New Ryzen 3700x upgrade 3 times worse performance?
|
At the moment i play with a 2500K and a GTX 1050 TI and can't complain.
All you can do is updating all incl. Bios. New Generation of Hardware like the Ryzen can have some bugs. Don't believe the Hype!
| |
" Yes I have updated bios ,graphic card driver,window 10 to 1903 ,played World of Warcraft ,Pubg and someother games .all gone well but Poe. | |
|
Ryzen 3700x at stock boost settings
16 gigs of 3600HZ ram 1070ti latest drivers/bios for both GPU and CPU. latest windows update I'm having similar issues, I've been playing POE on a 3770k paired with a 1070ti for quite a while. I recently upgraded to a 3700x, and performance is actually worse overall(in POE). I was farming glaciers with vaal ice nova, and upon popping the monolith and vaal ice novaing, I would nearly lag out completely, depending on the size of the monolith (of course I expected to lag, I'm shattering and exploding a ton of mobs all at once). I'd recover maybe 75% of the time to be able to get the 2nd vaal ice nova off in time to not die. After updating to the 3700x and attempting the same thing, I would completely lag out and be unable to get that second vaal ice nova off. I MAYBE recover 15% of the time or so to get that 2nd ice nova off, but much more rare than before. I'm not expecting to be able to vaal ice nova huge packs without some sort of lag, my friend with a worse computer is able to do it, while sharing screens in discord without lagging out like I was, but I honestly wasn't expecting to lag out worse than with my 3770k. That's not a real case scenario, so I'm not too worried about lagging out there. However, now even through general gameplay (right now I'm tornado shot), if I explode a pack with any more than maybe 15-20 mobs, I lag/freeze. Monoliths are really rough, I need to clear parts of them at a time or risk freezing up for a second or two, which wasn't a problem with my 3770k. Overall the game is playable, but it's a worse experience than my 7+ year old CPU, I don't know if it's a hardware issue, the newness of the new zen series, POE, or a mixture of all 3. | |
|
I'm working on making some graphs & video at the moment, but basically what I'm seeing is that the game isn't limited by CPU [processing power] or GPU or thermals. Overclocking CPU doesn't increase performance and the GPU is far from being fully utilised. The only remaining culprit can be memory/memory latency. Once a guide comes out for Ryzen 2 memory tweaking then I'll adjust the timings & see what the result is.
| |
|
Hmmmm.. you will get some more FPS but i don't think Mem Tweaking makes a unplayable Game playable.
" Try with disabled SMT. Don't believe the Hype! Last edited by bruts on Jul 27, 2019, 8:12:35 AM
| |
|
I've tried with SMT off, I've tried different power plans, I've played with PBO on, PBO off, 4.2 all core OC, tried reverting BIOS, I tried --nosound just to see if it was sound related. I've been tweaking/messin around with whatever I could think of but if there's been a difference it hasn't been noticeable.
| |
" Am at c14 on timings @3200mhz I'll try overclocking to 3600mhz@c14 later I doubt it will do much tho |
|
|
Just tried it at 3600mhz@c14 and its still the same :(
|
|
|
New videos:
https://www.youtube.com/watch?v=0nC7WdZbNaI&feature=youtu.be https://www.youtube.com/watch?v=hgkSJxCS3Bw&feature=youtu.be First is everything stock, 1080p, default graphics settings in PoE (ie significantly lower than before.) Second only change is enabling PBO+AutoOC for the CPU. Video descriptions have links to a CSV of the raw monitoring data from HWINFO. Also tried: - disabling one CCD of the CPU to see if inter-core communication was causing an issue (essentially making the CPU a 6-core). Thought it might also help it run at a higher boost clock by reducing thermals. Result: slightly reduced performance overall. Framerates didn't jump around as much, but the lows were even worse, now <30. Still sat at <60 doing legion/breaches. - Looked at memory latency/bandwidth. According to AIDA64, everything is working as expected (around 49 read/write bandwidth, 73ns latency). I'm also in the doubtful camp about memory timings/overclocking changing much. I guess the only remaining test I can do is SMT off to see what the core utilisation numbers look like. It's possible with SMT that the % number is misleading and there is a single-core bottleneck even when it's saying the max usage is 80%. Last edited by mawtle on Jul 27, 2019, 3:42:50 PM
| |
|
Can you make a test with vsync enabled ? Just to see if you drop below 60 when it's enabled.
|
