Most of us just don’t care about the season theme this time around. We need the itemization and especially the aspects to actually drop, meaning the bottleneck on the ancestrals has to go. The devs could have kept ancestrals as the progression item while bringing back sacred items and having sacred top out at the same item power as normal items, but able to roll the higher tier of aspects that people are currently unable to get this season.
Unless they fix that, among a few other choices they flip flopped on going from S5 to S6, I couldn’t give a rat’s petootie about S7 and I suspect a great many players feel the same way. What good is a new season theme if the core parts of the game are so abysmally unrewarding that it all just doesn’t matter?
We’ve got a wooden rocking chair underneath the stairs here in our house. It has a solitary cushion on it, and atop that cushion sits an old china doll that’s around a hundred years old. My grandpa wants to give it to my cousin but I want it because it reminds me of my great grandparents and that cousin already took their piano when they passed away (something I wanted as I actually play it where my cousin does not). She got her keepsake, so I want mine since I didn’t get anything from my great grandparents’ house except a clock radio because nobody gave any thought to what I might have wanted.
Because of x3d, I actually suspect that what WoW really likes is low memory latency.
But the theory of using ARM is that with the increased power efficiency, won’t have to downclock to avoid bouncing off thermal constraints, thus enabling higher performance.
I don’t think this is actually the case. I know it was rumored to be true, but looking at the official specs sheet, the 9950x3d and the 7950x3d have the same cache.
But the reason for the driver is cross-CCD cache hits, which can reduce performances by effectively negating most of the benefits of the extra cache, which would still apply if both caches had the extra cache.
None of the x64 CPUs are thermally throttling unless their cooling is insufficient, so they still rule the roost when it comes to IPC. And ARM gets hot under load. You need only look at what happens with Apple’s machines. The default fan curve lets the machine cook itself to death. Apple’s Macs won’t even turn the fan on until the CPU hits 90°C. And once it does that, the fans turn off again once below that temp, only to come back on, but by the time the fans can start cooling things down again, it’s too late - the machine thermally throttles. It’s a back and forth nightmare. And Apple went a step further - as of a specific OS update (early Sequoia I believe), they issued a firmware update that blocks modifications to the fan curve until the fans come on for the first time, meaning no matter what, the machines will hit the 90ºC threshold at least once, and if the user hasn’t installed something like MacsFanControl or iStat Menus and set up a much more aggressive curve, it just bounces up and down. ARM is only more efficient - it isn’t cooler running and still needs adequate cooling out of the gate or it too throttles.
It is actually the case. Only one of the two CCDs prior to the 9xxx series had the 3D cache at all. If the OS was installed prior to the user upgrading to one of the X3D CPUs, it would often put high demand threads on the non-3D cached CCD, drastically reducing performance. You can get around this by going through a rather cumbersome set of steps to “refresh” the OS, but the simplest way to avoid the mess was a clean OS install with those X3D CPUs. But with the 9xxx series both CCDs have the same 3D cache and it no longer matters whether you do a fresh OS install or not since either CCD will offer the same performance. AMD also flipped the physical layout - the cache and CCD have been switched, with the cache sitting below the CCD, letting the CCD make direct contact with the IHS to dissipate heat better. Cross CCD cache misses are a thing of the past now that they each have their own cache pool. Once the driver determines which CCD has the “priority cores” (the ones that operate more stably at higher clock rates), that is what determines where threads are put and which CCD has parked cores.
I love old furniture. I also have my Great Grandfather’s rocking chair. My uncle was TICKED when my grandmother gave it to me. It was perfect when my kids were babies, although we had to have it re-upholstered as it was stuffed with straw, which was beginning to poke out. My mother also has my great grandmother’s hall tree, which I will get some day.
It’s amazing to me that some of these pieces are now over 100 years old, and functioning just fine. While the stools I purchased 5 years ago need to be replaced. There’s no saving them.
My old people brain can’t read the language you are speaking. It makes my head hurt. My computer works, or my husband fixes it. That’s my computer skills.
