Overclock.net banner

101 - 120 of 185 Posts

·
Registered
Joined
·
18 Posts
I was able to pick up a NIB 3900x and NIB 3900XT recently for under retail, the XT $25 more than the X. Based on OC experiences with each chip and price which would you keep? The 3900x is a late 2019 batch number. Only dabbled with an early release 3600x so far which was underwhelming but not bad (reached 1867 IF but couldn’t post above that and all-core OC was relatively low).
If XT is just $25 extra than X, I would go for XT because it seems they are slightly better binned chips. And various users over Internet have claimed that the later lot of chips hold up higher frequency better than 2019 batch.

I have 3900x bought in Oct 2019 and I believe my chip is average binned, I can run all core 4.25 Ghz at 1.25v stable (not Prime 95 stable). My CR20 score is around 7400 multicore which keeps changing up or down based on the voltage I fiddle with. But I feel my chip runs more stable around 1.3v.

I had even tried Auto OC mode, or EDC bug and had seen all 6 cores on my CCX0 able to reach 4.6 Ghz though not together same time but individually.

Sent from my Redmi Note 7 Pro using Tapatalk
 

·
Registered
Joined
·
22 Posts
About stability, I havent tested it to much yet, its probably not prime95 stable, but I havent had the time to test it proper, I guess I need more voltage for prime95 stable, but I dont like prime95 cause no programs will ever stress your cpu that much that prime95 does.

I am gonna fiddle a bit more and try to lower vsoc voltage etc etc
I run the default AIDA64 stress for an hour or so. There are certain torturetests that will crash an AIDA stable system (P95 probably can) but since I don't run any workloads that are mission critical, I consider AIDA stable good for a daily driver. My prev posted config i'm running [email protected] 24/7 (100%GPU and 8 CPU threads for total of 40-50% 24-7 CPU load) without issues. Giving 24 cpu threads puts FAH into some wild computations reminiscent of P95, so I don't let it run full tilt. I am at 313h uptime at the time of writing and have done quite a few gaming sessions, run a bunch of VMs, video rendering, done some benching and ofc [email protected] for over 300h.


I was able to pick up a NIB 3900x and NIB 3900XT recently for under retail, the XT $25 more than the X. Based on OC experiences with each chip and price which would you keep? The 3900x is a late 2019 batch number. Only dabbled with an early release 3600x so far which was underwhelming but not bad (reached 1867 IF but couldn’t post above that and all-core OC was relatively low).
I got my XT for $30 more than the X and I believe it's money well spent. While you may get a lucky X, the XTs are binned X's which suggests a very recent X will be binned as such and if it performed better it would be considered an XT. Not that it really matters for the few % perf gain, but a big thing for me is if I decide to sell the chip in a year, the XT may depreciate more slowly.

In fact i bet someone could do a cost/benefit analysis to show what $30 for an XT would do for performance compared to spending $30 more on better cooler/fans
 

·
Registered
Joined
·
22 Posts

·
Registered
Joined
·
31 Posts
SMT isn't going to really show you any benefits in CPU-Z. Some games CAN see a benefit keeping 1T1C parity, others not so much.


SMT ON I have better SC & MC (clock speed being the actual factor not SMT):

https://valid.x86.fr/hnt5v8
Almost all games cannot understand flows.
Games think they are loading 2 cores, but in reality they are loading 2 threads with one core.
Overclocking the processor with disabled SMT is more difficult.
When the pipes (SMT) are turned off, the delays increase.
Your test says nothing. Many here have pressed a button and boast of overclocking the processor.

https://www.3dmark.com/3dm/49692395

Disabling SMT is necessary for online games. :)


.
 

·
Registered
Joined
·
22 Posts
Almost all games cannot understand flows.
Games think they are loading 2 cores, but in reality they are loading 2 threads with one core.
Overclocking the processor with disabled SMT is more difficult.
When the pipes (SMT) are turned off, the delays increase.
Your test says nothing. Many here have pressed a button and boast of overclocking the processor.

https://www.3dmark.com/3dm/49692395

Disabling SMT is necessary for online games. :)

.
I disagree ALL games but agree SOME games.

Your test says nothing. Many here have pressed a button and boast of overclocking the processor.
If you look at what I wrote above I posted my test to show that SMT != CPUZ performance.... you are (abrasively) agreeing with me...

When the pipes (SMT) are turned off, the delays increase.
I think you mean decrease :)

Disabling SMT is necessary for online games. :)
No :) You might find it works for you in what you play but it's not necessary :)

Anyways, I'm not saying you are wrong.... Hyperthreading offers marginal performance boosts in most modern workloads and some decrease in others. We can agree to 80% agree since we are 2 dogs barking at the same moon. All I was saying is that not all games/workloads that's the case... and... you could turn off half your cores to get a gain in performance, would likely help far more than disabling hyperthreading.
 

·
Registered
Joined
·
31 Posts
If you look at what I wrote above I posted my test to show that SMT != CPUZ performance.... you are (abrasively) agreeing with me...
I think you mean decrease :)
No :) You might find it works for you in what you play but it's not necessary :)
You just think so.
Look in the 3dmark test what my processor frequencies are. :) :) :)
With the SMT on, I have a memory latency of 69 ms.
With the SMT disabled, I have a delay of 73ms
I don’t want to prove anything to you and show you tests.
I assure you that without the included SMT you will lose to me in tests.
Moreover, if I turn on the SMT, then you will also lose.
 

