9900X is likely similar to 7900X which means it'd be among the worst (if not the worst) performing Ryzen SKUs in gaming. I wouldn't read too much into it personally. I think Daniel Owen got the right settings but the memory configuration and uCode could've been borked for all we know.9900X could be dragged down by inter CCX latency if the game uses more than 6 cores.
Thanks, I missed that.
The x900 SKUs are just useless.9900X could be dragged down by inter CCX latency if the game uses more than 6 cores.
I'm not sure, but even if it is scheduled on the wrong CCD for the 9950X3D isn't it still basically gonna be 9950X performance?For the 1 CCD parts, I think its pretty obvious that the X3D parts are tempting. But with the 2 CCD parts, I'm definitely gonna go with the one that does not rely on a flimsy thread prioritization routine, and is available earlier, the 9950X. I of course can't claim to know if the majority of potential buyers think like me though.
They did all benches with DDR5-6000 according to footnotes. They did the same nearly 2 years ago with ZEN4, so I wouldn't expect memory performance to be any better, otherwise they would've benched it with something better.Is there any indication it supports higher speed DDR5 than previous versions?
I'm not too optimistic on that.
Can anyone guess what has been improved on the motherboard to make that possible?
The x900 SKUs are just useless.
Maybe the mobo makers need time to get their UEFI code up to snuff. Makes sense if they got final CPU samples from AMD not too long ago.otherwise they would've benched it with something better.
Higher quality traces leading to the DIMM slots from the CPU socket? Some sort of shielding to reduce signal interference from the high frequencies? adroc may know better.Can anyone guess what has been improved on the motherboard to make that possible?
What to do with dies that have only 6-7 functional cores..?.
Beside if priced adequately they can offer a very good bang for the buck for people who are not primarly gamers, there s even some tasks where they should be very good given that they retain the full L3 cache, that s 33% more L3 per core than the 9700X/9950X, not counting that they have more bandwith/core in respect of the IOD BW.
I think AMD sells 10x if not higher quantity of x600x than x900x. And AMD also introduced non-x and 7500f parts. Which I think is enough for harvesting.
Next, they could introduce 7600x3d to sell more of the 6 core parts.
Depends, they are a good choice when running multiple virtual machines - without shelling out for an x950.The x900 SKUs are just useless.
It s a given that 6 cores sell quite more than 12C, but the margin is not the same, a 12C sell for 2x more money, so i would think that they prefer to make say 200$ out of 3 pieces of silicon (IOD + 2 CCD) rather than 100$ out of two (IOD + 1 CCD), hence one 12C get them more net income than three 6C since there s still the packaging and deliveries costs to account for.
Looking at the attractivity a 7950X is currently around 485€ in Germany while a 7900/7900X is at 335€ and the 7700X at 275€, so for whom want the throughput and is not willing to throw 485€ in a CPU that s a relevant offering, currently the 7900X sell almost as much as the 7950X, and after a price reduction the 7900X3D sold 80% more than the 7950X3D.
10 is for the ones sold in trail, the boxed ones are at 40 and 7950X boxed at 50,I am surprised such a large number of 7900x3d sold - 210. Maybe some special discount.
Only 10 of the 7900x sold. So the SKU can just as well be dropped.
BTW, buyers of these would probably switch to 9600x if the price is right
If you are thinking about buying shortly, it may be a good idea to wait for about 1-2 months and Zen 4 may be on sale. MicroCenter in the US has 7800x3d on sale for $319, down from $369 last week.
If we had no information other than Mindfactory numbers, you'd have to assume that AMD is the trillion dollar company and Nvidia/Intel are left fighting for the scraps.
I thought intel was eliminating the K skus? I could be wrong. Please confirm.The K models (unlocked multipliers) will have 125w TDP. The regular Arrow Lake CPU's will be 65w TDP. Zen 5 will be neck and neck with Arrow Lake with regards to power consumption. That's why 3nm would have given Zen 5 a decent edge in efficiency.
See belowApparently there’s frame gen turned on or something. Results seem too good to be true, because the results put the 7900XTX into 4090 territory I think.
Well, I will be on the upgrade train, but my train is not from Back to the Future, it will be a more slow walk. I have the 7900xtx. I just need now to get the ddr5, new mobo's cause why go with the 670e's? and a zen5. hopefully the 9800x3D will be out by the time my slow walk hits the stop sign. I will let y'all know.Now we need a hero in the forum with a 7900XTX and Driver Version 24.5.1 who can run that bench and post the numbers here. Hero where are you?
I haven't been following this for a while, so I might be missing something.
The zen5c dies look too long to just be 16 Zen5c cores unless it is a really different layout. Is there supposed to be something else on there, like more cache, an NPU, or something?
Me too. I think of two things: security patches and/or BIOS problems.So i'm still very much confused about the discrepancy between the 32% specint 2017 and the 16% figure directly from amd. Weird.
The former figure was given by a random internet poster. Oh boy...So i'm still very much confused about the discrepancy between the 32% specint 2017 and the 16% figure directly from amd. Weird.
I was assuming that it is still 2 CCX per die like it is in zen4c. Do you have something indicating that they changed that? It seems unlikely. In zen 4, the infinity fabric connection seemed to be along the long edge of the cpu die. Looking at the weird aspect ratio of Zen5c, it looks like it may be 2 CCX abutting on the short edge rather than the long edge. The infinity fabric interface may be on the edge towards the IO die or in the middle between 2 CCX. It would be along a shorter edge, so it would be a completely different layout from the regular zen5 die. Zen5c is a really long, narrow die; kind of reminds me of Zen 1, which essentially had two abutting along the short edge. It looks like they put a lot of work into the package routing. The long, narrow die looks like it helps with that.It's a single CCX, so while Zen4c had CCXs side by side, imagine what happens if you just double the number of cores in a single CCX.