But how does the 9950x3d have 2 extra 64MB 3dvcache dies in addition to the base 64MB of cache from the 9950x and only have 128MB of L3 cache per AMD’s website?
Don’t feel too bad. I’ve been doing tech support of varying types for fourty years, starting with the Commodore PET computers back in the 2nd grade.
The 9950X3D has 64 MB L3 cache per CCD. It isn’t in addition to what you’re thinking of as the “base” L3. The non-3XD has a two CCDs with shared L3 and the 99xxX3D CPUs each have 64 MB L3 per CCD. The 9800X3D uses a single CCD with 3D cache. The 9950X (non-3D) has 64 MB L3.
The MacBook Airs are terribad for gaming unless you put them in a freezer (I’m only half joking there) since they have no fans at all. The Airs throttle left and right after only a short time under load.
You’re right about there being two sets of cache - I missed one of the lines in the description, but it isn’t quite the way you think. The 32 MB portion is on both CCDs and is not 3D cache and is built in to the baseline portion of the Zen 5 core complex (each 6/8 cores on each CCD has 32 MB inline L3). Each CCD does in fact have its own 64 MB of dedicated 3D V-cache. AMD’s marketing probably avoided adding that line on the site itself because, as you may have noticed, it causes confusion. The reason for the 96/32 split would be that only one CCD is accessed at a time with regard to the non-3D cache.
The reason for 3D cache on both CCDs is to eliminate the issue of having the OS put threads on a non-3D cache enabled CCD like in the previous X3D lineup (if you didn’t do as noted above to properly prepare the OS first).
It’s a weird setup, but in the end in simplistic terms, it goes VROOM. I find it both hilarious and pretty sad at the same time that Intel’s Arrow Lake performs worse than Zen 5 even when paired with CU-DIMMs. That’s how much having both CCDs with 3D cache makes a difference. FWIW, even reading the Zen 5 architecture details can be a daunting task. AMD certainly did some shuffling this time around. The biggest downside to AMD is that because their CPUs don’t utilize CU-DIMMs (yet), they’re still very finnicky with anything faster than DDR5 6000. That’s another reason for the 3D cache on both CCDs. I’m amazed they reduced the cycle latency so much on the v-cache relative to Zen 4. 3.5 cycles doesn’t look like much at first glance, but it adds up in highly repetitive branch prediction scenarios (e.g., games).
I’m actually beginning to see why it took AMD so long to get Zen 5 out the door. The engineering had to be insane. Intel on the other hand, did a slight change to its architecture, killed SMT, reduced max boost clocks, and even with the new CU-DIMMs is a distant second place, and it really hurts games, a lot. Sadly I’m going to need to lean on Arrow Lake when upgrading my dentist’s computers because there’s not a chance in hell I’m risking mission critical systems, some of which have to remain online 24/7, to Intel’s 13th/14th gen nightmare and OpenDental apparently has at least some issues with AMD as they recommend Intel for full compatibility (or it could just be they didn’t want to buy AMD stuff to vet…who knows).
In any case, AMD marketing and under the hood show two different things. I wouldn’t recommend Intel to anyone except for compatibility reasons in the current gen.
Talk about a blast from the past … decided to fire up my EA launcher and see if I could load some old saves of Dragon Age: Origins. Took some work but finally found some old character backups. My last saves were from 2014.
Took a bit longer to reacquaint myself with the controls, lol.
I love that game and missed all the sassy chatter between party members.
Yeah, it would have been up earlier but they encountered an issue with Plunderstorm. Hopefully, we’ll see that resolved and that mode can be brought up later today.
I was going nuts trying to find out why fast start wasn’t an option shown to me in the advanced power settings and then it finally dawned on me - it isn’t shown because hibernation isn’t enabled. Back when this system ran Windows 7 I disabled that because desktops don’t need it and I prefer clean boots on cold boots and warm restarts. I never noticed anything under Windows 10 because this is now the default behaviour if not on a laptop or other battery powered device (like my dentist’s iTero machine).
It’s comforting to know that I had the foresight to make things less dirty from the beginning.
