3900X for $499.99 at Amazon

I know the 3950x is the new hotness ... but I think I'm gonna stick it out until next year for whatever is next. It's AMD's Tock cycle right? So that should be have improvements / refinements to the existing (Tick) 3xxx series right?
 
dvsman said:
I know the 3950x is the new hotness ... but I think I'm gonna stick it out until next year for whatever is next. It's AMD's Tock cycle right? So that should be have improvements / refinements to the existing (Tick) 3xxx series right?
Click to expand...

They don't have do tick/tock BS of Intel. The next chips are Zen 3, all new.
 
jmilcher said:
Partially true... Zen 3 supposedly will have new architecture. However...

Tick tick is a thing now for AMD.
https://www.extremetech.com/computi...ecture-for-zen-3-adopts-intel-tick-tock-model
Click to expand...

I fundamentally disagree with the tick/tock BS since Intel was giving 3%-5% gains and calling it a new generation. They'd been pulling this crap for decades. AMD on the other hand is making serious and major gains with each successive iteration. I don't see that as the same thing.
 
thesmokingman said:
I fundamentally disagree with the tick/tock BS since Intel was giving 3%-5% gains and calling it a new generation. They'd been pulling this crap for decades. AMD on the other hand is making serious and major gains with each successive iteration. I don't see that as the same thing.
Click to expand...
I’m pointing out the nomenclature is being used in reference to AMD’s frequent improvements and releases. And at one point Intel was doing the same thing. During the Core 2 days, the tick tock was relevant and was used in a similar way.

overall it’s the new buzz word. And you can define it however you’d like.
 
jmilcher said:
I’m pointing out the nomenclature is being used in reference to AMD’s frequent improvements and releases. And at one point Intel was doing the same thing. During the Core 2 days, the tick tock was relevant and was used in a similar way.

overall it’s the new buzz word. And you can define it however you’d like.
Click to expand...

Yea, I know, it is what it is. But it just fails to encapsulate the gains AMD is putting out.
 
I hope the prices come down about the time im ready to make my move to the 3900x
 
I have actually ordered another one. I'm not happy with the infinity fabric/memory controller on my 3900x. It also clocks pretty low compared to everything else I've seen. I guess it could be my memory, but b dies are b dies, right? I want to see if I can get a better one. It will drive me nuts if I don't try. I had to go through 3 2700k Sandies to fine a good one. The second one was the charm. I sold the first one and returned the last one to Amazon. The second one was the best clocker. I may or may not be selling my first 3900x.
 
kamikazi said:
I have actually ordered another one. I'm not happy with the infinity fabric/memory controller on my 3900x. It also clocks pretty low compared to everything else I've seen. I guess it could be my memory, but b dies are b dies, right? I want to see if I can get a better one. It will drive me nuts if I don't try. I had to go through 3 2700k Sandies to fine a good one. The second one was the charm. I sold the first one and returned the last one to Amazon. The second one was the best clocker. I may or may not be selling my first 3900x.
Click to expand...

No, not all are real B-dies. Dies are graded, but ppl often assume just cuz it's Samsung, it's real B-dies. What sticks did you get btw?
 
thesmokingman said:
No, not all are real B-dies. Dies are graded, but ppl often assume just cuz it's Samsung, it's real B-dies. What sticks did you get btw?
Click to expand...

G.Skill Trident Z RGB 4000 19-19-19-39 1.35v (F4-4000C19D-32GTZR). I verified B Die status on the B Die finder website. My computer won't even boot on xmp settings. It requires a CMOS reset. At 3733 with fclk at 1866, SoC at 1.1v, I need 1.43v on the memory for 15-16-16-32 to be stable. Fclk 1900 is a no go. I'm getting 8d errors on the mobo every so often resuming from hibernation or on reboots. Probably 1 in 4 or 5 boots requires the retry button. I've read that's an fclk stability issue. I'm sure it's fine at 1800 fclk and 3600 RAM speeds. I had some stable settings there. But this is HardOCP. I'm not running at 3600!

B Die.JPG
 
Last edited:
kamikazi said:
G.Skill Trident Z RGB 4000 19-19-19-39 1.35. I verified B Die status on the B Die finder website. My computer won't even boot on xmp settings. It requires a CMOS reset. At 3733 with fclk at 1866, SoC at 1.1v, I need 1.43v on the memory for 15-16-16-32 to be stable. Fclk 1900 is a no go. I'm getting 8d errors on the mobo every so often resuming from hibernation or on reboots. Probably 1 in 4 or 5 boots requires the retry button. I've read that's an fclk stability issue. I'm sure it's fine at 1800 fclk and 3600 RAM speeds. I had some stable settings there. But this is HardOCP. I'm not running at 3600!
Click to expand...

