-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Poll: Ryzen 7950X3D, 7900X3D, 7800X3D
- Thread starter kindai
- Start date
Associate
My knowledge on DDR5 isn't quite where my DDR4 was, but if you're going 2x16GB just now make sure to get an a-die kit. I think pretty much everything 6200+ is going to be a-die, possibly even 6000+. The new m-die gen 2 is quite good looking in the newest 2x24GB kits, but IMO most fun is still going to be had with a-die and how tight some of its timings can get. We haven't quite reached a place where 48GB is needed as standard, 32GB is still holding in there (all the 2x24GB kits are m-die gen 2).
As for the binnings, from my experience so far it seems most of the a-die bins are quite homogenous. As in, you could buy one of the newest 8200 kits and someone with a 6000 kit gets the same results as you. Especially if you're sticking to 6000~6400 1:1. So it's really going to be silicon lottery it seems for timings rather than "just throw money at this bin and you'll win".
I nabbed this kit from OCUK as it's on sale right now, but I'd probably say just buy the 7200 kit and save a bit more money https://www.overclockers.co.uk/team...600c34-7200mhz-dual-channel-ki-my-0b5-tg.html
DDR5 looks like it's going to have an interesting next year or two in terms of its maturity and what is incoming, so I wouldn't recommend throwing lots of money at what is currently about (like going for an 8200 kit).
Then we've got the new line of AMD CPUs that could, hopefully, start hitting things like 6400 1:1 more consistently, let alone 6600.
Last edited:
Thanks! Can you turn off rgb on this one?
The biggest problem is that in 2 years we will be looking at ddr6 already
Caporegime
Yeah looks good. Did you buy it from OcuK?Should be fine
It's pretty difficult to kill memory, within reason with voltage/temp it's more instability that is your enemy as it will corrupt data.
IIRC there were some DDR4 kits sold at retail rated for 1.6v. My DDR4 daily was 1.55v with a 1.55v rated kit. There may be some scope to come down a little from 1.65v, but from what I've seen 6400 C26 is going to need around it to be stable.
@LtMatt My averages
Associate
Yup, both were. The first was a B-Grade 7950x3D they accepted a £469 offer on, but sadly its IMC is just rough even though it's a not bad chip at SP111. Second I bought new for the offer price at the moment, I think the 7950x3D's have dipped below £500 before but given I've gotten a chip that seems to have a decent IMC and SP rating it's not like paying £560 wasn't worth it. With the IMC lottery some people would pay a premium just to have something decent
Thanks! Can you turn off rgb on this one?
The biggest problem is that in 2 years we will be looking at ddr6 already
Yeah I turned it off with OpenRGB. I refuse to install ASUS software, it's literally a virus. ASUS Armory Crate should be branded as such. Probably adds like +10 to latency just having that crap running on your computer
EoL (or end of generation) RAM kits are where it is at though. That's when I bought my 4000C14 2x16GB b-die kit. Absolute monster, though, pricey! Going to sell it now and it'll likely recoup more than I paid for it, or at least get it all back, given I think it's the best bin of dual rank b-die that was ever released and not many went out.
Silly timings like this, though my daily ended up at 1.55v because the 1.63v above was simply to push tRFC below 110ns. Yeah, a drop from around 117~120ns to 109ns required a voltage bump of 0.08v
Last edited:
Associate
Because I am currently forced to use my mobile right now, imgur is of course refusing to upload lol, try this link instead.
Corpo Plaza
Corpo Plaza (or Corporate Plaza) is a sub-district of City Center in Night City. Corpo Plaza is the jewel of neopostmodern, neomilitaristic architecture, paradoxically owing its style to the tragedy that occurred in 2023 during the Fourth Corporate War. The terrorist attack on the former Arasaka...cyberpunk.fandom.com
This area, when driving under the glass circle during the day, is without a doubt the most demanding location in the game.
Update:
Here found a video of the area on YouTube.
Cyberpunk 2077 RTX 3090 5800X3D 32GB CPU Bottleneck test 4k 1080p
GPU: NVIDIA GeForce RTX 3090CPU: AMD Ryzen 7 5800X3D 8-Core ProcessorArbeitsspeicher: 32 GB RAM (31.92 GB RAM verwendbar)Aktuelle Auflösung: 3840 x 2160, 60H...youtu.be
Well guys I found it what is causing the low GPU usage, I did more testing after playing Helldivers 2 and after noticing frame drops and low GPU load.
I had even done a fresh install of Windows 11 yesterday with no luck... Turns out Windows "Virtualization-based security" was the issue....
I forgot to disable "Memory integrity" in Windows security. After disabling it, my FPS has increased by 40%!!! in Helldivers 2 when CPU bound.
As for Cyberpunk my frame rate has increased to 190fps in the same area I was getting 120-140fps. Making my performance the same as Hardware Unboxed's Phantom Liberty results when CPU bound with a 7800X3D...
https://youtu.be/7KZQVfO-1Vg?feature=shared&t=260 at 4:20.
