EightyDuce

If you think you have a cooked board or ram, or both, I'd say return/exchange while you can. I just recently built a new system and turned out to have a borked 7700X ( bad MC) and a bad board (B650E-F strix) had a bad primary m.2 slot. Also, as for the PCIe speed, I mean check it with GPUz and make sure it's running at 16x rather than just being in the top-most slot (I assumed that's where it was installed). Doesnt happen often but the board can drop the PCIe speed if there is a hardware issue or possibly a conflict with shared PCIe lanes between slots and m.2 drives. Hopefully you get it sorted, I know it can be a pain sometimes to troubleshoot.

Is your primary PCIe slot running at 16x? Just to check make sure your motherboard isn't cooked in some way and you're not running on a 4X slot.

Some 7950X3D leakage. https://www.hd-tecnologia.com/segun-amd-el-ryzen-9-7950x3d-es-6-mas-rapido-que-el-i9-13900k-y-16-mas-rapido-que-el-ryzen-9-7950x-en-juegos/ Looks like about 6% better than 13900K (not KS) and 16% better than 7950X. Take with some salt, but appears to be a leaked AMD review guide.

So, I guess I sort of have an answer to my own question of why DDR5 timings, at least on AM5, are trash. Apparently they way it works is that boards take the JEDEC timings and scale them with RAM speed... So very much a crude approach which I suppose guarantees function, but at cost of performance.

Also it appears that hardware unboxed did a video using some of his settings. It would seem AM5 is sensitive to secondary and tertiary timings, even more so than primary timings. It also appears that memory manufacturers are using very relaxed timings beyond the primaries for whatever reason and thus leave a lot of performance on the table. Even EXPO profiles are effected as they aren't long enough to incorporate all of the timings which once again leaves performance on the table. Unknown how much this effects DCS, but honestly, there's no reason not to do this easy tune and grab up performance... Unless your memory can't handle them (which it should be able to no problem for Hynix), you're guaranteed a net performance gain.

Figured I'd throw this in here for those that have picked up some Hynix M or A die kits for AM5. Buildzoid "cookie cutter" settings for AM5 using memory with hynix chips. This is significantly better than any XMP/EXPO profile and is a very good baseline for doing further tuning....or you can just plug them in and call it a day, collecting free performance boost.

Gotcha. Yeah, many of us waiting on the X3D performance. Though at this point I'm fully vested in AM5, but benchmarks will dictate if I'll be staying with my 7700X or going X3D; whether it's the 7950X3D or 7800X3D. This is the kit I got, it's 299 now which is bonkers for 64gb. According to the manufacturing code they are hynix A die. G.SKILL Ripjaws S5 Series 64GB (2 x 32GB) 288-Pin PC RAM DDR5 6000 (PC5 48000) Desktop Memory Model F5-6000J3238G32GX2-RS5K https://www.newegg.com/g-skill-64gb/p/N82E16820374411?Item=N82E16820374411&Source=socialshare&cm_mmc=snc-social-_-sr-_-20-374-411-_-02112023

There are quite a few good 64gb kits, at least for AMD. I just grabbed a 64Gb DDR5 6000 CL32 for $349. Again that will depend if you want Intel or AMD.

Speed VS. latency is largely going to depend on whether you're on Intel or AMD. AMD sweet-spot is 6000 +-200, so you would want DDR5 6000 with CL28-32, for X3D chips it's less important as they try to keep things out of RAM so the benefit of lower latency RAM is less prominent. For Intel, you want as fast as you can afford (6800+). Now, you can score a fairly inexpensive kits of Hynix A-die (DDR5 6400-6600) and OC them to 7000+. That being said, the benefit of spending time to overclock/tune RAM is likely to yield negligible real-world noticeable performance (outside of benchmarks).

Since they announced that 79X0X3D would hit before 7800X3D, I've been kicking aorund the idea of getting a 7950X3D. For the most part I have no need for this many cores/threads (I did couple of years ago when I encoded my entire movie library, that was a chit show). But I'm starting to rationalize to myself that it may be a good idea to have best of both worlds; a vcache CCD for most games that take advantage of it and a high clocking non Vcache CCD for games that don't. My only concern, one that has been echoed by many, is the potential hardware scheduler issues which we won't know is valid or not until reviews hit.

