Review Intel Core i9 10850K Review (TechPowerUp)

[Zucker2k]

With a retail price of $450, the Core i9-10850K is $50 cheaper than the Core i9-10900K. That's a 10% reduction in cost for barely a few percent in performance. Definitely consider the i9-10850K if you're in the market for an i9-10900K. What's probably more important than $50 for many people is that availability of the i9-10850K seems to be much better than for the i9-10900K. Especially in the States, as Intel CPU supply is much worse than in Europe, where you can just go out and buy an i9-10900K without having to wait for the stars to align. That's why I'm also not surprised to see terrible pricing in the U.S., no doubt from greedy merchants. Looking at our performance numbers and the pricing, I would definitely prefer the i9-10850K over the Ryzen 9 3900XT, and possibly even the Ryzen 9 3900X, for gaming and general productivity. Professionals working with rendering and simulation apps, or other similarly demanding apps, should definitely consider AMD for their rigs, as the higher thread count can make a difference. We reviewed the Core i9-10900 only recently and liked it very much. However, it's kind of obsoleted by the i9-10850K because of the small price difference. $10 more gets you an unlocked multiplier, much higher base clock, and higher power limit—just the boxed cooler is missing. Definitely worth considering. On the other hand, if the i9-10900 drops to $400, it would compete with the Ryzen 7 3800XT in an interesting reversal of 10-core Intel vs. 8-core AMD at the same price.

Intel Core i9-10850K Review - Just as Good as the i9-10900K

The Intel Core i9-10850K is the company's latest Comet Lake processor. It's just 100 MHz slower than the Core i9-10900K flagship, but much more affordable, and with better availability. In our i9-10850K review, we're taking a close look at both gaming and application performance to determine...

www.techpowerup.com

 

[Edrick]

Are people still dropping $500+ on 5 year old architecture and a 6 year old node?

 

[Makaveli]

Are people still dropping $500+ on 5 year old architecture and a 6 year old node?

Considering intel is still making record profits per quarter yes.

The average consumer doesn't know what an architecture is or a node.

 

[TheELF]

Are people still dropping $500+ on 5 year old architecture and a 6 year old node?

People are dropping $500+ on CPUs that have similar IPC but can barely surpass 4Ghz all core, so yeah of course.

 

[Atari2600]

Are people still dropping $500+ on 5 year old architecture and a 6 year old node?

Unfair - no fan of Intel and its underhand methods, but if its still churning out calculations as quick as the competition at similar wattage, then process and architecture are irrelevant.

 

[UsandThem]

Unfair - no fan of Intel and its underhand methods, but if its still churning out calculations as quick as the competition at similar wattage, then process and architecture are irrelevant.

Very true in most accounts, but the security issues/concerns of older architecture is a concern for some.

Yes they are all eventually patched with firmware and software updates, but at the expense of performance.

 

[ondma]

Very true in most accounts, but the security issues/concerns of older architecture is a concern for some.

Yes they are all eventually patched with firmware and software updates, but at the expense of performance.

How much "performance loss" is highly dependent on workload. Yes, some I/O intensive workloads have shown a decrease, but in gaming and day to day use, performance loss is minimal (probably not even noticeable without running benchmarks).

 

[Zucker2k]

How much "performance loss" is highly dependent on workload. Yes, some I/O intensive workloads have shown a decrease, but in gaming and day to day use, performance loss is minimal (probably not even noticeable without running benchmarks).

What's more, Intel is still competitive on a five year old architecture and a six year old process, even with all those mitigation patches! Makes you wonder how much the competition has been sleeping all these years.

 

[DrMrLordX]

What's more, Intel is still competitive

No, they aren't.

 

[TheELF]

No, they aren't.

That's right since competitive would imply them being close but 5Ghz + all core compared to 4Ghz + all core is not even close.
AMD is cheaper per core but quantity is not quality.

 

[A///]

You're right. AMD isn't great for pure gaming. Anything else, they're brilliant.

