AMD Zen 2 + Radeon RX 5700 Series For Linux Expectations

[oiaohm]

Calling the L3 cache "gamecache" is so dumb. I hate these marketing guys sometimes.

https://www.anandtech.com/show/14407/amd-ryzen-3000-announced-five-cpus-12-cores-for-499-up-to-46-ghz-pcie-40-coming-77

No gamecache is not L3. L3 per chiplet is 32 Megs. L2 is 0.5 of a meg per core active in the chiplet.
So 70 Meg example is 64 Megs of L3 and 6 megs of L2.

Gamecache and Smartcache are both basically totalling. AMD does at least provide L2 and L3 cache information if you want it.

Intel ark.intel.com just gives you Smartcache and leaves out what is L3 and L2.

[LeJimster]

https://www.anandtech.com/show/14407/amd-ryzen-3000-announced-five-cpus-12-cores-for-499-up-to-46-ghz-pcie-40-coming-77

No gamecache is not L3. L3 per chiplet is 32 Megs. L2 is 0.5 of a meg per core active in the chiplet.
So 70 Meg example is 64 Megs of L3 and 6 megs of L2.

Gamecache and Smartcache are both basically totalling. AMD does at least provide L2 and L3 cache information if you want it.

Intel ark.intel.com just gives you Smartcache and leaves out what is L3 and L2.

Ok, I guess you're technically correct. We get they've doubled the L3 cache it to help reduce memory latency and therefore help framerates. But the term "gamecache" is still dumb, as if the cache only exists for game workloads.

[sophisticles]

I personally think people are going to be very disappointed with Zen2, there's been a lot of build up and hype, but everything I have seen points to the 8C/16T offerings being slightly faster than existing AMD processors on a clock-for-clock basis, I think the 12C/24T and 16C/24T offerings will be slower over-all than the current Threadripper lineup due using 2 channels for memory instead of 4, though they may be a bit faster in certain cache dependent workloads or possibly AVX related workloads, assuming AMD got their act together and beefed up their SIMD units.

I think AMD will enjoy about 2 months of "glory" with this release and then Intel will smack them right back to reality with either Ice Lake or Comet Lake.

Navi on the other hand looks very interesting, especially from a video perspective, assuming of course that ISV's impliment the new features in a timely manner.

[starshipeleven]

Also, minor rambling: Can you stop being all about "gaming"? I want to work, compile, and sometimes leave my computer on overnight! Whenever I see "gaming" I always think "oh, this thing is not 100%-reliable and at some point it will fail when not used for something that is gaming (where people just sit in a computer and play a game for like a few hours then turn it off and do something else but sadly it is not what I do all the time (yes I do play games at times but not like aaaallllllll-the-time...))"

Your definition of "gaming" is wrong, how did you come up with this bullshit.

Gaming boards are more durable than "office" ones as they are designed to withstand overclocking (i.e. being stable when pushed to the limits, with insane CPU power draw), high-end cards that at times may draw more than the PCIe spec mandates from the PCIe slots.

Also gaming boards and cards in general have decent heatsinks (or at least a better shot at it).

so basically are you saying that a "creator" is different from a workstation user?

Yes but that's not what you think. Seriously what the fuck made you think that "gaming" hardware is unstable.

The difference is in computing (and certification) power requirements.
A "creator" is usually fine with high-end CPUs and GPUs, a "workstation user" needs top-of-the-line CPUs and top-of-the-line GPUs, sometimes more than one such GPU.

[starshipeleven]

As long as it comes with support for Windows 7 and Linux, I'm happy about it.

FYI: windows 7 will EOL next year, and within a couple more years they will stop supporting it. So prepare to jump ship.

[jrch2k8]

I personally think people are going to be very disappointed with Zen2, there's been a lot of build up and hype, but everything I have seen points to the 8C/16T offerings being slightly faster than existing AMD processors on a clock-for-clock basis, I think the 12C/24T and 16C/24T offerings will be slower over-all than the current Threadripper lineup due using 2 channels for memory instead of 4, though they may be a bit faster in certain cache dependent workloads or possibly AVX related workloads, assuming AMD got their act together and beefed up their SIMD units.

I think AMD will enjoy about 2 months of "glory" with this release and then Intel will smack them right back to reality with either Ice Lake or Comet Lake.