·
Registered
Joined
·
31 Posts
I run the default AIDA64 stress for an hour or so. There are certain torturetests that will crash an AIDA stable system (P95 probably can) but since I don't run any workloads that are mission critical, I consider AIDA stable good for a daily driver. My prev posted config i'm running [email protected] 24/7 (100%GPU and 8 CPU threads for total of 40-50% 24-7 CPU load) without issues. Giving 24 cpu threads puts FAH into some wild computations reminiscent of P95, so I don't let it run full tilt. I am at 313h uptime at the time of writing and have done quite a few gaming sessions, run a bunch of VMs, video rendering, done some benching and ofc [email protected] for over 300h.
Arithmetic performance is locked for Ryzen 9 3900X, the 610 Gflops mark cannot be overcome under any conditions and at any frequencies.
 

·
Registered
Joined
·
236 Posts
Hi all wanted ask if 5.0ghz on 3900xt with a water cooler is stable for people here? I hope i can set the boost to 5.0ghz with 3600hmz ram?
 

·
Registered
Joined
·
2,208 Posts
Hi all wanted ask if 5.0ghz on 3900xt with a water cooler is stable for people here? I hope i can set the boost to 5.0ghz with 3600hmz ram?
Are you thinking single core or all core?
 

·
Registered
Joined
·
236 Posts
Just wanted to check with you guys while running prime 95 i see it only pushing 4.2ghz max seen the cpu hit 4.3ghz max under cinebench. What can i use to hit that 4.7ghz spot?

Noticed the temps stay around 30-50degress then hit 72 degress max so far under load ive yet to test gaming i use a water cooler.

Never seen my bus speed go above 99.8mhz either
 

·
Registered
Joined
·
41 Posts
The 3900XT will only hit 4.7 on very specific, and rather light, loads. You'll mostly see 4.7 if you leave HWMonitor open on your desktop with a web browser open. I've even seen 4.8+ on one of my cores. Did I notice that blazing fast speed? Nope.

Under heavy single-dual core usage, the most you'll see is 4.5. Under heavy multi-core usage, you'll see 4.0 to 4.3. That's stock or PBO+AutoOC. You might get a little better all core than that with a manual OC, but you're not going to see 4.7 that way, ever.
 

·
Registered
Joined
·
91 Posts
One thing that have hit me is do you really earn performance with an 3900XT vs 3900X, the XT version doesnt like BCLK oc, doesnt like edc bug or anything!

So do you really earn performance when using PBO? Or do you only earn some performance when Manual OC the XT version cause of lower voltage?

What do you guys think? I am trying to think outside the box, My temps are under 70 deegres celcius so it should boost better imo! But what I have read is if you BCLK oc a normal 3900X that has more headroom cause of the XT version is so maxxed as it is. You can get similar performance or better.
 

·
Registered
Joined
·
236 Posts
The 3900XT will only hit 4.7 on very specific, and rather light, loads. You'll mostly see 4.7 if you leave HWMonitor open on your desktop with a web browser open. I've even seen 4.8+ on one of my cores. Did I notice that blazing fast speed? Nope.

Under heavy single-dual core usage, the most you'll see is 4.5. Under heavy multi-core usage, you'll see 4.0 to 4.3. That's stock or PBO+AutoOC. You might get a little better all core than that with a manual OC, but you're not going to see 4.7 that way, ever.

Undervolting should help prolong cpu life right? will i see any issues comes from my ram with undervolting as well? Ram is 32gb @3600mhz 16-19-19-37.

Gonna try undervolt and see if i can get cores running at 4.5ghz. Is there anyway to hit a constant boost clock of 4.5ghz. After a few days of watching hwmonitor and speccy all corres hit 4.3ghz and only 2 might touch 4.7ghz.

From my digging it seems that amds cool and quite function might be holding things back as well? any other settings i should enable? For instance precision boost overclock is on auto in bios
 

·
Registered
Joined
·
18 Posts
Undervolting should help prolong cpu life right? will i see any issues comes from my ram with undervolting as well? Ram is 32gb @3600mhz 16-19-19-37.



Gonna try undervolt and see if i can get cores running at 4.5ghz. Is there anyway to hit a constant boost clock of 4.5ghz. After a few days of watching hwmonitor and speccy all corres hit 4.3ghz and only 2 might touch 4.7ghz.



From my digging it seems that amds cool and quite function might be holding things back as well? any other settings i should enable? For instance precision boost overclock is on auto in bios
May be if your LLC is left on auto then it could be controlling things more conservatively and hence low boost. Try manually setting your LLC to lvl 3 and see if that helps.

If I leave all my settings to auto atleast 6 of my cores randomly boost close to 4.6 Ghz.

Sent from my Redmi Note 7 Pro using Tapatalk
 

·
Registered
Joined
·
1,523 Posts
What can i use to hit that 4.7ghz spot?
Here's an exaple you can try in PowerShell:

foreach ($loopnumber in 1..2147483647) {$result=1;foreach ($number in 1..2147483647) {$result = $result * $number};$result}
 

·
Registered
Joined
·
236 Posts
Noticed i hit 1.5v on cpu voltage currently troube shooting it i suspect thats why im not seeing higher clocks cpu gets hot really fast during benchmarking then slows down.
 

·
Registered
Joined
·
236 Posts
Here's an exaple you can try in PowerShell:

foreach ($loopnumber in 1..2147483647) {$result=1;foreach ($number in 1..2147483647) {$result = $result * $number};$result}

No idea what your talking about
 
101 - 120 of 185 Posts
Top