...vs Rocket Lake. Not really a hard thing to do, given 14nm vs. "Intel 7".
This is not the point, you don't get it. Intels up to is not a real up to in many cases.
...vs Rocket Lake. Not really a hard thing to do, given 14nm vs. "Intel 7".
The RPL number was never specified as being single thread, and multithreaded will surely demonstrate better gains. Would be very surprised if single thread ends up even 10% better.His point is that they said, "Up to 20% ST performance" improvement for Alderlake and the average ended up around 19%. So the up to statement was really the upper range for the average single thread improvement. If the same holds true for RPL, then the average single thread improvement will be 10+%. We'll have to wait and see how it goes.
Yes, you are right, I got it confused with all the ST discussion recently. I will say, if they can only manage low double digit MT improvement (like 10 - 15%), that would be pretty disappointing. Unless they're taking the opportunity to bring the power consumption back down a bit at the same time.The RPL number was never specified as being single thread, and multithreaded will surely demonstrate better gains. Would be very surprised if single thread ends up even 10% better.
You are the first that's expecting lot out of Meteorlake desktop.Raptor Lake will advance in many areas, but until MTL-S, I'm not expecting anything decent out of Intel.
That's obviously not happening. If you go down from 250W to 200W, you are bound to lose potential performance. If people are accepting of 250W, why go down?Unless they're taking the opportunity to bring the power consumption back down a bit at the same time.
I do think there will be socketed Meteor Lake parts, but I doubt they'll be of much interest to this forum.I expect something decent out of MTL-M and U, but nothing much more than Tigerlake did for desktops. They'll have NUC parts and BGA desktop, that's about it. It's Arrowlake that they might do something significant for desktops.
There is one possibility. Namely, a design or process that limits the Vmax of the processor to such a degree that power is effectively capped lower. Like, Apple's cores max around 1.0V, right? Even in devices with an effectively unlimited power budget, peak clocks are no higher than laptop, tablet, or even phone parts. I have no idea if Intel is pursuing such a design, but it's at least possible.That's obviously not happening. If you go down from 250W to 200W, you are bound to lose potential performance. If people are accepting of 250W, why go down?
If they make a radically better chip, I wouldn't be surprised if it still ends up being 250W.
The same with GPUs. The TDP ceiling is going up. 600W is in the cards for consumer because HPC GPUs are going to exceed a kilowatt.'
Increased efficiency NEVER translated into lower power. Yes you can choose to get lower power devices, but every segment that gets more efficient it just increases the dynamic range. LEDs are 10x+ more efficient than incandescent but flashlights nowadays are about 10x brighter.
The lighted ball for Times Square used incandescents and was only on for New Year's Eve. They changed to LED and is now on for all year long. That's what happens and is a natural progression. I bet you that mentality is across the board. You need to put effort to go against that trend, since everybody does it and doesn't come naturally.Also, for your lightbulb example, a modern LED replacement for an incandescent bulb has substantially lower absolute power draw. Even if you could cool an LED bulb with incandescent wattage, it would be obnoxiously bright.
At the risk of getting too off topic, the examples you cited are outliers. For the vast majority of household and commercial lighting, on a per household basis, switching to more efficient technologies has more than offset any increases in total amount of lighting.The lighted ball for Times Square used incandescents and was only on for New Year's Eve. They changed to LED and is now on for all year long. That's what happens and is a natural progression. I bet you that mentality is across the board. You need to put effort to go against that trend, since everybody does it and doesn't come naturally.
When you look at the LED lights out there, it's mostly used for higher brightness. Yes some are lower power sure, but now you have many more then you did back in those days. I know people with more than half a dozen of them. And you stick them in every room, back and front of the house, and you have them on for much longer.
Not an exception for me either. I loved flashlights and lights in general. Used to have incandescents and they were probably 50 lumens at the best. Now my top lights are 800 lumens.
It's all used for higher dynamic range. High end CPUs used to be 2-3W and didn't even have passive cooling. It didn't need lot of power saving features since the max power was already pretty low.
That's it. Direct replacements may be lower power but it always brings up new use cases and overall it's not lower. Yea housing lights may be a bit lower, but what about overall? We used to go outside and many of us nowadays spend 3-4 hours a day on our computers that's pretty much plastered every corner now.I think there's a more interesting discussion to be had for datacenters, and particularly AI, where growth in compute demand is vastly outstripping increases to efficiency.
Now add the power consumption. . .![]()
CB R23 MT
6P 5ghz = 15555
4E 4ghz = 4 179
So you can do match
8P 5ghz = 20 688
8E 4ghz = 8 358
If Raptor Lake has no IPC gains in productivity then
13900K = 20688+8358+8358 = 37404
13900K will score 37404 with P cores at 5ghz and E cores at 4ghz.
That will be a cool 37404 points at some 300+ Watts or more. What an achievement.![]()
CB R23 MT
6P 5ghz = 15555
4E 4ghz = 4 179
So you can do match
8P 5ghz = 20 688
8E 4ghz = 8 358
If Raptor Lake has no IPC gains in productivity then
13900K = 20688+8358+8358 = 37404
13900K will score 37404 with P cores at 5ghz and E cores at 4ghz.
Depends on clocks and whether E-Cores are decoupled in voltage from P-Cores and Ring. I doubt they will have trouble scoring 35k @ 230ish watts or whatever that PL2 is nowadays. AMD could probably score 32K @ 100W, but Intel will force them to burn power to beat them.That will be a cool 37404 points at some 300+ Watts or more. What an achievement.