 

[Markfw]

That's right since competitive would imply them being close but 5Ghz + all core compared to 4Ghz + all core is not even close.
AMD is cheaper per core but quantity is not quality.

And some people have still not learned that GHZ does not equal processing power.

 

[.vodka]

My 3900X is doing over 15 GB/s in AES. Overclocked to 4.4 GHz, it's doing over 20 GB/s.

This is a stock 3900x (no PBO), well cooled, with fast memory/fclk (3800MHz CAS16, tuned secondaries and tertiaries)

It's mostly a memory speed benchmark. It says so right there, "these tests take place in RAM". TPU probably forgot to load the XMP profile and all the Zen2 CPUs are running DDR4-2133 to get results that low, lol

Having said that, I wonder how much further it'd go with some CCX overclocking. Too lazy to do it, I'm waiting for 1usmus' automated tool to release anytime soon.

 

[DrMrLordX]

That's right since competitive would imply them being close but 5Ghz + all core compared to 4Ghz + all core is not even close.
AMD is cheaper per core but quantity is not quality.

Hi, 2016 called and wanted their meme back. Try benching any Intel Comet Lake-S against an aging 3950X and tell me how they're "competitive". AMD is cheaper? As if.

@.vodka

Oh okay, I only run DDR4-3666 CAS/CL14. Seems like it wants speed over timings.

 

[Zucker2k]

Hi, 2016 called and wanted their meme back. Try benching any Intel Comet Lake-S against an aging 3950X and tell me how they're "competitive". AMD is cheaper? As if.

@.vodka

Oh okay, I only run DDR4-3666 CAS/CL14. Seems like it wants speed over timings.

Skewed math is skewed. So, AMD brought down an HEDT chip onto desktop to claim throughput supremacy. Well, I got news for you; you'll need to shell $700+ for the chip alone. I hope you enjoy browsing youtube with your 32 threads while losing to an i5 10400 at only a quarter of the price in gaming, and some light-threaded stuff.



Trolling isn't allowed.


esquared
Anandtech Forum Director

 

[Makaveli]

Skewed math is skewed. So, AMD brought down an HEDT chip onto desktop to claim throughput supremacy. Well, I got news for you; you'll need to shell $700+ for the chip alone. I hope you enjoy browsing youtube with your 32 threads while losing to an i5 10400 at only a quarter of the price in gaming, and some light-threaded stuff.

Maybe gaming isn't the priority with his build?

 

[ondma]

Maybe gaming isn't the priority with his build?

That may be true. However, he seems obsessed with denigrating gaming as a valid scenario (i.e. "corner case") in order to claim, (in his opinion) intel chips "aren't competitive".

 

[A///]

And some people have still not learned that GHZ does not equal processing power.

Indeed. If you go back a mere 12 years, you would have seen a lot of first hand accounts of people moving from Pentium 4 to what was the first generation Core series under Nehalem. Even back about six to eight months after the majority of the original Core 2 Duo desktop processors launched, you would have people arguing that their Pentium 4 was just as good. I'm 100% certain you'll be able to recall this given you've been a hardware enthusiast most if not all of your life.

 

[Makaveli]

Indeed. If you go back a mere 12 years, you would have seen a lot of first hand accounts of people moving from Pentium 4 to what was the first generation Core series under Nehalem. Even back about six to eight months after the majority of the original Core 2 Duo desktop processors launched, you would have people arguing that their Pentium 4 was just as good. I'm 100% certain you'll be able to recall this given you've been a hardware enthusiast most if not all of your life.

I remember it I've been on this site longer than 12 years.

 

[A///]

I remember it I've been on this site longer than 12 years.

In the P4 days I'd make a cardboard shroud with ducttape and get all the exhaust heat and direct it to my feet. At the time my desktop sat on the floor. We had central air, but my feet had poorer circulation back then so it would keep them nice and warm due to the overclock I had. My first winter of my Core 2 Duo I found myself reaching for double socks and a blanket. At the time I considered dremeling the top portion right above my heatsink and placing a grate with exhaust fan so I could set my feet atop my case to warm them up and layback. Back then there were keyboards and mice with long ps/2 cords. I want to say wireless was available then but the response times were trash.

