AMD Announces New AGESA 1.2.0.2, 105W cTDP for 9700X and 9600X, Intercore Latency Improvements

[Kirederf]

Level 1 said that the Memory support has improved with X870E but we don't have reviews yet. That is good what are your timings?

I'm running a Corsair Dominator kit 2x24GB 8000C38. This one.

That's because X670(E) uses the same chipset silicon as X870(E) - the Promontory 21 (+ Promontory 21).

Yes I know; I'm fully aware of that. I was just curious if my board was able to do it. Previously it did not. 8000MT/s was promised before but it never actually ran great.

 

[izy]

What is the difference between enabling PBO and increasing PPT and this new setting for 105W? Does it do something PBO couldn't?

 

[LittleBro]

I guess the PBO is a bit more sophisticated and a bit risky for this update.

Today's chips tend to boost as much as possible until some limit is met. This works for my Zen 3: when I raise PPT limit without touching anything else, the only difference is that few other cores boost a few hundred MHz more until the CPU package hits the newly set TDP limit or the EDP limit (whatever comes first). I've never tried this myself but maybe if you set 200W TDP limit for 65W CPU, it will try to boost all cores to reach max. boost clock limit and still it may never really reach 200W limit.

With PBO you also change voltage curve and max. boost clock. PBO is not guaranteed and can render your CPU unstable. This could mean another potential fiasco for AMD so I think they will skip using PBO for this update.

 

[Geofrancis]

If you're referring to X3D cache, then you're wrong. X3D cache is exposed to CPU cores in the same manner as regular L3 cache. In other words, 3D cache is in fact regular L3 cache but stacked vertically (three times) rather than being placed planarly.

no im talking about the intercore latency improvemtns that have nothing to do with vcache,

 

[Chrispy_]

Only workloads that resulted in relatively low clock speeds at the 65 W TDP saw significant performance increases with PBO. In TPU's review, Blender was the standout with a 15% increase at the cost of more than doubled power consumption. Given that result, the impact of the 105 W setting is unlikely to be significant enough to warrant the higher power consumption.

Well, I guess it's there as a preset option now

65W is efficient, 105W is crazily inefficient but 15% faster is faster if that specific workload matters to you - and as always a manual tune with PBO+ and some curve optimiser will likely get those who are willing to tinker the absolute best of both worlds.

 

[phints]

Well, I guess it's there as a preset option now

65W is efficient, 105W is crazily inefficient but 15% faster is faster if that specific workload matters to you - and as always a manual tune with PBO+ and some curve optimiser will likely get those who are willing to tinker the absolute best of both worlds.

Agree. 105W TDP + PBO + Curve Optimizer -15 will get you about +10% performance over stock 65W with roughly the same power consumption/thermals as stock. I do something similar even with my 4 year old 5800X, I use PBO + CO -20 and no wattage increase over stock.

 

[jpvalverde85]

I tried the new FW3.08 AGESA 1.2.0.2 on my ASRock B650I Lighting, but it has a bug which reduced my rear USB-C 10Gbps port to 480Mbps (USB 2.0) speeds. I flashed it back to FW2.10 AGESA 1.1.0.3 and my full port speed returned. There are a few firmware versions in between, but I didn't bother to see if they had the bug or not. I'll stay on FW2.10 for now.

This... I assembled just yesterday an R7 7700 over an Asus Prime B650M-A WIFI, first thing done updating BIOS, loaded the beta one with AGESA 1.2.0.2 and found the same with some USB 3.0 ports, speed degraded to USB 2.0 spec. I can't believe that AM5 still needs so much ironing out after 2 years on the market.

 

[Kapone33]

AMD tries to play marketing here, to play us to buy new boards - to think that we get support for something new only with newer boards. Don't get fooled.
8000 MT/s on Zen 4/5 is only a very little (meaning negligible) bit faster than 6000 MT/s. That's because of P'o's memory controller unable being to run in 1:1 mode beyond 6400 MHz.
Only with 7800 MT/s and more you're actually starting to overcome latency penalties caused by running in out of synch mode (2:1).

Yes it is marketing but still based on the IMC on the 9000 series chips. Just like when we went from 2933 to 3200 on 1700 to 2700 and then 3600 on 3300x but 5000 had to wait until the build quality of some X570 and X570S boards to support 4000. Unless we see some innovation in the IMC it will not improve. I agree with what you said about X3D. Don't forget about Smart Access Memory. Combine that with X3D and you are laughing. I really wonder how many people have 7900X3D (same Cache as 7950X3D) with 7900XT (20 GB VRAM buffer) then 6000 30 timings RAM (Team) and laugh at people that tell me Gaming at 4K on high is unachievable and I am lying about enjoying the performance of my PC. Now the slowest drive in my system is an SSD but it has 8 TB of storage. SSD RAID 0 is still my favourite form of RAID.

This... I assembled just yesterday an R7 7700 over an Asus Prime B650M-A WIFI, first thing done updating BIOS, loaded the beta one with AGESA 1.2.0.2 and found the same with some USB 3.0 ports, speed degraded to USB 2.0 spec. I can't believe that AM5 still needs so much ironing out after 2 years on the market.

This is true but you did not get a B850 board either. On AM4 it was not until X470/B450 that rock solid performance and reliability were realized. Over time all boards will get the same support but there has to be an incentive for you to buy. Unfortunately I think AMD has been watching too much Youtube and missed the mark on USB 4.

 

[PenguinBelly]

Agree. 105W TDP + PBO + Curve Optimizer -15 will get you about +10% performance over stock 65W with roughly the same power consumption/thermals as stock. I do something similar even with my 4 year old 5800X, I use PBO + CO -20 and no wattage increase over stock.

105W cTDP = 140W PPT
65W cTDP = 88W PPT

Whichever way you slice, they are not "roughly same" power consumption/thermals.