News Rocket Lake Flaunts Up To 18% Higher Single-Core Performance Than Core i9-10900K

[TerryLaze]

To be fair, it did take Intel almost six years and a back-port to 14nm to get that 19% due to perpetual delays and hiccups on 10nm to get there which is rightfully worthy of ridicule. To make things worse, 7nm got delayed by six months too, so 10nm may not be the end of Intel's years-long process delays either and now Intel is seriously looking into outsourcing more production to bridge gaps.

To be fair intel had the palm cove in a 10nm CPU, the i3-8121U, since 2018.
The question is if they didn't bring it to desktop because the production would be too expensive and would have bad yields or if they were just making way too much money from 14nm on desktop to change it.

 

[RodroX]

lol this went soo off rute for just a leak of a supposed intel CPU which may or not be and early sample of the next Rocket Lake chip. Theres really no need to argue this much.

And all fanboys should really take a long break and relax. Most of us really want both AMD and Intel to keep launching products and getting better and better, why? easy we win, the consumer.

 

[InvalidError]

To be fair intel had the palm cove in a 10nm CPU, the i3-8121U, since 2018.
The question is if they didn't bring it to desktop because the production would be too expensive and would have bad yields or if they were just making way too much money from 14nm on desktop to change it.

Pretty sure the answer to that question is: 10nm was junk, yields and performance so bad that Intel gave up on making first-gen 10nm CPUs' IGP usable and officially canned driver development for it some weeks ago. Intel itself admitted that 10nm was nowhere near where it wanted to be and even 10nm+ is a watered-down version of the 10nm process Intel wanted to get to back in 2016 just to get some usable products with scaled-back performance out the door. There is no desktop 10nm because after all of the 10nm complications, Intel decided to bet on a skip to 7nm about two years ago to avoid the sunk cost of temporary 10nm fabs... and now it gets the double-whammy of 7nm delays on top.

Most of us really want both AMD and Intel to keep launching products and getting better and better, why? easy we win, the consumer.

If consumers were really winning, AMD and Intel wouldn't be jacking up prices to pad their profit margins by an extra 10-20%. The shareholders are winning, not consumers.

 

[logainofhades]

I would rather think that most people aren't bothering to upgrade at all for 20% gains and skip a few generations instead - in the 90s where CPUs were getting much faster much quicker, most people upgraded only every 3-4 years, usually to something that was both more than twice as fast and cheaper.

Enthusiasts' willingness to throw absurd amounts of money at tiny incremental upgrades is nuts.

With how fast GPU's are getting, a 19% IPC improvement is pretty big, and needed for those cards to be fully utilized, for high refresh rate monitors, at 1080p and 1440p. I am highly considering upgrading my 3700x, to a 5000 series CPU, next year, and possibly an RX 6800, to replace my RTX 2060.

 

[spongiemaster]

To be fair intel had the palm cove in a 10nm CPU, the i3-8121U, since 2018.
The question is if they didn't bring it to desktop because the production would be too expensive and would have bad yields or if they were just making way too much money from 14nm on desktop to change it.

Palm Cove was such a failure, it doesn't even have a Wikipedia page and there are cancelled architectures that have Wikipedia pages. Palm Cove was a tick in Intel's tick-tock model and did not have any IPC improvement over Sky Lake. It added AVX-512 instructions, which I'm sure were tremendously useful in a low power dual core mobile CPU.

 

[RodroX]

If consumers were really winning, AMD and Intel wouldn't be jacking up prices to pad their profit margins by an extra 10-20%. The shareholders are winning, not consumers.

So you don't consider the advance in core counts in the Desktop and HEDT segments and the corresponding price drops peer core as a wining?, fair enough. But for many people that weren't able to pay +$ 1500 for an intel X or Xeon part just to get a mere 8 core its a wining.

Outisde the US and perhaps a few European countries its not easy to access technology, and the used market is very very small and shady. So we do appreciate that AMD was able to get back on thier feet and bring a competing product with decent prices (specially if you consider the extra resources and the low price platform access).

Some of my friends, work colleagues and clients that are related to programming and content creation are very happy to be able to snag a 6c/12t, or 8c/16t cpu for stupid prices considering what you used to pay for them (if you could) just a couple of years ago.

 

[InvalidError]

So you don't consider the advance in core counts in the Desktop and HEDT segments and the corresponding price drops peer core as a wining?

What I consider winning?

• in 2000, I bought a 650MHz P3 Coppermine for $350
• in 2004, I bought a 3GHz Northwood P4/HT for $280
• in 2008, I bought a 3GHz C2D E8400 for $220
• in 2012, I bought an i5-3470 for $170

More everything, still decreasing prices. That is consumers winning.

In 2020, the same $170 going by launch MSRP of modern parts lands me in i3 territory on Intel's side and a Ryzen 3400G on AMD's side since most other sub-$200 AMD SKUs are out of stock. The only current-gen things I can "upgrade" to without paying more are ~30% faster than an eight years old i7-3770. Grossly underwhelming.

