9950X and 9950X3D is going to be terrific.

Not sure I get the joke
The first Pentium D chip (dual core) was just 2 Pentium 4's "glued" together to make a dual-core processor. The 2 "cores" on the die couldn't communicate with each other directly - they communicated via the Northbridge. They were horrible. This was back in 2005.
 
The first Pentium D chip (dual core) was just 2 Pentium 4's "glued" together to make a dual-core processor. The 2 "cores" on the die couldn't communicate with each other directly - they communicated via the Northbridge. They were horrible. This was back in 2005.
Ah, the Summer of the Turion 64.
 
Ah, the Summer of the Turion 64.

Now thats a name I havent heard in a long time. I got a laptop back then with a single core turion proc, it was pretty good once I took vista off it and put XP that it should have had. It didnt have nearly enough memory to run vista, only 1 gig which was really bad. I upgraded it to 2 gigs which was the max it could use, and with the return to XP it ran really well.

Later I ended up getting another similar laptop, with a bigger screen and an actual dual core turion that was supposed to have a bad screen. Turned out that someone had turned the backlight down to zero, and turning it back up it ended up working great.
 
Now thats a name I havent heard in a long time. I got a laptop back then with a single core turion proc, it was pretty good once I took vista off it and put XP that it should have had. It didnt have nearly enough memory to run vista, only 1 gig which was really bad. I upgraded it to 2 gigs which was the max it could use, and with the return to XP it ran really well.

Later I ended up getting another similar laptop, with a bigger screen and an actual dual core turion that was supposed to have a bad screen. Turned out that someone had turned the backlight down to zero, and turning it back up it ended up working great.
I took a turion 64 dual core to Iraq and it worked great, the next time I came back with a core 2 duo with a 7900gs go and that was a terrible mistake. Those were a long few weeks for a reliable in hot weather computer showed up again. We played bioshock, battlefield 2 and mods for it, fallout 3 and new vegas, Empire Total War, and some other things on a parade of laptops. I learned the hard way to buy 2 or 3 at a time with 16 and 24 port network switches. Laptops, fruits, socks and books were all I asked for from home.
 
Latest leaks out of China show Zen 5 as having up to 128 MB L3 cache, which is the same as the current 7950X3D. This strongly suggests that the top Zen 5 part will be structured the same way- one CCD with v-cache, one CCD without.
 
Latest leaks out of China show Zen 5 as having up to 128 MB L3 cache, which is the same as the current 7950X3D. This strongly suggests that the top Zen 5 part will be structured the same way- one CCD with v-cache, one CCD without.
Why can't they split it across both CCDs?
 
R24 ST:

3950x: 72 (2019)
5950x: 95 +32% (2020)
7950x: 125 +32% (2022)
9950x: 170 +36% (2024)


18-19% a year increase average (double every 4 years rate...), would be quite impressive
Damn that's like Nvidia's average GPU performance annual increase over that time frame. Would be really wild if AMD could match and/or maintain this cadence for CPUs.
 
Damn that's like Nvidia's average GPU performance annual increase over that time frame
Timing is a bit different and GPU generation sku way more changing and hard to match, but say from 2018->2022, NVIDIA went from a 445mm² 2060 to a smaller 380mm² 4080 that has +200/230% the performance (i.e. 3.3 to 1 ratio) or so despite being smaller, AMD st was about +55%/60% from 2019->2023, GPU are still probably moving faster (and their price tag).
 
Why can't they split it across both CCDs?
They can, but the effect would be diminished. In most games, there are one or two primary threads that are the bulk of the bottleneck. You want those one or two threads to get as much of the cache as possible so there's less pulling of data from RAM. If you split that among both CCDs, inter-CCD communication latency comes into play. It will also castrate the single CCD part as compared to the 7800X3D.
 
They can, but the effect would be diminished. In most games, there are one or two primary threads that are the bulk of the bottleneck. You want those one or two threads to get as much of the cache as possible so there's less pulling of data from RAM. If you split that among both CCDs, inter-CCD communication latency comes into play. It will also castrate the single CCD part as compared to the 7800X3D.
Gotcha. I guess I'll wait for the 9850X3D, or whatever it ends up being called, if I can squeeze one more year from my current rig.
 
I already have a AM5 Asus X670-P WiFi board. I'll just update the bios like normal and get a new cpu...yay. The integrated graphics on my 7900X shit the bed and I need something to replace it.

Do you guys know when they will first be in stores? I live 45 minutes from the Cambridge, MA Microcenter. Hurry up and sell these yo!
 
I already have a AM5 Asus X670-P WiFi board. I'll just update the bios like normal and get a new cpu...yay. The integrated graphics on my 7900X shit the bed and I need something to replace it.

Do you guys know when they will first be in stores? I live 45 minutes from the Cambridge, MA Microcenter. Hurry up and sell these yo!

They haven't even been officially announced... Nor even hinted at. All we have are people leaking otherwise confidential corporate info. Chances are these won't release until august or September.
 
Remind me; on modern motherboards, if you install a CPU that is newer than the BIOS can you at least power up the system with the 'unrecognized CPU' to update said BIOS?

I know way back when you could.
 
I already have a AM5 Asus X670-P WiFi board. I'll just update the bios like normal and get a new cpu...yay. The integrated graphics on my 7900X shit the bed and I need something to replace it.

Do you guys know when they will first be in stores? I live 45 minutes from the Cambridge, MA Microcenter. Hurry up and sell these yo!

Good to know AMD brought back their IGP on non APU desktop CPUs. Does AMD have an equivalent to quick sync or NVENC?
 
Remind me; on modern motherboards, if you install a CPU that is newer than the BIOS can you at least power up the system with the 'unrecognized CPU' to update said BIOS?

I know way back when you could.
No.

If you are worried about getting a board which does not have a compatible bios on it when you pull it out of the box, you have two options as a consumer:

1. AMD can send you a dummy CPU
2. buy a board with a feature often called "Bios Flashback" or something similar. With this feature, you plug the board into power. Then insert a thumb drive loaded with the bios file, into a specifically designated USB port. And then press a button on the back panel of the mobo to start the process. Be sure to read the directions, as it is often a highly specific process.
 
AMD mandated this feature to be on all AM5 motherboards and to my knowledge, it is, with a single exception. Asrock's B650E PG-ITX WiFi somehow got away without it.
That's what convinced me to jump to AM5 after building an AMD rig for my nephew, last month.
 
Back
Top