I meant not familiar if the achieved 1600ish GFLOPS was remarkable in any wayLinX is vanilla Linpack, which is used in all sorts for testing software like OCCT for instance.
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
I meant not familiar if the achieved 1600ish GFLOPS was remarkable in any wayLinX is vanilla Linpack, which is used in all sorts for testing software like OCCT for instance.
He he!encouraging data on floating point multithreaded perf
16 core 9950X almost matching latest 32 core threadripper
Alleged AMD Ryzen 9 9950X 16 Core "Zen 5" CPU Benchmarked In AIDA64 With DDR5-8000 Memory, Up To 45% Faster Than 7950X
AMD's flagship Ryzen 9 9950X CPU with 16 Zen 5 cores has allegedly been tested in AIDA64 benchmark by a user at Anandtech Forums.wccftech.com
and also an utter failure of ability to calculate percentages
I can't even figure out what they did to arrive at those numbers
and a hint that Zen5 will appreciate tuned low latency ram in the end? not familiar with LinX
Zen6 it's rumoured, but might need a change of socket too.I was hoping for AMD to increase the core count to 32 cores 16 per CCX
any eta for Zen 6?Zen6 it's rumoured, but might need a change of socket too.
I expect roughly 18 months give or take as usual.any eta for Zen 6?
Intel's new wonder - aside from talking up their foundry - is the Atom, née E core, team. Time will tell but after their P cores grew so big in size and power they are betting big on the E cores. Smaller than Zen C cores. But no AVX-512 must be hard sell as that was Intel's wonder for many years!Lot of caveats all mentioned there, this mostly only showcases AVX-512 in extreme case, where a workload can keep floating point units fed as much as possible.
But still that illustrates power efficiency and high sustained load clocks.
Intel is definitely screwed for productivity
meh the comment about changing socket is annoyingI expect roughly 18 months give or take as usual.
For work I have an Alder Lake laptop and plenty of programs don't like the E cores including the database version we are currently on. There was one query which was twice as slow compared to the previous Haswell laptop which in every spec was slower. Plus the Haswell 37W M CPU was fine with a 45W, the Alder Lake 25W U CPU throttles unless given a 65W supply!
Intel PR are currently busy selling no more SMT though! So overall a huge switch - AMD are now the SMT and AVX-512 champions while Intel are claiming dropping SMT lets then chase better ST.
I don't think its likely for Zen 6meh the comment about changing socket is annoying
Agreed, and they already extended support for new processors on AM5 until 2027.I don't think its likely for Zen 6
One big reason to change socket would be a shift to DDR6, probably won't be ready in 2027.
CAMM2 DDR5 will be all the rage in between so we will definitely see new motherboards and high speed DDR5
Sorry, it had been a very long, hot, and tiring day.I meant not familiar if the achieved 1600ish GFLOPS was remarkable in any way
Yes, all Win11. Unsure what's the story with mssql and Alder Lake but while we're on the same old version as our apps, I did try that query with the latest version plus all the updates. Same results. Fiddling with core affinity or even turning off the E cores didn't help either. It's not a huge deal - it wasn't a production query - but a bit of a mystery and unexpected surprise.Are you using Windows 11 on the Alder Lake laptop?
I've so far had no problems with Windows 11 and a 14700K - but we've other systems here on Windows 10 or 11 with 12th or 13th gen CPUs which the E cores can cause some issues seemingly but it is hard to compare like for like. Also some of the newer laptop CPUs are a bit of a con in some ways in that once you exceed about 75% CPU core utilisation the boost drops right out and/or you hit power limits before maxing out all the cores so you don't really gain any performance beyond a certain point or can even lose a bit of performance if maxing the CPU out compared to just disabling say a couple of cores.
Personally I think dropping hyperthreading is the wrong direction - while the average performance uplift from it might be relatively low percentages there is also plenty of applications which can see 1.7-1.95x performance gains from it, or just run smoother with it despite minimal performance uplift.
Yes, all Win11. Unsure what's the story with mssql and Alder Lake but while we're on the same old version as our apps, I did try that query with the latest version plus all the updates. Same results. Fiddling with core affinity or even turning off the E cores didn't help either. It's not a huge deal - it wasn't a production query - but a bit of a mystery and unexpected surprise.
9700x TDP going back up again, from 65w to 120w
Looks like amd finally realised I was right and moving to 65w is a stupid move
ive never understood the 65w tdp on dekstop parts, as long as they dont push it to far the other way...9700x TDP going back up again, from 65w to 120w
Looks like amd finally realised I was right and moving to 65w is a stupid move
65 watt is the capacity of those crappy $10 coolers OEMs put in their prebuilds, so I think that's why they still exist even though realistically the number is meaningless.ive never understood the 65w tdp on dekstop parts, as long as they dont push it to far the other way...
ditch the coolers, a lot of people buy there own anyway so wasted material, plus amd save loads of rnd and material costs.. i guess they want to cover all bases hence adding it in. if your smart enough to buy your own cpu and socket it, you should be smart enough to buy a separate cpu cooler65 watt is the capacity of those crappy $10 coolers OEMs put in their prebuilds, so I think that's why they still exist even though realistically the number is meaningless.