You're running the IF out of spec that's why. There is no gain going above 1:1 on the IF and yer actually nerfing it. Ideal is 3600mhz with tight timings. You can do up to 3800 and manually setting IF to 1:1. But that's about it. This has nothing to do with the 3900x IMC, directly that is.
 
thesmokingman said:
You're running the IF out of spec that's why. There is no gain going above 1:1 on the IF and yer actually nerfing it. Ideal is 3600mhz with tight timings. You can do up to 3800 and manually setting IF to 1:1. But that's about it. This has nothing to do with the 3900x IMC, directly that is.
Click to expand...

I'm running 1:1, fclk:mem speed at 1866. I really can't get any tighter timings at 1800. 16-15-15-32 at 1.43v. It's very temperamental with any slight changes. Best I've OC'd so far is 4.4 on the good chiplet and 4.3 on the craplet at 1.325v on the CPU. I may see what I can do at the 1800 fclk.

I've read about some people buying multiple Ryzens and having one be completely fine at fclk 1900 with memory at 1.35v and then the other being a mess at anything over 1800 even with high SOC and memory voltages. The infinity fabric is definitely not equal across all chips. Not saying you don't know what you're talking about. You for sure have done well with your system. I'm willing to try it out just for kicks and to satisfy my own curiosity.
 
kamikazi said:
I'm running 1:1, fclk:mem speed at 1866. I really can't get any tighter timings at 1800. 16-15-15-32 at 1.43v. It's very temperamental with any slight changes. Best I've OC'd so far is 4.4 on the good chiplet and 4.3 on the craplet at 1.325v on the CPU. I may see what I can do at the 1800 fclk.
Click to expand...

I bet your fclk is adding stress to it. I'm running 1800 and can clock to 4.5/4.4/4.35/4.35 at 1.32v. This nets me a lil under 7900 in R20 multi. I should add imo my chip is about average.
 
thesmokingman said:
I bet your fclk is adding stress to it. I'm running 1800 and can clock to 4.5/4.4/4.35/4.35 at 1.32v. This nets me a lil under 7900 in R20 multi. I should add imo my chip is about average.
Click to expand...

I just set mem to 3600 and fclk to 1800. I am cinebench stable with 4.4 on the good chiplet and 4.3 on the craplet. Going to 4.45 on the good chiplet throws an instant error in CB. Going to 4.5 reboots the computer. This is at 1.325v on the cpu.
 
There's 4 CCX across two CCD. Are you sure you've found the right order for best to worse?
 
As much as I want one of these, I think I’m going to hold out for the next gen. Worst case scenario, I’ll just buy a 3900x from someone upgrading. My 2700x is enough for now.
 
thesmokingman said:
There's 4 CCX across two CCD. Are you sure you've found the right order for best to worse?
Click to expand...

No idea, I've just been OCing per chiplet. Chip 0 at 4.4, chip 1 at 4.3. All core at 4.3 is about the best I can do. Mem is stable at 3733 on overnight memtest run at boot time with no errors.
 
kamikazi said:
No idea, I've just been OCing per chiplet. Chip 0 at 4.4, chip 1 at 4.3. All core at 4.3 is about the best I can do. Mem is stable at 3733 on overnight memtest run at boot time with no errors.
Click to expand...

There's 4 CCX, so you're missing two more? Are you doing this with Ryzen Master? I do it via bios.
 
I'm overclocking each entire chiplet, so ccd0 at 4.4 and ccd1 at 4.3. I am using Ryzen Master. I've tried in BIOS as well, no difference on behavior at my settings. I've been using Ryzen Master for quick tests. Once I get close, I'll switch to BIOS to lock it in.
 
kamikazi said:
I'm overclocking each entire chiplet, so ccd0 at 4.4 and ccd1 at 4.3. I am using Ryzen Master. I've tried in BIOS as well, no difference on behavior at my settings. I've been using Ryzen Master for quick tests. Once I get close, I'll switch to BIOS to lock it in.
Click to expand...

Yer supposed to oc per CCX not CCD.
 
thesmokingman said:
Yer supposed to oc per CCX not CCD.
Click to expand...

Didn't realize that. I thought going for each whole chiplet was the way to go. Do you just do trial and error to figure out which CCXs do what? I know that CCD 0 is the always the best overall chiplet, right? On my CPU CCD1 is no good past 4.3 for sure.

When I ran Prime for about 10 minutes at 4.4 ccd0/4.3 ccd1 with CPU voltage at 1.325v, my computer rebooted and the BIOS indicated it had shut down due to a high temp warning. HWINFO showed temps never got over 72, both from the Ryzen sensors and the Asus ones. What gives?
 
kamikazi said:
Didn't realize that. I thought going for each whole chiplet was the way to go. Do you just do trial and error to figure out which CCXs do what? I know that CCD 0 is the always the best overall chiplet, right? On my CPU CCD1 is no good past 4.3 for sure.

When I ran Prime for about 10 minutes at 4.4 ccd0/4.3 ccd1 with CPU voltage at 1.325v, my computer rebooted and the BIOS indicated it had shut down due to a high temp warning. HWINFO showed temps never got over 72, both from the Ryzen sensors and the Asus ones. What gives?
Click to expand...

