There were a few sensational stories when the first couple of fried CPUs got posted, but the evidence clearly showed bad and forced socket insertion. Then ASRock seems to have been shoving too much volts through it in their awful BIOS implementations, but I guess users can also do this unwittingly since many boards allow you to push dangerous OCs and volts...
-
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.
Was going to create a separate thread, but figured I'd ask here... Is it worth me upgrading from a 7800x3d for my use case? I'm mostly playing mainstream games e.g. Exp33, KK2, Stalker 2, Space Marines 2 etc. and toying around running a 32b Q4 LLM, and some smaller financial models in Excel. My current specs are:
- 7800x3d
- 32gb HyperX DDR5 6000
- Zotac Rtx 5090 Solid
- Asus RoG Strix b650-A
- 1200w Asus Thor PSU
- Gigabyte m32u 4k 144hz
I normally would hold out for AM6, but 9800x3d prices have really come down lately and probably about as low as they'll go, and I really don't see me upgrading any other components for the next year or so.
- 7800x3d
- 32gb HyperX DDR5 6000
- Zotac Rtx 5090 Solid
- Asus RoG Strix b650-A
- 1200w Asus Thor PSU
- Gigabyte m32u 4k 144hz
I normally would hold out for AM6, but 9800x3d prices have really come down lately and probably about as low as they'll go, and I really don't see me upgrading any other components for the next year or so.
Last edited:
Yeah that would be my thinking, you already have vcache and would only really be getting a small ipc increase. Wait for zen 6
Associate
- Joined
- 11 Sep 2020
- Posts
- 1,446
- Location
- London
I made that upgrade as I was CPU bound in so many games trying to max my 240hz monitor. Now i've switched to a 5090, I am even more obviously CPU bound. Zen 6 X3D cannot come soon enough
Last edited:
Which games out of interest?I made that upgrade as I was CPU bound in so many games trying to max my 240hz monitor. Now i've switched to a 5090, I am even more obviously CPU bound. Zen 6 X3D cannot come soon enough
Soldato
- Joined
- 26 Apr 2004
- Posts
- 9,821
- Location
- Milton Keynes
I think the key point on whether Zen 6 X3D is worth the upgrade will be if they increase the size of the core complex, and if they improve the IMC/memory connectivity. From everything I've read and seen, the IPC for Zen 4->5 is a decent jump, but the MINIMUMS of the 7800X3D and 9800X3D are basically identical at the same clock speed, meaning the bottleneck is likely not the IPC or the core itself, but its connectivity outside the CPU, especially for regularly used data that does not fit in the cache, or does not make sense to be in the cache. Its also why non-X3D Zen 5 really didn't see much gaming improvement over Zen 4, as it was essentially IO/latency starved, which the X3D cache helped with a lot.
If Zen 6 can finally rework the IMC/IO die to really decrease latency and improve throughput, even if that is just physically moving it closer to the CCDs, that'll have a pretty good overall impact I think, and reduce the bottleneck there, even if the large level 3 cache does a good job of hiding it a lot of the time.
@LtMatt - I see you've got VERY high frequency RAM (whilst still low latency) compared to the vast majority of people running Ryzen 9***, how much improvement (ballparked) did you find you got?
At what point roughly did you find memory clock speed overcame breaking the 1:1 rule.
If Zen 6 can finally rework the IMC/IO die to really decrease latency and improve throughput, even if that is just physically moving it closer to the CCDs, that'll have a pretty good overall impact I think, and reduce the bottleneck there, even if the large level 3 cache does a good job of hiding it a lot of the time.
@LtMatt - I see you've got VERY high frequency RAM (whilst still low latency) compared to the vast majority of people running Ryzen 9***, how much improvement (ballparked) did you find you got?
At what point roughly did you find memory clock speed overcame breaking the 1:1 rule.
Last edited:
Soldato
- Joined
- 2 Jan 2012
- Posts
- 12,426
- Location
- UK.
Agreed, not worth the cost, could be worth waiting for something better. The 10800X3D rumored spec may be a far more worthy upgrade, 12 cores and a more streamlined design.