Kind of wish I kept all of the beige boxes I had in those days. You can get a pretty penny for them now if you sell to enthusiasts doing sleeper builds or do one yourself.

 

[Markfw]

Indeed. If you go back a mere 12 years, you would have seen a lot of first hand accounts of people moving from Pentium 4 to what was the first generation Core series under Nehalem. Even back about six to eight months after the majority of the original Core 2 Duo desktop processors launched, you would have people arguing that their Pentium 4 was just as good. I'm 100% certain you'll be able to recall this given you've been a hardware enthusiast most if not all of your life.

Not to go too OT, but I started being a computer enthusiast when I overclocked a 286 using a crystal.

On-topic....The point that keeps getting twisted is that GHZ is not equal to processing power. Yes, Intel currently has a small lead in most (not all) games, but a 3950x is not a gaming chip. Gaming is faster IMO on Intel due to latency, not the 5 ghz number. And the 10850k might very well be a decent gaming chip

Now in anything BUT gaming, meaning desktop, HEDT and server uses, Intel is NOT competitive, look at all the benchmarks out there. Gaming is just a different scenario, you can't compare the 2.

Without valid numbers, let me just throw up a scenario for you. DON'T quote my numbers, just trying to make a point.

So with both at 4 ghz, single core, a 10900k vs a 3900x lets say, the 3900x wins be a decent margin, so its more efficicient and faster in anything. Now in reality the 10900k can go to 5 ghz, so it more than makes up for that loss in speed in gaming, but looses in everything else. But the next gen Intel ? Most likely will not go that fast from what I have read.

So keep this discussion to gaming or total computing speed, don't mix cases. Otherwise its trolling IMO.

 

[A///]

On-topic....The point that keeps getting twisted is that GHZ is not equal to processing power. Yes, Intel currently has a small lead in most (not all) games, but a 3950x is not a gaming chip. Gaming is faster IMO on Intel due to latency, not the 5 ghz number. And the 10850k might very well be a decent gaming chip

Now in anything BUT gaming, meaning desktop, HEDT and server uses, Intel is NOT competitive, look at all the benchmarks out there. Gaming is just a different scenario, you can't compare the 2.

Without valid numbers, let me just throw up a scenario for you. DON'T quote my numbers, just trying to make a point.

So with both at 4 ghz, single core, a 10900k vs a 3900x lets say, the 3900x wins be a decent margin, so its more efficicient and faster in anything. Now in reality the 10900k can go to 5 ghz, so it more than makes up for that loss in speed in gaming, but looses in everything else. But the next gen Intel ? Most likely will not go that fast from what I have read.

Yeah. Not to get too off topic either, but right now I'd compare the situations between the two companies to that weird analogy that floated back when Intel launched Conroe. If my memory isn't too bad, it was that a Pentium 4 running at excess of 3.8 Ghz while powerful and hot, was like a Ferrari going up and down a mountain carrying two people and some golf clubs. The new Core 2 Duo was touted as a performance SUV that while slower, carried more people and baggage. It was a really dumpy analogy but it got the point across. Slower, but more transference from the base to the peak and back again per stop.

I'd say right now that latency due to how gaming works amplifies that issue quite poignantly. If you recall, there was a SiSoft dataset published back in late December showing core latency in multithreaded workloads. Z1 and Z+ had nearly the same latency numbers while Z2 experiences a great reduction. Intel's then 9900K had the tightest latency in terms of good and bad core pairings on the processor die during any given MT task. I think AMD can fix this, but I do suspect Intel will do some magic to get their new core architectures out onto client and DC ASAP. All I will say it's an amazing time to be a consumer. I don't think there's been a product in the last three years that has left me underwhelmed! Light in the wallet, yes.