Cache isn't allocated per core. All cores on the CCD have access to the same pool of cache. As for thermals, if there is any difference, it would be negligible (couple of degrees). The reason they down-clocked (less power, less heat) the vcache CCD is to stay within the thermal envelope that allows non cache CCD to hit higher speeds.

I think It's going to depend on chip yields and binning, but it's more likely than not, that 7900X3D will be 6+6 (8 with 2 cores disabled/dsfective/etc).

You heard it here first folks, if you have DDR4 CL16 RAM, good luck being able to play at 4K. Time to sell your 4090s and 7950x boys and girls, yall are doing it wrong.

Seems to has the same speculations as those swirling around since the anmoucment. Clock speed, "overclocking", hardware scheduling.... We don't know until people start leaking or reviews come out. My 7700X will still be in the return window when highend X3D chips hit, from that 7800X3D performance could be more readily extrapolated. At that point I should be in a better position to decide whether to keep the 7700X or send it back and wait for 7800X3D.

This. If you're buying B. die and running XMP, you're doing it wrong. He's paying 40% more money for 3-5% more performance. Yup. 64Gb DDR5 6000 CL32 hynix M die (can get it down to CL28) is $319 on the egg.

To be honest, I don't think someone your your situstion is the target audience for this type of upgrade unless you have money to burn. Realistically, you're probably looking at 15-20% improvement? Don't think it's worth it. Overclocking potential on these will be minimal/negligible. If you're building a new systems, upgrading an old system where 5800X3D is not available such as in my case coming from 7700K, it makes more sense.

The staggered release is annoying but not unexpected. High-end skus typically hit first in hopes of getting people to buy what they otherwise wouldn't. 7950X3D is going to sell well, especially at $699. I was really hoping for Feb 14 for 7800X3D.... But alas. Two more months as I continue to stare at the box of parts in my closet (case, rads, waterblocks, motherboard, ram, nvmes)

7900 and 7950X3D - Feb 28. 7800X3D - April 6

As much as the dude-man poster above wants it to be the case, you're not in some "bottleneck of hell?" bound by your RAM. He posts how he had a massive performance gain by going to CL14 DDR4, but in reality, vast majority of his gains come from 5600X to 5800X3D, where the latter is a gaming monster in comparison. Furthermore, contrary to his lack of understanding and regurgitating, of information, 5800X3D performance does not come from low latency RAM, but the masisve cache. This large low latency cache allows the engine to keep more of the low-level "instructions" on the chip and not put in the RAM. Long story short, take everything he copy-pastes into every thread with a large block of salt, as alot of his statements aren't rooted in reality or at best are a misrepresentation of information.

Don't know if I'm being optimistic, but I want to think FE would be $2K with AIB being $2200-2400. Titant variant if it sees the light of day, probably a $3k-3500, with a $4K Strix card lol. As for RTX5XXX, should be TSMC 3nm node, so we may see some good performance; power still probably not pushing past 600W. That being said, 4090 will last me until the 6 or 7 series comes haha.

... 4090ti is coming. Guru3d link similar info Clickbait-ish title (kinda), but some alleged specs leaked for a 4090Ti. Big takeaways, 96mb L2 and 24Gbps GDDR6X (significant bandwith increase). Which begs the question, why only 21Gbps GDDR6X on the 4090, other than sell you Ti later. So likely 10-15% over 4090. Curious to see the price.

Realistically, CAS latency primarily effects first word delay ie. 7.7ns for 3600CL14 vs 9.0/9.6ns 6000CL28/6200CL30. Once the delay is met, DDR5 is faster than lower latency DDR4 and at today's price it will do it for $50 less. What this translates to is virtually indistinguishable impact to the end-user outside of benchmarks. Current gen CPUs with arguably mid-range (6000CL32) are already matching and beating last gen CPUs on what you call "Optimized" aka 3200/3600CL14. At this point, unless you're on a seriously tight budget or staying on a DDR4 only platform , you'd have to do some serious mental gymnastics to justify DDR4 over DDR5.

For $700, I would grab it and run. Hell of a deal. You should be able to get acceptable VR performance. You may have to lower some eye candy, but it's not going to be an etch a sketch quality by any means.

Good to hear. Can your post your DCS settings?

You're right, haha. I initially said 2x faster than 3070 in my first post which was correct, then apparently fell asleep in my head. So yeah, 2x faster than 3070 and 1.5x3080.