AM5 AGESA 1.0.0.7b

[ir_cow]

Made some more progress tonight. I can confidently say DDR5-8000 is no go for the ASRock X670E Taichi and my CPU for the voltages I feel safe with.

DDR5-8000 SoC 1.3 / VDDIO-MEM 1.45V - Fails y-cruncher 2.5b

DDR5-7600 SoC 1.275v / VDDIO-MEM 1.35V - pass y-cruncher 2.5b - fails memtest.

DDr5-7200 SoC 1.2v / VDDIO-MEM 1.35V. - passes y-cruncher 2.5b - passes memtest.

I can probably get 7600 stable, but I don't like the idea of VDDIO-MEM 1.4+ for long term use.

Time to try out the GB X670e Master

 

[tabascosauz]

Made some more progress tonight. I can confidently say DDR5-8000 is no go for the ASRock X670E Taichi and my CPU for the voltages I feel safe with.

DDR5-8000 SoC 1.3 / VDDIO-MEM 1.45V - Fails y-cruncher 2.5b

DDR5-7600 SoC 1.275v / VDDIO-MEM 1.35V - pass y-cruncher 2.5b - fails memtest.

DDr5-7200 SoC 1.2v / VDDIO-MEM 1.35V. - passes y-cruncher 2.5b - passes memtest.

I can probably get 7600 stable, but I don't like the idea of VDDIO-MEM 1.4+ for long term use.

Time to try out the GB X670e Master

Is VDDIO MEM just VDIMM? If so, seems like more a question of whether Hynix A likes it long-term or not.

afaik don't think bz's 8000 profile was stable either? Not yet at least. 7200 seems similar to Intel (well obviously Intel retains all the performance cause it's still Gear 2)

 

[ir_cow]

VDDIO-MEM is the IMC voltage rail. Like VDD2 for Intel.

The memory im using is binned to 8000. Work up to 8400 on Intel.

 

[AusWolf]

I skipped 1.0.0.7a and still on 1.0.0.7. and it works absolutely perfectly. Wondering if I should update to 1.0.0.7b.
ASUS claims "Improve system stability" so maybe there are some good behind the scenes changes.

If it works perfectly, I'd leave it as it is. I'll stay on 1.0.0.7 as well.

 

[tabascosauz]

VDDIO-MEM is the IMC voltage rail. Like VDD2 for Intel.

I can't figure out which is which lol. I know there's a UMC rail but not the name. VDDIO MC? VDDDIO MEM? VDDIO MEM S3?

Thought it was VDIMM because Ryzen Master refers to it as MEM VDDIO. Unhelpful as always, AMD.

How far can you get the kit stable on Intel?

 

[ir_cow]

I can't figure out which is which lol. I know there's a UMC rail but not the name. VDDIO MC? VDDDIO MEM? VDDIO MEM S3?

All three depending on the motherboard. Technically ASRock is only one correct in the naming. ASUS just calls it memory controller voltage. Typical ASUS lol. It's fine if you already know the real name for it, but doesn't help for useful information between motherboards without knowing it already.

How far can you get the kit stable on Intel?

8400 passes y-cruncher, but only can run it from a cold boot. So 8200 is my 24/7 limit for any kit with the 13900K CPU I have right now. Maybe then next z790 /14th gen refresh will do better. The CPu voltages are my limit right now. I like sticking with long-term stable / safe voltages for mem controller and not just a benchmark run.

 

[blackie]

Checking Asus website for my B650E-F and still no 1007B available...
Could you please share which Asus model do you have with already updated AGESA bios? thnx

I skipped 1.0.0.7a and still on 1.0.0.7. and it works absolutely perfectly. Wondering if I should update to 1.0.0.7b.
ASUS claims "Improve system stability" so maybe there are some good behind the scenes changes.

 

[TheLostSwede]

Gigabyte is currently still doing the last of the verifications and should have the first updates out by sometime next week from what I was told.

This is obviously on the 1.0.0.7a AGESA, but I just popped the memory in and enabled XMP (no EXPO profile), suggesting that there's a huge difference between boards.
In fact, I'll admit I was surprised at how easy it was.

 

[Chry]

Checking Asus website for my B650E-F and still no 1007B available...
Could you please share which Asus model do you have with already updated AGESA bios? thnx

Heya, my B650 also doesn't have an update yet. Was just speaking in general and in anticipation.

 

[chrcoluk]

We will see. I am doing some A/B comparison tonight after work. But its not 2/3 IMC, it is 1/2 (Gear 2 -2:1). Just like AM4 and DDR4, only when you reached DDR4-5066 did it start to beat match 3600 CL14 once you pasted Gear 1. So if the same applies, we are looking at DDR5-8400 before it beats / matches DDR5-6000 (Gear 1).

Is there any speed/timing combo on DDR4 in gear 2 that outperforms 3200CL14 gear 1 you tested?

 

[ir_cow]

Is there any speed/timing combo on DDR4 in gear 2 that outperforms 3200CL14 gear 1 you tested?

If its looser like 3200 CL18, 5066 either ties or beats it. https://www.techpowerup.com/review/acer-predator-vesta-rgb-ddr4-3600-2x-16-gb/6.html

 

[chrcoluk]

Gigabyte is currently still doing the last of the verifications and should have the first updates out by sometime next week from what I was told.

This is obviously on the 1.0.0.7a AGESA, but I just popped the memory in and enabled XMP (no EXPO profile), suggesting that there's a huge difference between boards.
In fact, I'll admit I was surprised at how easy it was.