Soldato
- Joined
- 26 Apr 2004
- Posts
- 9,821
- Location
- Milton Keynes
As a side aside, think I settled on a tune for my ram. I did try seeing if I could get the timings lower, but the RAM already gets warmer than I feel comfortable when the RAM is stress tested so don't want to push more voltage (albeit it's passed hours of YCruncher VT3, OCCT's memtest, 24 hours of Karhu, and multiple hours of TM5 Absolut Anta777 profile), so have one of those little clip on ram coolers on route just to lower temps a bit and ensure long term stability.
Edit: I might try tREFI 65535 ONCE the cooler is fitted, but I don't want to do that without the RAM cooler.
Edit: I might try tREFI 65535 ONCE the cooler is fitted, but I don't want to do that without the RAM cooler.
Last edited:
Soldato
- Joined
- 26 Apr 2004
- Posts
- 9,821
- Location
- Milton Keynes
OK Ram cooler is now fitted
If this ram config turns out to be stable, I'll probably stick here, as ultimately I'm after best gaming performance and 0.1/1% lows, not benchmark records, and the primary timings don't want to push lower.
Timings currently being tested. Its a 64GB Dual Rank kit, so I'm pretty happy with this if stable; given how far it'll go, I'd be very surprised if it wouldn't also clock quite high with the right combo of timings and voltages, but no real need for me to push that right now.
Purdy!: Unfortunately due to the size of the Artic pump head, and all the guff on the mobo, its at a very slight angle, but looks nice, and hides some of the few remaining cables!
If this ram config turns out to be stable, I'll probably stick here, as ultimately I'm after best gaming performance and 0.1/1% lows, not benchmark records, and the primary timings don't want to push lower.
Timings currently being tested. Its a 64GB Dual Rank kit, so I'm pretty happy with this if stable; given how far it'll go, I'd be very surprised if it wouldn't also clock quite high with the right combo of timings and voltages, but no real need for me to push that right now.
Purdy!: Unfortunately due to the size of the Artic pump head, and all the guff on the mobo, its at a very slight angle, but looks nice, and hides some of the few remaining cables!
Last edited:
Which RAM cooler did you go with in the end? I'm running a 9800X3D build in an Ncase M2 and a 265k build in a Fractal North and was definitely a little wary of the RAM temps on my 265k build - just because I'm pushing them over 8000mhz at CL38 currently and whilst stable, occasionally they can look a little toasty!OK Ram cooler is now fitted
If this ram config turns out to be stable, I'll probably stick here, as ultimately I'm after best gaming performance and 0.1/1% lows, not benchmark records, and the primary timings don't want to push lower.
Timings currently being tested. Its a 64GB Dual Rank kit, so I'm pretty happy with this if stable; given how far it'll go, I'd be very surprised if it wouldn't also clock quite high with the right combo of timings and voltages, but no real need for me to push that right now.
Purdy!: Unfortunately due to the size of the Artic pump head, and all the guff on the mobo, its at a very slight angle, but looks nice, and hides some of the few remaining cables!
Appreciate it's off topic too - but how's the North XL? I'm really tempted to swap from the non-XL to it, just for the top mounted 360mm AIO support.
Soldato
- Joined
- 26 Apr 2004
- Posts
- 9,821
- Location
- Milton Keynes
Well I'm calling that stable - overnight TM5 absolut, OCCT's default 1 hr, several hours of Y-Cruncher VT3, and over 24 hours of Karhu, error free
@t2na
Just one of the generic most common sort on the rainforest, mine was under the 'brand' Hopbucan, but a load of sellers were selling them, I just happened to pick that one due to prime meaning it'd arrive quickly.
It basically slots round the ram clips and then just connects to argb and fan connectors and away you go. Nice and simple, not too noisy, ram dropped by a good 5-8 degrees, which Im happy with.
Ill be honest having looked back it looks like Corsair was selling these at one point with some extra trim, but looks like fundamentally the same oem design.
Re the XL, I really like it, although I've got the RC variant, which only has room for a 120mm rear fan, but generally really like it. Its a nice, modular case, and unlike most, has a 20gbps front usb c.
Last edited:
Soldato
- Joined
- 26 Apr 2004
- Posts
- 9,821
- Location
- Milton Keynes
I was looking back at some older posts from last year, and I noticed that some people were getting drastically lower AIDA64 latency results, even with similar or worse RAM timings.