You won't need it to burn 300W if they want just 30% improvement like @SmokSmog calculates.That will be a cool 37404 points at some 300+ Watts or more. What an achievement.
If Intel can improve efficiency by 10+% overall with RPL, I could see it hitting over 37000 points in R23. Otherwise, I think it's going to come in more like 34000ish range.You won't need it to burn 300W if they want just 30% improvement like @SmokSmog calculates.
Raptor Cove is going to end up being few % faster per clock. Simply lowering the clock(and voltage a bit too) to make it equal to Golden Cove will free up quite a bit of TDP for the extra E cores.
Also assumes there will be zero which makes little sense as refreshes get 100-200MHz at same TDP. 1185G7 to 1195G7 is 3.4% increase in base clocks and 4.2% increase in max Turbo. If you get few % of efficiency on the Gracemont portion, again they can keep the clocks same to reduce the power per core a bit too.
You can see here the 8+0 with no power limit is just 10% faster in average MT applications compared to 125W limited 8+0.
![]()
Intel Core i9-12900K, i7-12700K & i5-12600K im Test: Leistung und Effizienz P- vs. E-Core
Intel Alder Lake im Test: Leistung und Effizienz P- vs. E-Core / Wie effizient ist die Hybrid-Architektur?www.computerbase.de
However, Intel doesn't need to cut P core power by half to 125W.
In fact it shows that 8+0 can use nearly 250W by itself! That's why with extra E cores it's worth it to back down the P cores slightly and get massive boost in MT performance at the same TDP. Alderlake has to max the P cores to squeeze the last few % of performance. Raptorlake will be comparably much more efficient even if nothing gets better because of this.
There were some rumors that Alderlake was to be mobile only. Clearly you can see how much they have to push 8+8. 8+16 will be far better.
Due to Power Budget/TDP the 12900K is closer to 27,000. The 13900K/KS Might also be Power Budget Constrained and a more realistic number would be 36,000 MT Points in CBR23![]()
CB R23 MT
6P 5ghz = 15555
4E 4ghz = 4 179
So you can do match
8P 5ghz = 20 688
8E 4ghz = 8 358
If Raptor Lake has no IPC gains in productivity then
13900K = 20688+8358+8358 = 37404
13900K will score 37404 with P cores at 5ghz and E cores at 4ghz.
P core will be at 5.8ghz and e cores 4.5ghz plus ipc.. so 40k to 42k range when the 13900ks high quality silicon 😏😀![]()
CB R23 MT
6P 5ghz = 15555
4E 4ghz = 4 179
So you can do match
8P 5ghz = 20 688
8E 4ghz = 8 358
If Raptor Lake has no IPC gains in productivity then
13900K = 20688+8358+8358 = 37404
13900K will score 37404 with P cores at 5ghz and E cores at 4ghz.
Not in an MT workload. The clocks will probably be closer to a 12900KS.P core will be at 5.8ghz and e cores 4.5ghz plus ipc.. so 40k to 42k range when the 13900ks high quality silicon 😏😀
I think a lot depends on what clocks they can achieve in 10W per core region with CB23 load. It takes Intel ~15W per core to achieve 5Ghz on P cores, so i would not be surprised that AMD can hit 5Ghz on new process @ 10W.7950X will supposedly hit ~41k in CBR23 @ 170W PPT.
It means they will be closer to each other than Zen3 and ADL are today. So it will be tradeoffs as usual: homogenous core vs heterogenous core and dealing with scheduler woes. Or having monolith chip versus chiplet based one.Only 15% increased ST vs Zen3 means RPH will be slower than RPL for most consumer applications?
A nearly 29K score in R23 for the 5950X requires an overclock to 4.5 GHz all core. I'd be surprised if AMD is claiming the 40+% over that, more likely it's against a stock 5950X that scores in the 25.5K range.7950X will supposedly hit ~41k in CBR23 @ 170W PPT.
A 3950X can hit 26k stock. Are you sure you're talking about the right CPU?A nearly 29K score in R23 for the 5950X requires an overclock to 4.5 GHz all core. I'd be surprised if AMD is claiming the 40+% over that, more likely it's against a stock 5950X that scores in the 25.5K range.
Not at stock:A 3950X can hit 26k stock. Are you sure you're talking about the right CPU?
Not at stock:
Actually not really. The only bearing memory has on a 5950x for CB R23 is higher FCLK/memory eats into the limited power budget and forces lower clocks. At stock jedec 2133/1066 FCLK I will see slightly higher all core boost and around 26.5k scores. At my preferred 3800/1900 FCLK I get lower boost with stock power limit and see ~25,500.![]()
Cinebench R23 Scores [Updated Results]
Cinebench R23 is the newest instalment of the most popular CPU-Rendering Benchmark Cinebench. We have the Score Results for all modern Processors.www.cgdirector.com
Cinebench R23 (Multi-Core) CPU benchmark list
Cinebench R23 (Multi-Core) CPU benchmark listwww.cpu-monkey.com
I see a lot of variation here in CBR23 scores. Might be down to memory, but I think 26k is on the slow side for 5950X.
In any case, until we see more numbers it'll be hard to know exactly what 42% faster is . . . but then we haven't seen any numbers at all for 8+16 Raptor Lake.