AMD Ryzen 9 7950X3D Memory Scaling: Does DDR5 Performance Matter?

Thx for this article Steve. Another conclusion which is worth mentioning is that 7950X3D paired with DDR5-6000 CL30 performance, against 13900K, F, S etc is even higher, so this round AMD 7950X3D is the clear winner against anything Intel has on the market in gaming, productivity and efficiency. Though is 'loosing' on price :)
 
Does anyone know if an x670e mainboard that is certified for higher RAM clocks (up to 6400) would actually benefit from those 400 extra MHz over the 6000 RAM tested here? Currently looking at one of the higher end ASUS ones...
 
Yep, it behaves EXACTLY the same way as my R7-5800X3D. Before I bought it, I looked around at different tests with regard to RAM speed and gaming performance with it because my RAM is just DDR4-3200. I never cared much about buying the fastest RAM possible because in 99% of cases, it's just not worth the (sometimes MUCH) higher price (hell, I used 16GB of DDR3-1333 for my FX-8350 and it was just fine). When I bought new RAM for my X570 Pro4, DDR4-3200 offered the most MHz per dollar so that's what I got. Ironically, DDR4-3200 also happens to be the sweet spot for the R7-5800X3D.

I've never noticed a difference in my PC's performance between my DDR4-3200 kit and the DDR4-2400 kit that I used with my X370 Killer SLI. I don't even bother enabling XMP because even when I did, it made no difference that I could feel. Therefore, I just let Windows run the RAM at 2400MHz instead of 3200MHz because it's less power draw and probably easier on the silicon. If I ever encounter a situation where a game's performance is suffering because of RAM speed (who knows, could happen, eh?), then I'll enable XMP again.

When I bought my R7-5800X3D, I thought to myself "Hmm, since I'm not buying a new motherboard or DDR5 maybe it would be a good time to up my RAM to 32GB from 16GB." and started looking around. I found an open-box 16GB AData XPG GAMMIX D20 kit for $40 at Canada Computers. The timings all looked correct to match my existing Team Vulcan kit so I grabbed it.

FUNNY THING:
After hearing god-knows how many (well-meaning) people giving me "Dire warnings about using two brands of RAM <scary ghost moaning>" which I completely ignored (I buy by spec, not by brand), I installed the RAM. The PC posted, Windows loaded and Task Manager showed 32GB at 2400MHz. I restarted the PC, entered BIOS and enabled XMP. Loaded Windows and brought up Task Manager again. This time it said 32GB at 3200MHz so I ran Passmark to test the RAM stability. Passmark remained perfectly stable so I went back into BIOS and turned XMP back to auto (2400MHz).

That's why I only buy by spec. If the numbers all match, that's science. The brand is just marketing BS that often costs money for nothing. Remember that ALL of the RAM in the world is made by Samsung, Hynix or Micron no matter whose "brand" is on it. :laughing:
 
Does anyone know if an x670e mainboard that is certified for higher RAM clocks (up to 6400) would actually benefit from those 400 extra MHz over the 6000 RAM tested here? Currently looking at one of the higher end ASUS ones...
The problem of the Ryzen is the slow L3 , maybe on purpose ? I have 3200 Mhz DDR4 and when I put it at 3800 Cl 15-15-15-30-45 , system is slower on benchmarks than DDR4 at 3000 Mhz Cl 14 or 15 . In game I wouldnt says the same thing , but improvements with high clocked ram is not sure for me . One time I had a good improvment with DDR 4 at 3600 . Infinityfabric runs at 3200 Mhz at normal speed ( clock at 1600 Mhz ) . I think windows 10 has a rôle in various perfs because I usually never had same perfs at a defined clock ram .
 
Yep, it behaves EXACTLY the same way as my R7-5800X3D. Before I bought it, I looked around at different tests with regard to RAM speed and gaming performance with it because my RAM is just DDR4-3200. I never cared much about buying the fastest RAM possible because in 99% of cases, it's just not worth the (sometimes MUCH) higher price (hell, I used 16GB of DDR3-1333 for my FX-8350 and it was just fine). When I bought new RAM for my X570 Pro4, DDR4-3200 offered the most MHz per dollar so that's what I got. Ironically, DDR4-3200 also happens to be the sweet spot for the R7-5800X3D.

I've never noticed a difference in my PC's performance between my DDR4-3200 kit and the DDR4-2400 kit that I used with my X370 Killer SLI. I don't even bother enabling XMP because even when I did, it made no difference that I could feel. Therefore, I just let Windows run the RAM at 2400MHz instead of 3200MHz because it's less power draw and probably easier on the silicon. If I ever encounter a situation where a game's performance is suffering because of RAM speed (who knows, could happen, eh?), then I'll enable XMP again.