Something I have always suspected board variance (IMC also on CPU side). Ram speeds remain one of those variables that brings a new level to silicon lottery.

If its looser like 3200 CL18, 5066 either ties or beats it. https://www.techpowerup.com/review/acer-predator-vesta-rgb-ddr4-3600-2x-16-gb/6.html

Ok sticking to my gear 1 then. Thanks.

 

[Kapone33]

So I went on AMD and downloaded the latest chipset. Wow, my CPU now only goes to 5.6 GHz from 5.7 but now the voltage is locked at 1.25 max on my 7900X3D vs 1.35. Now there are no 88+ C spikes in Gaming. AMD has me hooked just as much as Disney as Secret Invasion this week was insane for Marvel fans (Avengers Annual 10 is the greatest single Marvel comic). I can't wait to get my hands on one of those APUs with RDNA Graphics. If they can sing at 35 Watts imagine 65 Watts or the best thing about AMD. If you apply aggressive cooling it will push (The processor) itself to give you that satisfying smile that some PC upgrades can bring. How many of us loved the jump from HDD to SSD?

 

[ir_cow]

Just waiting for everyone else to release patch B So far publicly its only ASRock. I guess in the meantime I will use the Taichi for DDR5-7200 reviews. Dang. AMD just created more work for me lol.

Update: Did some A/B benchmarks with a RTX 4090. There is a good dip once you switch to Gear 2. Can't recommend going above DDR5-6000 (or your IMC 1:1 limit).

(2x16 GB) G.SKILL Trident Z5 DDR5-7200 F5-7200J3445G16GX2-TZ5RK (34-45-45-115-127)
(2x 16 GB) G.SKILL Trident Z5 NEO DDR5-6000 F5-6000J3038F16GX2-TZ5N (30-38-38-96)

LOW:

High:

 

[Mussels]

Now I'm waiting for peoples CPUs to die with VDDIO_mem at 1.45v+ lol.

Working on getting DDR5-8000 stable, but I don't see the benefit since it's Gear 2 now.

I saw that DDR5 9000 was plausible, but there was no real performance gain


It's good the platform can do it, but we need the knowledge of what the sweet spot in each gear is

 

[AusWolf]

I saw that DDR5 9000 was plausible, but there was no real performance gain


It's good the platform can do it, but we need the knowledge of what the sweet spot in each gear is

I don't know why anyone would want to run gear 2 with any RAM when 6000 MHz in gear 1 is faster than 7200 MHz in gear 2.

 

[lsevald]

A while back I got a 64GB kit thinking I might as well, since 6000ish is max anyway . I spent a few days dialing it in, and I guess it is still holds up. It can do slightly tighter primaries, but I opted for stability and lower voltages (this was in the middle of the Vsoc controversy). The cores runs -15CO on avg (P95 AVX2 and SHA3 stable) which is ok, but the IOD is nice. It has been running flawless like this for ~2 months now:



I'm seeing a couple of claims that the new agesa also makes it a little easier to run 1:1 6000-6400...

 

[Mussels]

I don't know why anyone would want to run gear 2 with any RAM when 6000 MHz in gear 1 is faster than 7200 MHz in gear 2.

because some value at some point, might match or beat it.
Math wise, it's going to depend on the divider they used

current best is 2000\6000
so 2000\8000? 9000?


It's a longevity thing for the platform rather than a 'useful now' situation

 

[Chry]

Sorry for noob and off-topic question, but how do I check if my RAM is running at gear 1 or 2? The board is as in signature.

 

[AusWolf]

Sorry for noob and off-topic question, but how do I check if my RAM is running at gear 1 or 2? The board is as in signature.

It's not your RAM, but the CPU memory controller.

Check out the "memory" tab in CPU-Z. If the uncore/NB frequency is half of your effective memory speed (for example, 3000 MHz if you're running 6000 MHz RAM), that's gear 1, if it's a quarter (that is, 1500 MHz for 6000 MHz RAM), that's gear 2. I don't even think this is labelled as "gears" on AMD. This terminology is an Intel thing (although, the concept is the same).

 

[Chry]

It's not your RAM, but the CPU memory controller.

Check out the "memory" tab in CPU-Z. If the uncore/NB frequency is half of your effective memory speed (for example, 3000 MHz if you're running 6000 MHz RAM), that's gear 1, if it's a quarter (that is, 1500 MHz for 6000 MHz RAM), that's gear 2. I don't even think this is labelled as "gears" on AMD. This terminology is an Intel thing (although, the concept is the same).

So looks like it's gear 1, so that's nice?

 

[ir_cow]

View attachment 306141
So looks like it's gear 1, so that's nice?

Running DDR5-5600 it seems.

Unrelated:
Something unexpected happened. This XPG 7200 kit has a EXPO profile, but if I run AID64, it restarts the computer instantly. But it passes y-cruncher and memtest5 (25 cycles). Kinda weird. Any idea?

 

[Chry]

Running DDR5-5600 it seems.

Unrelated:
Something unexpected happened. This XPG 7200 kit has a EXPO profile, but if I run AID64, it restarts the computer instantly. But it passes y-cruncher and memtest5 (25 cycles). Kinda weird. Any idea?

Could you test the kit without expo and see if the same happens?

 

[ir_cow]

I figured it out. CPU voltage and LLC is different per MB. Just had to adjust it for the Taichi.

 

[KLMR]

And the boot times? Any improvement?