GPUs & CPUs & Enthusiast hardware: Questions, Discussion and fanboy slap-fights - Nvidia & AMD & Intel - Separe but Equal. Intel rides in the back of the bus.

Overly Serious said:
Yes, you can. But I'm hoping channels like Hardware Unboxed are mostly puling out a fair and representative sample. In any case, the best Intel can say is that it's close.
Click to expand...

This is why, to your earlier point, it's a bad marketing stunt. We've known for over a year now, that games vary widely in which are memory-bound and which are compute-bound at extreme fps, so whoever at Intel thought blasting as much power as they could through the chip and do any better than "mixed results" should probably just be fired.

Because, aside from not being a good value, it's not succeeding at its primary purpose, which is to boost the brand.
 
  • Like
  • Thunk-Provoking
Reactions: Lone Wandering Courier, jeff7989 and Overly Serious
The Mass Shooter Ron Soye said:
I hope we see some better stunts from Intel soon, and change is coming. Arrow Lake is tile-based, and it's conceivable that Intel could put Adamantine L4 cache on Arrow Lake desktop chips as their own answer to X3D.
Click to expand...
I've only been following this loosely this time around but I am keen to see Arrow Lake chips. Still forecast for late this year, I believe? And increasing rumours of AMD revealing Zen 5 this Summer, presumably available to buy before the year out as well. So will be interesting to see how the PR war heats up between the two companies. I'm about ready for a full system upgrade (am on first gen Zen chip) and tempted to buy now but am drawing it out a little to see what Zen 5 looks like. Mainly for the expected larger instruction set but also just common sense.
 
A newly-discovered, Microsoft-branded SSD suggests the tech giant may be – or has been at least - exploring new ways to optimize its data center storage
The leaked images of a Microsoft Z1000 SSD show a 1TB NVMe M.2 drive, apparently boasting sequential read speeds of up to 2,400MB/s and write speeds of 1,800MB/s.

The Z1000 SSD, originally revealed by @yuuki_ans on X, is made up of a mix of components from various companies, including Toshiba NAND flash chips, Micron's DDR4 RAM cache, and a controller from CNEX Labs, a company best known for its work with data center hyperscalers.
Click to expand...
 
  • Thunk-Provoking
Reactions: bigoogabaloogas and The Ugly One
Who is the 14900KS even for? With such low volume I think Intel should have just seeded a bunch of overclockers with them and dominated whatever benchmarks matter.
 
  • Like
Reactions: Overly Serious
KS has always been for the tiny group that needs to win the silicon lottery with their CPU and is willing to waste lots of money. I'm perfectly fine with it.
 
Almighty Space Duck said:
With such low volume I think Intel should have just seeded a bunch of overclockers with them and dominated whatever benchmarks matter.
Click to expand...
Still get wrecked by threadripper, basically.

On the high end intel and amd aren't even close, chiplets OP
1710713883326.png
 
  • Dumb
Reactions: George Lucas and The Ugly One
The Ugly One said:
Y-yeah but what about when you control for an arbitrary weight class distinction [barely holding back tears]
Click to expand...
Insofar as the 14900KS has not yet apparently had any good passmark runs, it would appear based on other related intel chips on the same chart it will likely still lose to ryzen series chips that are circa 2/3 of the cost.
1710721812701.png
 
  • Dumb
Reactions: George Lucas, The Ugly One and Forbidden Existence
Betonhaus said:
That line at the very bottom is a Xeon processor, Intel's very best.
Click to expand...

Intel's very best right now is the Xeon Max 9480, which still gets beat by an EPYC 9684X in most workloads. Neither are on the chart. Neither have any bearing on Core vs Ryzen.

geckogoy said:
Insofar as the 14900KS has not yet apparently had any good passmark runs, it would appear based on other related intel chips on the same chart it will likely still lose to ryzen series chips that are circa 2/3 of the cost.
Click to expand...

That's not what your chart says. Why do you think the distinction between server and desktop CPUs is arbitrary?
 
Last edited:
I'm trying to figure out why a 9th gen I7 still goes for $100+. All I want to do is make my NAS/Plex box set for the future, not question if I should upgrade to a later gen for a few dollars more.
 
You're MDNWU said:
I'm trying to figure out why a 9th gen I7 still goes for $100+. All I want to do is make my NAS/Plex box set for the future, not question if I should upgrade to a later gen for a few dollars more.
Click to expand...
The used market (in my experience so far) tends to reasonably closely price in performance per dollar, so you can hyper finagle and still get a bargain but generally going back a generation or three doesn't hugely effect what you pay vs what you get. Going to an older socket can help a bit more.
 
You're MDNWU said:
I'm trying to figure out why a 9th gen I7 still goes for $100+. All I want to do is make my NAS/Plex box set for the future, not question if I should upgrade to a later gen for a few dollars more.
Click to expand...
The best CPU for its socket/chipset always carries a premium. The 9900K also has the lowest potential memory system latency of any 8 core x86 CPU that I am aware of, so it is still intrinsically desirable for some very niche applications (theoretically). Tech Yes City made a few videos on this topic.

 
  • Informative
Reactions: You're MDNWU and The Ugly One
BTW, for those who don't know, the difference between server & desktop CPUs is data throughput. According to geckogoy's chart, the i9-14900k is "better" than a Xeon 6448Y because it has a higher clock speed. Except it's not that simple.

CoresClockMax RAMMemory bandwidthPCIe lanes
Xeon 6448Y32P2.1-4.1 GHz4096 GB300 GB/s80
i9-4900K8P + 16E3.2-5.8 GHz192 GB89.6 GB/s20

A chart of single-core (EDIT:) composite performance based on mainly compute-bound workloads that compares server & desktop CPUs is dumb. Yes, the i9 has a higher clock speed than the Xeon. But let's say I want to run an in-memory database across 32 server nodes, 8 TB per node and 100G Ethernet cards. You cannot physically build that out of desktop CPUs, and the workload scales extremely well across the thousands of cores in the system. The fact that your Cores and Ryzens can have higher clock speeds is irrelevant.
 
Last edited:
  • Informative
  • Agree
Reactions: 4090 Chad, George Lucas and seri0us
geckogoy said:
Passmark is not single core unless you run it in single core mode, which is not a chart I linked to
Click to expand...

Looks like a composite score that includes single thread.

1710784913783.png

Run a test on a 2 TB database using your Ryzen and see how it stacks up.
 
The Ugly One said:
Looks like a composite score that includes single thread.
Click to expand...
The single threaded performance lives in another chart from the one I was pasting

It seems like you never even heard of passmark before, and are now trying to argue memory bus is most important, except I was originally replying to 'what benchmarks matter' for a 14900KS, which would definitely be passmark.

Even if you control for the arbitrary server/desktop distinction, and in your case fail to notice the fact that HEDT/workstation chips that are not meant for severs are chart toppers anyways, AMD is still doing really well on raw processing power just on the Ryzen series by coming in at a much better price per performance and you just seem kindof dumb.
 
Last edited:
  • Dumb
Reactions: George Lucas and Forbidden Existence
You must log in or register to reply here.
Back