AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

[Lolik]

added troll easyrider to ignore list

 

[chroniclard]

here ya go - the crux of it was that der8auer couldn't get the cpu to maintain or even reach it's stated boost speed.

I see, ta.

Its almost like broadband with its "upto" speeds.

 

[emodan]

Managed to get my 8Pack RAM to CL14 at 3733Mhz but a friend has tested and found it better to stick with 3600 CL14. Will do some memory testing tonight as once I find the sweet point I'm done at last with my rebuild and can focus on games

 

[Thekwango]

I see, ta.

Its almost like broadband with its "upto" speeds.

seems so, at least at the moment.

 

[N19h7m4r3]

Really interesting that the dual-die design of the R9 3900X/3950X is consistently slightly faster than the single-die design of the R5 3600(X) and R7 3700X/3800X. I wonder if that is a result of the gimped memory write of the single-die designs? If it is, it's a justification for the decision I suppose: halving the bandwidth only affects performance in typical tasks by a tiny margin (a few % at most). Or maybe it's just the larger L3 cache per core?

It's likely the access to more Cache than anything else really. Two 6 core chiplets with each the same Cache as a single 8core chiplet like the 8 Cores.

 

[alec]

Or maybe it's just the larger L3 cache per core?

3600X has same cache configuration as 3900X

 

[Nghtmare]

3600X has same cache configuration as 3900X

No it has half the L3 Cache. 32 vs 64MB

 

[alec]

No it has half the L3 Cache. 32 vs 64MB

Same amount of cache per core. 3 cores share 16MB of L3
And cores can't read from L3 of other CCX, so it should make no difference to IPC whether there are 2 CCX or 4.

 

[Sargatanas2511]

I wouldn't be surprised if a future iteration uses some I/O die space for L4 cache. 256/512MB of L4 should help a lot.

 

[esoteric]

My cheap Taichi has been despatched

 

[chrcoluk]

I run the 8 Pack 3200CL14 at stock.. which is 14-14-14-31-1T with 1.360v/1.1v Soc on 2700X 4.2Ghz All Core on an X470 MSI Gaming Pro Carbon AC.

No issues at all..

To correct the thread from my earlier post regarding 1.45v.

My ram in my ryzen 2 rig is actually at 1.35v.
It was my b die in my intel rig I put at 1.45v (needed to push it to CL12).

This 3200CL14 8 pack ram has no thermal monitoring tho, so I couldnt monitor the temp effects, it seems tho on hot summer days with it at 1.45v the temp increase is enough to make my 65535 ram refresh cycle timing (trefi) an issue. On one hot day a couple of weeks back I had weird crashes that didnt go away until I rebooted and it probably was due to the ultra aggressive trefi I set, so that will either go down to 32768 or back to what I had before which was 11430 (oc.net suggest 12k as a safe value thats better than stock, 11430 is what I calculated to be jedec spec to match the higher clock speed of the ram so thats probably how they got to 12k). Basically how far trefi can be pushed is linked to the temperature of the ram, as hotter ram will get flipped bits quicker and its the reason why dram has to self refresh. I may also decide to keep 65535 trefi, but just back off the voltage and go back to CL14, as maxed out trefi has decent gains to both bandwidth and latency, whilst CL12 just helps latency.

 

[SiDeards73]

My cheap Taichi has been despatched

Mines been sat at home since last Wednesday just waiting on my 3800x to get dispatched so i can build.

 

[alec]

If the cores can only use the cache on their own CCX then 3 cores sharing 16MB vs 4 cores sharing 16MB = approx. 33% more cache per core for the 3900x. Not to mention some other variables at play here.

Yep, thats why I wrote 3600X and 3900X have same cahce config. The upcoming 3950X will be similar to 3700X.

Didn't watch IPC video yet, but if 3900X has higher IPC there, something must be special about 2 chiplet design. I doubt write bandwidth alone can cause that

 

[brightside229]

guys!11 3800x showing in stock at ocuk!

 

[Besty]

point is you buy 4133 kit to run 3600 with super tight timings ?? If i got more money to burn i wuld be on 4500mhz kit to run it at 3600 cl13 or something.

binning of B-Die ram kits for low latency CL/1T don't work that way, it never has.

best advice to save money and time, just buy the 8 pack kits.

if money and time unlimited - buy some 3600 C16 or C17 kits, non-RGB and bin those but it is a slippery slope. The peeps binning 100+ kits each have an approx 10% hit rate for good CL bins.

I am kinda happy this is now a 'thing' because it means that companies will stop trying to throw crap 4200+ C19 kits down our throats and get us some proper ram.

 

[Pete77777]

guys!11 3800x showing in stock at ocuk!

 

[Zeed]

binning of B-Die ram kits for low latency CL/1T don't work that way, it never has.

best advice to save money and time, just buy the 8 pack kits.

if money and time unlimited - buy some 3600 C16 or C17 kits, non-RGB and bin those but it is a slippery slope. The peeps binning 100+ kits each have an approx 10% hit rate for good CL bins.

I am kinda happy this is now a 'thing' because it means that companies will stop trying to throw crap 4200+ C19 kits down our throats and get us some proper ram.

Have You missed part where I upgraded from Binned 3733 mem kit to binned 4133 8pack ones ?? I highly dubt that Zen2 8pack mem kits will be running super tight timings 3633 at 1.45 - 1.5 volts xmp profile.

I looked around for 3600cl14 kit does not exist. I know That Gskill used to sell 3600cl15@@1.35 sometime ago
https://www.tweaktown.com/reviews/7...6gb-dual-channel-memory-kit-review/index.html

Thats best kit I'w seen but cant buy it anymore :/ those did 3600cl14/13/13/35



Ill try to move to 3600 14/13/13/35 t1 with this kit. got many options to find whats up

 

[Wingz]

oh nice new AGESA 1.0.0.3 AB BIOS just released for Gigabyte B450/x470 Boards

1. Update AGESA 1.0.0.3 AB
2. Improve M.2 SSD compatibility