Has something changed in AIDA64 or in the AGESA since last year? I just want to make sure before I save my profile and leave tinkering alone, I'm not leaving considerable latency/gaming performance on the table!
Edit:
OK ignore me, I seem to have worked it out - seems to be these "Core Tuning Config" settings that have been introduced in more recent BIOS.
Auto/enabled CAN result in slightly higher average FPS due to more advanced pre-fetchers/guidance, however LEGACY results in lower memory latency results and can result in higher minimums, looking at the testing I can see. Auto seems to opt for Level 2 on my motherboard.
I looked at the results I can find, and whilst I'll have to test it myself, it immediately dropped my memory latency result to sub-65ns using the legacy vs level1/2/auto settings. I mean it varies as always, but just got my best yet when rerunning at 63.4ns.
Running some quick benchmarks using Cyberpunk, I note that Level 1 and Legacy seem to get marginally better results than Level 2/Auto.
Anyone here played with this setting and found an optimal setting; or any specific outlier cases where setting it one way or another made a significant difference.
Has something changed in AIDA64 or in the AGESA since last year? I just want to make sure before I save my profile and leave tinkering alone, I'm not leaving considerable latency/gaming performance on the table!
Edit:
OK ignore me, I seem to have worked it out - seems to be these "Core Tuning Config" settings that have been introduced in more recent BIOS.
Auto/enabled CAN result in slightly higher average FPS due to more advanced pre-fetchers/guidance, however LEGACY results in lower memory latency results and can result in higher minimums, looking at the testing I can see. Auto seems to opt for Level 2 on my motherboard.
I looked at the results I can find, and whilst I'll have to test it myself, it immediately dropped my memory latency result to sub-65ns using the legacy vs level1/2/auto settings. I mean it varies as always, but just got my best yet when rerunning at 63.4ns.
Running some quick benchmarks using Cyberpunk, I note that Level 1 and Legacy seem to get marginally better results than Level 2/Auto.
Anyone here played with this setting and found an optimal setting; or any specific outlier cases where setting it one way or another made a significant difference.
Last edited:
Hi all, relatively new AMD user (Intel CPU's for last decade), had a couple questions for fine tuning a new setup.
New components:
CPU - 9800x3D
RAM - 64GB (2 x 32GB) Corsair Dominator Titanium - 6000 CL30
GPU - 4090
MB - ASUS ROG Crosshair X870E Hero
Everything is running fine just out of the box without tweaking BIOS etc but given as I'm new to AMD, I was wondering if there's any specific settings etc I should be turning on within the BIO (or Windows), tweaking of RAM speeds etc?
Generic post I know but thought I would start at a high level and work down through the details from there haha.
New components:
CPU - 9800x3D
RAM - 64GB (2 x 32GB) Corsair Dominator Titanium - 6000 CL30
GPU - 4090
MB - ASUS ROG Crosshair X870E Hero
Everything is running fine just out of the box without tweaking BIOS etc but given as I'm new to AMD, I was wondering if there's any specific settings etc I should be turning on within the BIO (or Windows), tweaking of RAM speeds etc?
Generic post I know but thought I would start at a high level and work down through the details from there haha.
Last edited:
There are some handy guides on YouTube. I like this one as a starter - then I made further tweaks myself from here:
Basically, use curve optimiser (under AMD Overclocking, not in AI Tweaker) in BIOS and run a negative curve. Every chip is different so yours could be stable anywhere from -10 to -30. I usually run Aida64 & Cinebench R24 to check stability cos you can roughly know in 30 minutes if it's not viable.
Start at -10/-15 and keep working up until it becomes unstable, then dial it back a notch. I'm running mine at -25 and it makes a lovely difference to how hot the chip actually gets and a decent bump in performance too.
For RAM, you'll have an option called EXPO Tweaked in BIOS for that Asus mobo. I'd suggest running benchmarks at stock, then EXPO - then EXPO tweaked so see what your best results are.
I'm running EXPO Tweaked in my Asus board + 9800X3D and multi-core score on CB24 is 1397 on tweaked vs. 1357 EXPO 2.
The video guides are a useful place to start though, it was an odd one to get my head round with AMD where you basically make most of your gains with a negative curve vs. on my 265k build where I'm pushing timings on ring clock or cpu clock to gain improvement.