When I bought my R7-5800X3D, I thought to myself "Hmm, since I'm not buying a new motherboard or DDR5 maybe it would be a good time to up my RAM to 32GB from 16GB." and started looking around. I found an open-box 16GB AData XPG GAMMIX D20 kit for $40 at Canada Computers. The timings all looked correct to match my existing Team Vulcan kit so I grabbed it.

FUNNY THING:
After hearing god-knows how many (well-meaning) people giving me "Dire warnings about using two brands of RAM <scary ghost moaning>" which I completely ignored (I buy by spec, not by brand), I installed the RAM. The PC posted, Windows loaded and Task Manager showed 32GB at 2400MHz. I restarted the PC, entered BIOS and enabled XMP. Loaded Windows and brought up Task Manager again. This time it said 32GB at 3200MHz so I ran Passmark to test the RAM stability. Passmark remained perfectly stable so I went back into BIOS and turned XMP back to auto (2400MHz).

That's why I only buy by spec. If the numbers all match, that's science. The brand is just marketing BS that often costs money for nothing. Remember that ALL of the RAM in the world is made by Samsung, Hynix or Micron no matter whose "brand" is on it. :laughing:
intel xmp suxx a lot , even if ram brand are Samsung, Hynix or Micron .
 
intel xmp suxx a lot , even if ram brand are Samsung, Hynix or Micron .
Yeah, that's what I thought. My system isn't Intel but for some reason my BIOS has an XMP setting. In any case, I forced my RAM to run at its rated speed to increase the chance of there being an instability issue during a Passmark run but everything was fine.

That was the point that I think you may have missed.
 
Just bought a new and system waiting for the parts. With a preorder for the new 7950x3d cpu. Hoping to have everything so I can finalize the build at the end of the month. Will have 128gb ram using 4x 32gb skill with the 30 timings. Haven't bought a new PC in 10 years as I converted to laptops. Old PC is really slow and too slow to bare. Exciting times. Was also looking at the new and laptop CPU due out around mid year 7945 I think is the CPU number? The dragon range. Keeping an eye on which laptop will give me the best bang for buck.
 
Is it possible that either Ryzen would have benefitted from the DDR5-7200 ram? Or even faster?

Not so much as the Infinity fabric tops out at 3000MHz, and using faster memory would mean an asynchronous IF memory ratio. 1:1 is ideal. So AMD is stuck with DDR5 6000. Maybe down the track some amazing DDR5 8000+ with CL24 or something might show some benefits even if run asynchronously, but it would monstrously expensive.
 
The problem of the Ryzen is the slow L3 , maybe on purpose ? I have 3200 Mhz DDR4 and when I put it at 3800 Cl 15-15-15-30-45 , system is slower on benchmarks than DDR4 at 3000 Mhz Cl 14 or 15 . In game I wouldnt says the same thing , but improvements with high clocked ram is not sure for me . One time I had a good improvment with DDR 4 at 3600 . Infinityfabric runs at 3200 Mhz at normal speed ( clock at 1600 Mhz ) . I think windows 10 has a rôle in various perfs because I usually never had same perfs at a defined clock ram .
If you're not running it at 1:1 you'll pay for it in a quite considerable latency uplift
so ram 1800MHz/3600MT/s and fabric 1800MHz shouldn't suffer the latency hit but
ram 1800MHz/3600MT/s and a fabric speed of 1600MHz takes a fairly sizable latency hit
no matter what the timings on your ram are
 
AMD pulled the same strategy in terms of pricing what comes $100 apart at launch and limited in units compared to the cheaper variant that no one wants?
 
People said that the 7950X3D isn't needed.

After seeing the reviews is it?

I guess if you have a top tier gpu and run 1080p or 1440p you will be happy. But for me, as a productivity cpu, the drop in performance across the board is in no way worth it ffor a few cache sensitive games where the 7950X was already fast enough. And you can run FSR or DLSS if you want to enormously boost fps.

Even with v-cache the 7950X still trails the 13900/700K in 1% lows in most games.

To me the only thing good about the X3D model is that with PBO + undervolting you can basically match the 7950X in productivity while using a lot less power. But not worth the $100+ price premium. 7900X3D just reviewed by Gamers Nexus makes even less sense. Then you have the lousy scheduler issues and have to rely on Xbox game bar.