Included a screenshot of the area I was testing in Cyberpunk 2077 Phantom Liberty.
imgur.com
Discover the magic of the internet at Imgur, a community powered entertainment destination. Lift your spirits with funny jokes, trending memes, entertaining gifs, inspiring stories, viral videos, and so much more from users.
imgur.com
My System
Windows 11
AMD 7950X3D
Corsair H115i ELITE CAPELLIX
Asus STRIX X670E-E
Corsair Vengeance 2x16GB @6000MHz 30-36-36-76 (Expo 2 Profile)
RTX 4090 FE
Corsair AX 1000W
Samsung 990 Pro 2TB NVMe
Last edited:
Hi all, after reading pretty much all this posts here and not finding the solution i would like to ask your help/sugestion in my case:
i have a 7950x3d with a kit hynix 64gb 6000mhz cl30 (30-36-36 76 1.4v) i had tuned curve to -18 all core and still dont get 5.7 on second ccd, i dont know why but all tests i had done from stock to curve optimizer the second ccd doesnt go much above the first...
any sugestions?
i have a 7950x3d with a kit hynix 64gb 6000mhz cl30 (30-36-36 76 1.4v) i had tuned curve to -18 all core and still dont get 5.7 on second ccd, i dont know why but all tests i had done from stock to curve optimizer the second ccd doesnt go much above the first...
any sugestions?
Associate
- Joined
- 30 Mar 2017
- Posts
- 857
Not 100% sure but isn't that 'normal' behavior that CCD1 usually boosts higher than CCD2?
I may be wrong but that is what I thought happened. CCD1 is the primary isn't it?
On the other hand I think lower clock speed can happen with slightly unstable voltages (i.e when dialing in curve optimizer or overclocking) , so would have to verify that I imagine.
Soldato
CCD1 is the X3D part. The 2nd CCD2 is the one with the higher boost
Caporegime
Interesting, thanks for the update. I can't say I've ever used Memory Integrity so explains why I never saw that.Well guys I found it what is causing the low GPU usage, I did more testing after playing Helldivers 2 and after noticing frame drops and low GPU load.
I had even done a fresh install of Windows 11 yesterday with no luck... Turns out Windows "Virtualization-based security" was the issue....
I forgot to disable "Memory integrity" in Windows security. After disabling it, my FPS has increased by 40%!!! in Helldivers 2 when CPU bound.
As for Cyberpunk my frame rate has increased to 190fps in the same area I was getting 120-140fps. Making my performance the same as Hardware Unboxed's Phantom Liberty results when CPU bound with a 7800X3D...
https://youtu.be/7KZQVfO-1Vg?feature=shared&t=260 at 4:20.
Included a screenshot of the area I was testing in Cyberpunk 2077 Phantom Liberty.
Max I seen in this area was 120-140fps. Now I can get 180-195fps. Almost a 50% increase at points....imgur.com
Discover the magic of the internet at Imgur, a community powered entertainment destination. Lift your spirits with funny jokes, trending memes, entertaining gifs, inspiring stories, viral videos, and so much more from users.
My System
Windows 11
AMD 7950X3D
Corsair H115i ELITE CAPELLIX
Asus STRIX X670E-E
Corsair Vengeance 2x16GB @6000MHz 30-36-36-76 (Expo 2 Profile)
RTX 4090 FE
Corsair AX 1000W
Samsung 990 Pro 2TB NVMe
Associate
Throwing this in here for anyone else that gets a 7950x3D and then maybe expects to begin with like -10~20 all core
The above is just testing potential curve offsets for my best cores, 2 on cache CCD and 2 on frequency CCD. The frequency cores, 8 & 9, showing -5 and -1. HYDRA doesn't run quite long enough with Prime95 to be called stable, but it's likely going to be thereabouts. Use Corecycler/Prime95/y-cruncher and all the usual goodies to actually test stability, not RyzenMaster or other "auto tuners" like HYDRA.
So yeah, if you stick a "baseline" curve in to begin your own testing and are immediately getting BSODs/issues, look to your best rated cores on the frequency side and maybe start them at like -5 lol. I just slapped in -15 all core to begin testing, couldn't get into Windows. -10 then booted but crashes almost instantly when beginning testing. You can now see how -10~15, especially on core 9, would immediately have caused issues in Windows/during boot. My best frequency core getting hit with requests and the curve value it should have is around -1.
Now for me to begin the very slow process of figuring out the rest and properly stability testing!
Good time of the year in the UK for the watercooling as well
"Why is the heating not on?"
"Because I'm doing benchmarks!*"
*and because of the bloody price of gas and electric right now
Last edited:
Associate
- Joined
- 30 Mar 2017
- Posts
- 857
Sweet water temp you got there. Is that running at full load? What's the setup and delta T?Throwing this in here for anyone else that gets a 7950x3D and then maybe expects to begin with like -10~20 all core
The above is just testing potential curve offsets for my best cores, 2 on cache CCD and 2 on frequency CCD. The frequency cores, 8 & 9, showing -5 and -1. HYDRA doesn't run quite long enough with Prime95 to be called stable, but it's likely going to be thereabouts. Use Corecycler/Prime95/y-cruncher and all the usual goodies to actually test stability, not RyzenMaster or other "auto tuners" like HYDRA.