Navi on the other hand looks very interesting, especially from a video perspective, assuming of course that ISV's impliment the new features in a timely manner.

we will have to wait until NDA are off but seeing the technical specs and the changes i strongly disagree and i suspect the 3900x will be at least as fast as the 2950x in productivity task and will trounce it at gaming/single threaded tasks. in the same sense i suspect the 3950x will hold the crown for a good chunk of 2019 if not part of 2020 as well but i expect at the very least it will compete neck to neck with the 7980xe under heavy overclock.

Remember one of the biggest issues performance issues with Zen and Zen+ was the CCX's having to communicate through infinity fabric through each other to access ram and that every SIMD operation needed 2 cycles(because the hardware could only do 128bits at once), that is why Threadripper with Quad Channel had an advantage in many scenarios but Zen2 added proper 256bits and the I/O chiplet should allow every CCX to access RAM with the same latency as the next, so Zen2 just need channels for bandwidth not access speed.

Threadripper 3000 series will be monsters if all the data AMD published is 100% accurate but as i said lets wait for the reviews to be sure since in theory everything is dandy

[Anty]

Calling the L3 cache "gamecache" is so dumb. I hate these marketing guys sometimes.

you are not alone

https://www.reddit.com/r/Amd/comments/bz8ngb/petition_against_gamecache/

best comment

4.7GHz PlayBoost™, 32K 8-way LootCache™, 512K TeamCache, 64MB GameCache™, 4K MicroHeadshotOpsCache™ with up to 8 fused GameOps™ going straight into the WinQueue® every round.

A massive 256bit wide FearPoints® calculator to boost benchmark scores, coupled with no less than double the GameLoad™/KillStore™-bandwidth (up from 16b to 32b!) really wins it for you.

Are you a streamer? Then check out the doubled bandwidth on the PlayCoolInterface-4.0 connection allowing you to stream games twice as fast into the trashcan on your SuperStreamDrive over NoobVictoryMachineEnchancement2™, now with added LEDs!

Needs fast memory? You've got it! The 67ns latency to the RGB MemeAccelerator™ running at up to 5130000 memes/sec (3200000 memes/sec stock speed guaranteed) provides you with enough speed to download any part of the Interwebz (including cat videos).

Turn it on by pressing the I-Win button on the front of the GameBox.

[LeJimster]

you are not alone

https://www.reddit.com/r/Amd/comments/bz8ngb/petition_against_gamecache/

best comment

lol.

And now we have nvidia with "Super" line up - What next? AMD responds with the SuperDuper series?

[schmidtbag]

I also rolled my eyes at GameCache. It might be acceptable in an APU, but c'mon... this is just cheesy.
That being said though, I do hope the APUs get these big caches. They desperately need it.

> Pricing on the Ryzen 9 3950X isn't yet available

$749 per the presentation

That price was a bit higher than I was expecting. Still nearly $500 cheaper than the 7920X (MSRP) but not exactly a bargain either. I actually see that as a good thing though - usually AMD prices their products based on competition. Intel has no competitor in this performance tier (the 7920X is a tier higher), so, $750 could be a sign of a pretty good product.

[skeevy420]

A "creator" is usually fine with high-end CPUs and GPUs, a "workstation user" needs top-of-the-line CPUs and top-of-the-line GPUs, sometimes more than one such GPU.

I'd like a workstation that supported multiple APUs so a workstation user wouldn't need multiple GPUs to fire up a VM for pass through. I imagine that would be a decent testbed for OS maintainers & developers, UI designers, etc when combined with NUMA, etc...or even office environments where two or more people could share the same system in a multi-seat setup since all we'd need is a monitor and USB hub for each APU.

Years back, I assumed we'd have something like that by now from either Intel or AMD. I'm a little sad that we don't.

I often wonder if pressure from game console manufacturers is why we don't have good APUs yet. I really doubt that MS or Sony wants an APU on the market that beats the PS4 or XB1 @1080p for less than the cost of the consoles.

Dear AMD,

If you're gonna do something with HBM this year, please do it with APUs and not GPUs. An APU designed around 1080p gaming or 2K-4K desktop work is better suited to deal with the 4GB HBM limit than gaming or professional GPUs.

Also, your model numbers really suck right now. It's hard to tell if someone is talking about the RX 5700 or if they typo-ed RX 570 or if they're talking about motherboards. How frickin' hard would it have been to call it an RX 670? It's like y'all hired Slackware's marketing team for this launch. Please, please don't tell me this is a pecker wiggling contest with Nvidia over who has the highest model number since only morons think stupid thoughts like "5700 is a bigger number than 2080 therefore it must be better...oh snap, there's the HD7970, it's HD with a bigger number so it has to be better ".