Right now I'm more interested in which AIB I'll pick out that's the least visually offensive for my Ampere card. They're all... puke. Alderlake will be a day 1 purchase for me, even if Intel is behind even more by then. I think the concept is meh, but the idea is radical. It may work in the long run in the future. I don't know, but I'm not a naysayer haha.

 

[tamz_msc]

Skewed math is skewed. So, AMD brought down an HEDT chip onto desktop to claim throughput supremacy. Well, I got news for you; you'll need to shell $700+ for the chip alone. I hope you enjoy browsing youtube with your 32 threads while losing to an i5 10400 at only a quarter of the price in gaming, and some light-threaded stuff.

Do you find it difficult to understand that the margins with which Zen 2 loses in ST to Comet Lake is much smaller than the margins with which Comet Lake loses to Zen 2 in MT?

 

[TheELF]

And some people have still not learned that GHZ does not equal processing power.

Hi, 2016 called and wanted their meme back. Try benching any Intel Comet Lake-S against an aging 3950X and tell me how they're "competitive". AMD is cheaper? As if.

Be it 2016, 1976 or 2256 work done by how many times you can do it in set time, it's called work activity or power.

It's instructions and it's per cycle
If you have low instructions and high clocks it's not that good,
If you have high instructions and low clocks it's not that good,
having both being high is the best you can do and if you can do that with low power draw even better.

GHZ does not equal processing power but GHZ times instructions does.

 

[TheELF]

On-topic....The point that keeps getting twisted is that GHZ is not equal to processing power. Yes, Intel currently has a small lead in most (not all) games, but a 3950x is not a gaming chip. Gaming is faster IMO on Intel due to latency, not the 5 ghz number. And the 10850k might very well be a decent gaming chip

Games are completely irrelevant at this time for comparing CPUs since even the highest GPUs get bottlenecked by the games themselves.
Wildly different core counts and wildly different Ghz give you very similar FPS.

Now in anything BUT gaming, meaning desktop, HEDT and server uses, Intel is NOT competitive, look at all the benchmarks out there.
...
...
So with both at 4 ghz, single core, a 10900k vs a 3900x lets say, the 3900x wins be a decent margin, so its more efficicient and faster in anything. Now in reality the 10900k can go to 5 ghz, so it more than makes up for that loss in speed in gaming, but looses in everything else.

Yes people, please do look at all of them and not just the click baity ones.
Running at 5Ghz instead of 4Ghz influences everything in the same way,clocks are clocks, running something 20% slower due to IPC(number of cores) but 20% more times in the same time due to Ghz will result in the same thing.
Overall the 10900k is very well balanced to the 3900x ,some things will run better on one other things will run better on the other but overall they are pretty much the same with the difference being the 3900x needing 2 more cores to do the same amount of work and the 10900k needing 20% more clocks to do so.

Intel Core i5 10600K + Core i9 10900K Linux Performance Benchmarks Review - Phoronix

www.phoronix.com

 

[DrMrLordX]

Skewed math is skewed. So, AMD brought down an HEDT chip

The 3950X isn't an HEDT chip. It has the same I/O die as the 3900X. AMD doesn't sell 16c in HEDT anymore.

Be it 2016, 1976 or 2256 work done by how many times you can do it in set time, it's called work activity or power.

It's instructions and it's per cycle
If you have low instructions and high clocks it's not that good,
If you have high instructions and low clocks it's not that good,
having both being high is the best you can do and if you can do that with low power draw even better.

GHZ does not equal processing power but GHZ times instructions does.

I'm not sure how that's rebuttal, but okay? Nice of you to dredge up Phoronix numbers again though. Do we have to go around again about why citing those numbers out of context is essentially useless? Meanwhile, there are plenty of other review sites that follow more-conventional benchmark routines that show the 10900k in a much worse light, but you won't be citing those, will you?

And if I may, can we get back to the point I was trying to make? Intel sells a cheaper alternative to the 10850k called the 9900k which is really what most people should buy unless they want the extra two cores and all the power draw that comes along with actually using them for something. If it weren't for someone opining about how "they're still competitive" we wouldn't be wasting our time discussing AMD chips in this thread.