacticalCoder 2y ago

> They all support ECC memory, with the caveat that it requires motherboard support. Is that different than the 7000 series? Basically: is this "ECC support all across the models" a 9000 thing or a DDR5 one? P.S: you didn't list the 9700X in your summary (the one I'm eyeing on) EDIT: ah no I see: you did list it but you made a typo and wrote 9600X twice.

Const-me 2y ago
Thanks, corrected the typo

nfriedly 2y ago

Typo: I think that second one should be 9700X, not 9600X again.

Tepix 2y ago

Finally the TDPs are going down instead of up! This is long overdue.

layer8 2y ago
TDP isn't a standardized measure, it has been keeping changing what exactly it means. I'm not saying they aren't improving now, but we'll have to see how it behaves in practice.

TacticalCoder 2y ago

> Ryzen 7 9700X: 8 Zen 5 cores, 32 MB L3 cache, 65W TDP, available Aug 8 Hmmm max TDP listed at 65W that's lower than my 7700X which listed max TDP at 105W I think. I know people have been running the 7700X with big fat fanless CPU heatsinks. So I take it that even if these TDP ratings are a bit of a lie, it's still not much and it should be possible to run a 9700X without any fan on the CPU heatsink?

112233 2y ago

Thanks. So nice to see official ECC support come back to the non-PRO parts.