Yea by ea CCX otherwise the whole CCD will be limited by the weakest CCX in it. I start with my all core voltage then bump the ratio on my best usually CCD0/CCX0, and raise its ratio to find the max. Then CCD0/CCX1, then CCD1/CCX0, etc etc. For ex my 3900x does 4.5CCX0/4.3CCX1 on CCD0 and the other one 4.4/4.3.
 
thesmokingman said:
Yea by ea CCX otherwise the whole CCD will be limited by the weakest CCX in it. I start with my all core voltage then bump the ratio on my best usually CCD0/CCX0, and raise its ratio to find the max. Then CCD0/CCX1, then CCD1/CCX0, etc etc. For ex my 3900x does 4.5CCX0/4.3CCX1 on CCD0 and the other one 4.4/4.3.
Click to expand...

Thanks, I'll give it a shot.
 
kamikazi said:
Tangoseal picked one up at Microcenter in Marietta, GA. There were two of them. Same store had shelves full of 3950x's as well.
Click to expand...
Sorry for OT - Would you consider picking and shipping me a 3950 if I PP or Venmo you in advance?
 
thesmokingman said:
Yea by ea CCX otherwise the whole CCD will be limited by the weakest CCX in it. I start with my all core voltage then bump the ratio on my best usually CCD0/CCX0, and raise its ratio to find the max. Then CCD0/CCX1, then CCD1/CCX0, etc etc. For ex my 3900x does 4.5CCX0/4.3CCX1 on CCD0 and the other one 4.4/4.3.
Click to expand...

So, I did some quick and dirty OC'ing to find the limits of my chip by CCX. First, I set infinity fabric fclk to 1800 and RAM to 3600. I backed off on timings to 16-16-16-16-36-52. In Ryzen Master, I set cpu voltage to 1.3125. I set all cores to 4250, then I went back and went through each CCX one at a time seeing how high I could go before throwing errors in a 5 minute CB20 loop. On the "good" chiplet, CCD0, CCX1 maxed at 4400 and CCX2 maxed at 4400. On the "craplet", CCD1, CCX0 maxed at 4300 and CCX1 maxed at 4400. I noticed that I got slightly higher temps running looking for maxes on CCD1. Once I found these, I set each CCX to max and tried a to look CB20 for 5 minutes, which is 8 loops. It scored 7689, but failed on pass 4. I raised CPU voltage to 1.31875 and it made the 5 minutes with a score of 7685, but on the very last pass, there are artifacts. Probably needs one more voltage bump for true stability Temps maxed around 75C.

If I set everything to default CPU wise, running CB20 for 5 minutes sees max clocks on any one core of 4017.8 which is the 40.3 multi with Bus clock at 99.8, but just for a millisecond. All cores seem to sit at 3992.8 which is the 40 multi. So what level silicon am I dealing with here? I got my other 3900x in and I'll be running the same exact tests Friday I hope.
 
kamikazi said:
So, I did some quick and dirty OC'ing to find the limits of my chip by CCX. First, I set infinity fabric fclk to 1800 and RAM to 3600. I backed off on timings to 16-16-16-16-36-52. In Ryzen Master, I set cpu voltage to 1.3125. I set all cores to 4250, then I went back and went through each CCX one at a time seeing how high I could go before throwing errors in a 5 minute CB20 loop. On the "good" chiplet, CCD0, CCX1 maxed at 4400 and CCX2 maxed at 4400. On the "craplet", CCD1, CCX0 maxed at 4300 and CCX1 maxed at 4400. I noticed that I got slightly higher temps running looking for maxes on CCD1. Once I found these, I set each CCX to max and tried a to look CB20 for 5 minutes, which is 8 loops. It scored 7689, but failed on pass 4. I raised CPU voltage to 1.31875 and it made the 5 minutes with a score of 7685, but on the very last pass, there are artifacts. Probably needs one more voltage bump for true stability Temps maxed around 75C.

If I set everything to default CPU wise, running CB20 for 5 minutes sees max clocks on any one core of 4017.8 which is the 40.3 multi with Bus clock at 99.8, but just for a millisecond. All cores seem to sit at 3992.8 which is the 40 multi. So what level silicon am I dealing with here? I got my other 3900x in and I'll be running the same exact tests Friday I hope.
Click to expand...

What is your LLC set to? I suggest setting your vrms on cpu and soc to extreme, raise your power to 130. I have my LLC set to 4 iirc.
 
thesmokingman said:
What is your LLC set to? I suggest setting your vrms on cpu and soc to extreme, raise your power to 130. I have my LLC set to 4 iirc.
Click to expand...
I believe that VRMS are already set to extreme. I have LLC on 3 though. SOC voltage is set to 1.1125 so that it sees 1.1 when loaded.
 
You must log in or register to reply here.
Back
Top