7800X3D will probably be ok, but it depends how much they have gimped the clocks. At least 7950X3D gas two ccd's and one runs at normal speed, 7800X3D will be pernanently gimped just to boost gaming.
 
After seeing the reviews is it?

I guess if you have a top tier gpu and run 1080p or 1440p you will be happy. But for me, as a productivity cpu, the drop in performance across the board is in no way worth it ffor a few cache sensitive games where the 7950X was already fast enough. And you can run FSR or DLSS if you want to enormously boost fps.

Even with v-cache the 7950X still trails the 13900/700K in 1% lows in most games.

To me the only thing good about the X3D model is that with PBO + undervolting you can basically match the 7950X in productivity while using a lot less power. But not worth the $100+ price premium. 7900X3D just reviewed by Gamers Nexus makes even less sense. Then you have the lousy scheduler issues and have to rely on Xbox game bar.

7800X3D will probably be ok, but it depends how much they have gimped the clocks. At least 7950X3D gas two ccd's and one runs at normal speed, 7800X3D will be pernanently gimped just to boost gaming.
Based on the microcenter stock it seems no one wants the 7900x3d but the 7950X3D are sold out on it's initial run. Reminds me of the 7900xtx vs 7900xt launch.
 
After seeing the reviews is it?

I guess if you have a top tier gpu and run 1080p or 1440p you will be happy. But for me, as a productivity cpu, the drop in performance across the board is in no way worth it ffor a few cache sensitive games where the 7950X was already fast enough. And you can run FSR or DLSS if you want to enormously boost fps.

Even with v-cache the 7950X still trails the 13900/700K in 1% lows in most games.

To me the only thing good about the X3D model is that with PBO + undervolting you can basically match the 7950X in productivity while using a lot less power. But not worth the $100+ price premium. 7900X3D just reviewed by Gamers Nexus makes even less sense. Then you have the lousy scheduler issues and have to rely on Xbox game bar.

7800X3D will probably be ok, but it depends how much they have gimped the clocks. At least 7950X3D gas two ccd's and one runs at normal speed, 7800X3D will be pernanently gimped just to boost gaming.

Seems like both the 16 and 12 core X3D parts are sold out everywhere. These chips aren't for a lot of us.
 
Seems like both the 16 and 12 core X3D parts are sold out everywhere. These chips aren't for a lot of us.
Interesting because AMD's website still has the 7900x3d in stock and all Microcenter have dozens of them in stock as well. Even Amazon has them but at scalped pricing. Although Bestbuy says it sold out.
 
Interesting because AMD's website still has the 7900x3d in stock and all Microcenter have dozens of them in stock as well. Even Amazon has them but at scalped pricing. Although Bestbuy says it sold out.
Those that are wanting and can pay high end, they want the highest end. That's why the rtx 4090 is selling well, the 4080 not. Though makes no sense to spend so much money for games.
 
Seems like both the 16 and 12 core X3D parts are sold out everywhere. These chips aren't for a lot of us.

Hype due to the 5800X3D. I doubt the 7950X or X3D are for most people, especially gamers, but probably the only ones buying them.
 
Is it possible that either Ryzen would have benefitted from the DDR5-7200 ram? Or even faster?
If one ignores the IF issues mentioned by @daffy duck then one can extrapolate where things might go from Steve's benchmark results. For the 7700X, the jump from DDR5-5200 to DDR5-6000, a 15% increase in bandwidth, resulted in the average game performance increasing by 7%; for the 7950X3D, the gain was 2%.

So if one assumes that these ratios continue for higher RAM speeds, then using DDR5-7200 would potentially increase the 7700X's gaming performance by 17% and the 7950X3D's by 5%. The 13900K's actual gains were a fraction under 6%, which lends some credence to such an extrapolation.
 
Hype due to the 5800X3D. I doubt the 7950X or X3D are for most people, especially gamers, but probably the only ones buying them.
What, I expect 7800X3D to be the go-to gaming chip for people going AM5 soon. Best gaming perf.
7600X/7700X will be the value choice.
 
Hype due to the 5800X3D. I doubt the 7950X or X3D are for most people, especially gamers, but probably the only ones buying them.
If you have heavy workloads and like efficiency they are damn nice. Granted I'd jump on a 7600X3D (which doesn't exist) without thinking twice. I don't need a space heater I want an efficient gaming CPU. Heck of a lot more appealing to me than a 13900K.

Are the DDR5 6000 CL30 and DDR5 CL32 widely available or are they like some of the low cas DDR4 modules that were available early on then skyrocketed in price later?
 
Back