We used to use Intel’s hottest CPUs for compiling some of the CAD stuff we got from engineers into the Unreal models we would publish. They got so huge that the Intel’s would crash after 16 hours and never complete the generation of the models. We switched to a system I custom built with AMD Threadrippers (just the 24 core) and we cut it down to 8 hours to full completion. If these were Epyc’s we would be done well before lunch time.
There’s something to be said about AMD’s instruction set and earth-shattering power.
I’d like to see Epyc’s get tried out. Intel seems to be playing catch-up with AMD. The specs on intel look good on paper but sometimes it’s not about a drag race, where speed matters but a tractor pull, where power reigns supreme.
It would be much appreciated if these timers (or ‘worst case timers’) are made visible front and center of the screen with some countdown timer or bar or circle, w.e. This prevents a lot of unnecessary clicking to then get a ‘wait’ message (which isn’t a pleasant user experience) . Right now the game is essentially putting you ‘on hold’ before you can do certain things without you knowing (until you click to do something etc.).
I wouldn’t mind being in one (as long as ships are being supplied instead of being asked to buy one). Lots of people say locking on a target and pressing f1 is stupid. In smaller fights, I need to worry about turret transversal and sometimes contribute nothing. In a cap chain if I fail to fix the chain when someone gets scrammed and fall behind, I am deadweight again. I can’t fly a cyno ship, but if I warped on grid 3 minutes before a fight uncloaked because of fat finger (like what PAPI did once… I assume it was fat finger anyways) and then cloak afterwards, then once again I’d be useless unless I was a bait. In small fights, when I forget to rinse gate gun fire, my ship is a wreck and I did nothing. In large fleet battles, I can press f1 and be confident I am contributing towards our victory. Or our client’s victory. I can’t even remember the last time we fought for ourselves in a context with large fleet battles. Other than those 3 losses, and usually we just don’t fight giant ones for ourselfs period.
Its obvious why they upgraded to intel 3rd gen, because 4th gen procs have paper launched and haven’t been validated by server vendors yet. A business does not care about the ego of people, it cares about continuity. There is a reason why it takes dell, lenovo, hp and cisco at least several months to put new tech into their product line: they have to write drivers, firmware and software for it as well as do thorough testing to ensure a minimum degree of stability and security.
The 6334 is the second fastest base clock CPU intel currently has. The only one faster is a platinum that is nominally faster but much more expensive. The idea that more cores is better than higher clock speed is just wrong. nearly all workloads are a combination of lightly parallel and single threaded operations. workloads containing zero sequential operations are extremely rare and higher clock is always more critical than more cores.
I have only found AMD database servers to be unmitigated disasters of AMD processor vintages of the last ~15 years even without getting into applications that are memory intensive enough to involve the most vulgur database swearword I know: NUMA.
Even in x86 there’s more to be said about. And EVE is in the deal proper. Being a physics simulation engine by nature, it may benefit well from special instruction sets, like various SSE, AVX et cetera.
I did and the problem returned instantly. I’ve been doing that trick between characters (way too) many times so perhaps somewhere along the way something went wrong. I did export my Overview and imported it to the fresh one, which worked just fine.
The UI itself took some 20 minutes to set up (mine’s not exactly standard) mostly due to having to set all the key binds again. Easy solve.
Has CCP considered hyper-converged infrastructure as a replacement for traditional SAN architecture?
I was a SAN administrator for 20+ years (mostly NetApp\Cisco\Dell) and switched to hyper-converged (Nutanix, VXRail) about 5 years ago. I doubted the sales pitches, but the cost savings were a priority at the time. I was surprised to see the performance improvements. The I/O was off the charts…
As you mentioned about “continuity”, almost all IT companies upgrade their hardware every 5-6 years. Therefore, we shouldn’t expect any upgrade until 2027 for “Rank-and-File” machines. Now let me do a comparison:
In 3 years they moved from Gen 2 to Gen 4 CPUs. So, 2nd Gen Xeon looks imbalanced, 3rd Gen - “okeish” and 4th Gen - just “wild!”. Server isn’t only the speed of CPU or RAM size, but also the overall data exchange speed between them and the external network. The only marketing cheating trick here is Intel7 mark which means the same 10 nm Tech built based on 7 nm Tech concepts. Intel has big issues switching to sub 10 nm Tech. In my opinion CCP could wait for a year or a year and half and then upgrade to that 4th Gen machine. In my opinion there shouldn’t be any serious compatibility or driver issues. Based on their specifications, they are using almost the same instructions, protocols, features. And the 8 or 16 core stations are quite stable. It’s like upgrading from DDR3 to DDR4 or from Intel i9 9900 to Intel i9 12900.
Don’t get me wrong. The Xeon 6334 looks like a very good CPU at the wrong time period. A good option to upgrade secondary or supportive servers but not main ones. As I also mentioned above, more cores does not always mean higher performance. There are tasks when that matters like while running multiple virtual servers or Docker containers under Linux Alpine OS, for example. But higher frequency is always better for EVE nodes, also cache size of course and data exchange speed between interfaces. EVE node isn’t a FTP-server or Web-server, responsiveness and stability under load is way more important than security, data correction and so on.
More on this @Baldeda_Maxi; there was yet another deployment today to Search where the team is making changes and adding logging to try and track down these issues.