So yeah, if you stick a "baseline" curve in to begin your own testing and are immediately getting BSODs/issues, look to your best rated cores on the frequency side and maybe start them at like -5 lol. I just slapped in -15 all core to begin testing, couldn't get into Windows. -10 then booted but crashes almost instantly when beginning testing. You can now see how -10~15, especially on core 9, would immediately have caused issues in Windows/during boot. My best frequency core getting hit with requests and the curve value it should have is around -1.
Now for me to begin the very slow process of figuring out the rest and properly stability testing!
Good time of the year in the UK for the watercooling as well
"Why is the heating not on?"
"Because I'm doing benchmarks!*"
*and because of the bloody price of gas and electric right now
Associate
Nah, that's pretty much idle hence how closely it's following ambient lol.
Doing some core testing just now but that basically tickles the loop because it's just focussing on one core at a time.
Case rear is the right side of a Lian Li O11 XL, so that's where the power supply and a vipers den of cables are stuffed
Associate
- Joined
- 30 Mar 2017
- Posts
- 857
Ah, I missed the ambient temp reading, very clever! Never really thought where the best place to put the header with the sensor is. Any ideas?Nah, that's pretty much idle hence how closely it's following ambient lol.
Doing some core testing just now but that basically tickles the loop because it's just focussing on one core at a time.
Case rear is the right side of a Lian Li O11 XL, so that's where the power supply and a vipers den of cables are stuffed
I never bother with those readings, but they are useful. I have an Aqua Computer Next Flow Meter which works pretty well.
I use a digital thermometer for a room temp reference and I am usually at around 4C Delta above ambient room temp at idle with x2 360 rads and a D5 pump, running a Ryzen 7900X and RTX 4090. In a Lian Li Evo Case.
I just upgraded to some Noctua NF-A12x25 PWM Chromax fans and Phanteks T30 fans and that completely changed the game.
Never expected fans to make such a massive difference, but I can run these on much higher speeds than my old Corsair ML120s and they are so quiet in comparison at higher speeds.
Last edited:
Associate
I have my ambient temp sensor wedged in at top intake
I run all intake on my O11 XL lol, cool air over rads is best, I have 4, 3 x 360 and 1 x 120.
I've got a custom front panel which replaced the glass one that has a design like the top piece to allow airflow out. Fans mounted behind it exhaust case air. Had to use the slim noctuas for that https://coldzero.eu/lian-li-pc-o11-dynamic-xl-front-grill-lines-clear
Rest of the case has the NF-A12x25's. 12 in total! Wasn't the cheapest
Last edited:
Associate
Tip for you all, don't even think about settling in with your new 7000 processors until you try to run something that can do Prime 95 HUGE FFT AVX2. After about a week of weird issues here and there and me spending hours working on a curve and getting contradictory results I believe I've found 1 or 2 of my cores on CCD1 fail stability testing at BIOS defaults.
OCCT settings.
Only way I can get the above to pass is by turning ON PBO and then adding a positive curve offset to Core 8 and 9. Test should pass without PBO on, let alone without a curve in use.
Back to OCUK it goes for a replacement
Caporegime
Hold on dude. Something is a bit off with that OCCT test as I’ve had five different X3d’s throw an error on that test. It’s the only test that throws these seemingly false errors. I thought I had a faulty CPU initially on an earlier 7850X3D, as it would fail on core 8/9 with everything stock, the I tried another 7950X3D, brand new, same error same cores (8/9). Tried a third one, same behaviour. at that point I stopped using OCCT and started using core cycler, y-cruncher etc.
Tip for you all, don't even think about settling in with your new 7000 processors until you try to run something that can do Prime 95 HUGE FFT AVX2. After about a week of weird issues here and there and me spending hours working on a curve and getting contradictory results I believe I've found 1 or 2 of my cores on CCD1 fail stability testing at BIOS defaults.
OCCT settings.
Only way I can get the above to pass is by turning ON PBO and then adding a positive curve offset to Core 8 and 9. Test should pass without PBO on, let alone without a curve in use.
Back to OCUK it goes for a replacement
If you get failures in core cycler and need to use +CO to fix then yeah maybe return it. But don’t judge it just on OCCT three second test as I think it creates false errors on 7950X3D for some reason.
Edit, let me add, I did not experience any system funnies like you tbf. These errors would occur even with +30 CO in OCCT bizarrely, even though the system was pure stock on
Everything and passed everything else.
Last edited:
Associate
Corecycler is the second screenshot running Prime95 HugeFFT.
Issue is I have been having games crash/reboots randomly and completely contradictory curve testing where one minute something is stable and next it isn't. I added a positive +15 to all cores and tests starting passing. I then narrowed it down to adding a positive +15 to core 9. Though core 8 seems iffy as well.
Wasted countless hours the past week just thinking the 7950x3D was a tough beast to tame only to finally narrow it down to some weird core behaviour on CCD1.
And yeah, gutted otherwise with the rating of this CPU and it having a good IMC.
Last edited: