Zen2's 7nm Complications: Why not All Ryzen 3000 Cores Are Created Equal

Gideon

2[H]4U
Joined
Apr 13, 2006
Messages
2,720
Nope its across the board. I have x570 VIII hero. Shamino today uploaded the 1.0.0.3ABB link and basically said don't expect better boost clocks from AGESA 1.0.0.3 FW version because 1.0.0.2 had the right boost clocks. So it looks like Agesa is the reason for the boost clocks and they just need to iron this shit out.
Most of the guys on x570 have not updated their bios that I had seen, but yeah if you updated from the shipping bios then yeah you can end up in the same boat. I only updated my bios on the x370 because I had memory issues even running at slower speeds, but with that bios my chip boosted to within 25 MHz of the max speed on the box. I would say with the new AGESA their goal has been compatibility and then I expect they will work on boost clocks. So I am with ya on the whole AGESA being a big part of it.
 

NKD

Supreme [H]ardness
Joined
Aug 26, 2007
Messages
8,063
Most of the guys on x570 have not updated their bios that I had seen, but yeah if you updated from the shipping bios then yeah you can end up in the same boat. I only updated my bios on the x370 because I had memory issues even running at slower speeds, but with that bios my chip boosted to within 25 MHz of the max speed on the box. I would say with the new AGESA their goal has been compatibility and then I expect they will work on boost clocks. So I am with ya on the whole AGESA being a big part of it.
Yea I think the old AGESA is on the CD if I am not mistaken lol. But cant find my damn portable CD drive lol. I was going to double check for the kicks of it.
 

ccityinstaller

Supreme [H]ardness
Joined
Feb 23, 2007
Messages
4,239
Nope its across the board. I have x570 VIII hero. Shamino today uploaded the 1.0.0.3ABB link and basically said don't expect better boost clocks from AGESA 1.0.0.3 FW version because 1.0.0.2 had the right boost clocks. So it looks like Agesa is the reason for the boost clocks and they just need to iron this shit out.

I was stuck on 1.0.0.2 until this morning. Boost clocks worked right for a while, but changing almost anything in the BIOS (mostly ram related) would break things. I could not get 3 different ram kits working, so I said F it and ordered a 32GB kit of Samsung B die based G. Skill. I certainly hope that with it and the new Bios I can at least get the rated speeds working. This CPU/mobo/bios does not like anything Micron based. I have wasted so much time that I could have paid for the new ram kit with 1/10th of the time I spent trying to get the Micron kits stable.

I am seeing my CPU reach 4360 (bclk is always at like 99.5~99.8 never 100 ever even with my old 1600/2700) on just the wraith max so hopefully once EK gets my block here I can get it under water and see what happens then.
 

NKD

Supreme [H]ardness
Joined
Aug 26, 2007
Messages
8,063
I was stuck on 1.0.0.2 until this morning. Boost clocks worked right for a while, but changing almost anything in the BIOS (mostly ram related) would break things. I could not get 3 different ram kits working, so I said F it and ordered a 32GB kit of Samsung B die based G. Skill. I certainly hope that with it and the new Bios I can at least get the rated speeds working. This CPU/mobo/bios does not like anything Micron based. I have wasted so much time that I could have paid for the new ram kit with 1/10th of the time I spent trying to get the Micron kits stable.

I am seeing my CPU reach 4360 (bclk is always at like 99.5~99.8 never 100 ever even with my old 1600/2700) on just the wraith max so hopefully once EK gets my block here I can get it under water and see what happens then.
Do you have the same board? I didn't spend much time with the original bios. But newere bios I have crucial ballistix 32gb 3200mhz kit running at 3533 16-17-16-32 at 1.4v. which is micro e-die I believe. That is not bad for 2 15gb sticks.
 

ccityinstaller

Supreme [H]ardness
Joined
Feb 23, 2007
Messages
4,239
Do you have the same board? I didn't spend much time with the original bios. But newere bios I have crucial ballistix 32gb 3200mhz kit running at 3533 16-17-16-32 at 1.4v. which is micro e-die I believe. That is not bad for 2 15gb sticks.

No I am using a upper mid-range x470, the prime pro since I owned the x370 Prime Pro prior which bricked it's bios and I needed 3x 16c slots since I was running triple water blocked VEGA 56s.

That is a very nice speed/timing setup. I have a Corsair 3000c15 kit and the same Crucial 32GB 3200 kit you have (Micron D and R die, respectively) and could not get them stable no matter what I did.


I managed to get the e die kit to 3400c16 across but it was only about 90% stable. I could do anything but pass a memory stress test over 2hrs. I tweaked it for a solid 20 hours over both kits and neither was stable even at their rated XMP speeds.

I could run 3600c15-16-17 on the E die but the old 1.0.0.2 would drop my IF speed to 840Mhz (and spikr the latency and drop ram benchmark results) despite the bios saying the 1800Mhz override for IF speed was properly set.


I got fed up and managed to pickup a G.skill c14 b die based kit (still 16GBX2) yesterday only to see Asus drop the new bios this morning. Since the micron based kits are going back I figured the extra $42 I spent for the B die was worth it and am not going to mess with trying to get it stable.
 

NKD

Supreme [H]ardness
Joined
Aug 26, 2007
Messages
8,063
No I am using a upper mid-range x470, the prime pro since I owned the x370 Prime Pro prior which bricked it's bios and I needed 3x 16c slots since I was running triple water blocked VEGA 56s.

That is a very nice speed/timing setup. I have a Corsair 3000c15 kit and the same Crucial 32GB 3200 kit you have (Micron D and R die, respectively) and could not get them stable no matter what I did.


I managed to get the e die kit to 3400c16 across but it was only about 90% stable. I could do anything but pass a memory stress test over 2hrs. I tweaked it for a solid 20 hours over both kits and neither was stable even at their rated XMP speeds.

I could run 3600c15-16-17 on the E die but the old 1.0.0.2 would drop my IF speed to 840Mhz (and spikr the latency and drop ram benchmark results) despite the bios saying the 1800Mhz override for IF speed was properly set.


I got fed up and managed to pickup a G.skill c14 b die based kit (still 16GBX2) yesterday only to see Asus drop the new bios this morning. Since the micron based kits are going back I figured the extra $42 I spent for the B die was worth it and am not going to mess with trying to get it stable.
honestly been looking at the gskill ones, I didn't spend much on my current kit so can easily resell it or save it for another build. But can't really find the gskill ones in stock anywhere. I am assuming you got the neo kits correct? Where did you get them.
 

N4CR

Supreme [H]ardness
Joined
Oct 17, 2011
Messages
4,687
Well, in reverse, if this was Intel they would have been dragged out behind the shed and shot.
Eh? It's the same for Intel, some cores are faster than others.

I'll bid on nothing new for $2, thanks.
 

ccityinstaller

Supreme [H]ardness
Joined
Feb 23, 2007
Messages
4,239
honestly been looking at the gskill ones, I didn't spend much on my current kit so can easily resell it or save it for another build. But can't really find the gskill ones in stock anywhere. I am assuming you got the neo kits correct? Where did you get them.
No, I got just the RipJaws 32gb kit that uses B die (14-14-14-28-1T). They are rated @3kc14 since u believe they are most likely dual rank to get the density in 2 DIMMs. I got them from Newegg for $190.

I am sure I can get then up to 3200c14-34/3600c16 but honestly if they run 3kc14 across I am good with that. The extra 22pts in CB20 are well worth the trade off for stability and having 32GB vs 16GB.

We are on the tail end of DDR4, and I just cannot see spending $140 for a 16GB b die kit when $190 doubled that. Those new 3800 kits are probably going to o be $200 for a 16GB kit.
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,171
The article is misleading and really not 100% accurate for 3600X.I have a 3600X which can hit 4525Mhz on all cores and up to 3CCX or 6 cores same time.
Hell I did a short single thread run and 3600X Hitting 4500Mhz Cinebench 20 and over 4400Mhz Max boost on 3CCX in a minute video.
That not how evidence works.
You cant prove a negative and especially not with a single sample.

or put it this way. Just because yours can does not mean all can, and that is the issue at hand. Yhat the same product apparently are not the same even under the same conditions.

-- Edit --
This is Why i always hated dynamic/boost clock. Cause you dont really know what you get. its never a promised./ just a well sometimes you get this speed.
It has to be well defined how those clocks on the product is working
 
Last edited:

Trimlock

[H]F Junkie
Joined
Sep 23, 2005
Messages
15,200
That not how evidence works.
You cant prove a negative and especially not with a single sample.

or put it this way. Just because yours can does not mean all can, and that is the issue at hand. Yhat the same product apparently are not the same even under the same conditions.

-- Edit --
This is Why i always hated dynamic/boost clock. Cause you dont really know what you get. its never a promised./ just a well sometimes you get this speed.
It has to be well defined how those clocks on the product is working
reminds me of the famous Onion article



untitled234.png
 
Joined
Feb 23, 2009
Messages
995
That not how evidence works.
You cant prove a negative and especially not with a single sample.

or put it this way. Just because yours can does not mean all can, and that is the issue at hand. Yhat the same product apparently are not the same even under the same conditions.

-- Edit --
This is Why i always hated dynamic/boost clock. Cause you dont really know what you get. its never a promised./ just a well sometimes you get this speed.
It has to be well defined how those clocks on the product is working
I have a larger sample size than the article,I own more than 1 3600X.
 

NKD

Supreme [H]ardness
Joined
Aug 26, 2007
Messages
8,063
No, I got just the RipJaws 32gb kit that uses B die (14-14-14-28-1T). They are rated @3kc14 since u believe they are most likely dual rank to get the density in 2 DIMMs. I got them from Newegg for $190.

I am sure I can get then up to 3200c14-34/3600c16 but honestly if they run 3kc14 across I am good with that. The extra 22pts in CB20 are well worth the trade off for stability and having 32GB vs 16GB.

We are on the tail end of DDR4, and I just cannot see spending $140 for a 16GB b die kit when $190 doubled that. Those new 3800 kits are probably going to o be $200 for a 16GB kit.
I guess i should be happy with what i got then. i just need to adjust subtimings to get little better latency I guess. I am running at 3533 16-17-16-36 for my E-die kit. It goes from 69-72ns. Its weird how my memory overclocks. I can't even run 3200 at 16-16-16-36 but I can run 3533 at 16-17-16-36 lol. I have stress tested it over 24 hours and did another 14 hours last night to be sure.

Haven't used the dram calculator, will try doing that next for subtimings.
 

ccityinstaller

Supreme [H]ardness
Joined
Feb 23, 2007
Messages
4,239
I guess i should be happy with what i got then. i just need to adjust subtimings to get little better latency I guess. I am running at 3533 16-17-16-36 for my E-die kit. It goes from 69-72ns. Its weird how my memory overclocks. I can't even run 3200 at 16-16-16-36 but I can run 3533 at 16-17-16-36 lol. I have stress tested it over 24 hours and did another 14 hours last night to be sure.

Haven't used the dram calculator, will try doing that next for subtimings.

That mirrors my experience as well. 3200c16 was a no go from the beginning but 3600 was working well, but the FClK bug was keeping me from considering even trying to get it 100% stable.


I will report back with what I can get the B die kit at.
 
Joined
Feb 23, 2009
Messages
995
Do not care dude also I did not read any of your post ,so I do not even know what your talking about.
 
Last edited:
Top