 

[RodroX]

What I consider winning?

• in 2000, I bought a 650MHz P3 Coppermine for $350
• in 2004, I bought a 3GHz Northwood P4/HT for $280
• in 2008, I bought a 3GHz C2D E8400 for $220
• in 2012, I bought an i5-3470 for $170

More everything, still decreasing prices. That is consumers winning.

In 2020, the same $170 going by launch MSRP of modern parts lands me in i3 territory on Intel's side and a Ryzen 3400G on AMD's side since most other sub-$200 AMD SKUs are out of stock. The only current-gen things I can "upgrade" to without paying more are ~30% faster than an eight years old i7-3770. Grossly underwhelming.

You could have gotten a Ryzen 5 3600 (which is 50% more cores and SMT enable + a huge bump in performance) for $165~$175.

Then again not everyone need it the core count increase and the corresponding price drops, so if you are happy and your Core i5 does everything you need for you why changing?

I do understand the frustration of having to deal with intel monopoly and product stagnation for many years (basically from 2011 till 2018). Thats the main reason I waited soo long to change my old Core i5 3570 for something new, a Ryzen 5 3600, and as you said for that money back in 2019 I could have only gotten a Core i3 instead. So it was no brainer, Ryzen all the way.

 

[InvalidError]

You could have gotten a Ryzen 5 3600 (which is 50% more cores and SMT enable + a huge bump in performance) for $165~$175.

Lowest price on PCPartspicker USA is currently $200. Lowest I've seen it in Canada was $230 around February, currently $280 including shipping.

As for whether my i5 still does what I need it to do well enough for my liking, I'd say 90-95% of the time. The remaining 5% takes hours now and would still take hours no matter what I may upgrade to, so I use the "go do something else" strategy backed by 32GB of RAM to make sure swapping won't make me acutely aware that I have stuff running in the background like it did back on my Core2 with 'only' 8GB.

Upgrade-wise, I may consider an i5-11400 next year. I'm not feeling particularly zen about Ryzen since my first experience with it (a friend brought parts for an 1700x/x370 build and asked me to put it together) was a no-boot with the store blaming us for frying components until they tried putting the exact same config together from fresh parts and couldn't get that to boot until a whole bunch of parts-swapping later.

 

[badger2k]

UserBenchmark	Octa-core Rocket Lake	Core i9-10900K	Core i7-10700K
1-Core	179	152	148

Core i9-10700K = 5.1 Ghz = 148 =>> 1 Ghz = 148 / 5.1= 29.02
Rocket Lake = 4.2 Ghz = 179 =>> 1 Ghz = 179 / 4.2= 42.62

1 Ghz Rocket Lake vs 1 Ghz 10700K = 42.62 / 29.02 = 147% = +47% IPC ?¿?¿??¿? I don't think so.

FAKE!!

The benchmark itself is genuine but the reported boost clocks are obviously incorrect. I estimate that single core boosts were around 4.8ghz in this instance.

 

[mitch074]

Lowest price on PCPartspicker USA is currently $200. Lowest I've seen it in Canada was $230 around February, currently $280 including shipping.

As for whether my i5 still does what I need it to do well enough for my liking, I'd say 90-95% of the time. The remaining 5% takes hours now and would still take hours no matter what I may upgrade to, so I use the "go do something else" strategy backed by 32GB of RAM to make sure swapping won't make me acutely aware that I have stuff running in the background like it did back on my Core2 with 'only' 8GB.

Upgrade-wise, I may consider an i5-11400 next year. I'm not feeling particularly zen about Ryzen since my first experience with it (a friend brought parts for an 1700x/x370 build and asked me to put it together) was a no-boot with the store blaming us for frying components until they tried putting the exact same config together from fresh parts and couldn't get that to boot until a whole bunch of parts-swapping later.

Well, early Ryzen motherboards were utter crap, apart from some select models from brands that put more stock in proper VRMs than RGB or plastic shrouds over I/O panels - and not the highest range at that. I have a B350 Bazooka on a test bench, it works wonders.
I built many Ryzen systems since 2017, the first and second ones were rather unstable for the first six months. After a BIOS upgrade and a newer Windows 10 release, they are now rock solid. Newer Ryzen hardware (since S2 2018 at least) always worked right away - no glitches, no BSOD, excellent performance.
I had trouble with Haswell-based systems though, where Windows 10 after 2017 crashed at boot in UEFI mode. Probably to be blamed on the mobo maker's lack of BIOS fixes, but that proves that it really depends on the mobo manufacturer - not the CPU itself.

 

[JayNor]

To make things worse, 7nm got delayed by six months too, so 10nm may not be the end of Intel's years-long process delays either and now Intel is seriously looking into outsourcing more production to bridge gaps.

Intel's Sapphire Rapids server chip is "broadly sampling", according to Intel, and includes implementations of both PCIE5/CXL and DDR5 support.

Apparently Intel's 10ESF has enabled Intel's processing bandwidth greatly over what is available from zen3, since AMD's second generation IF doesn't have the bandwidth to support PCIE5.