Intel Core i9-10850K Evaluation: The Actual Intel Flagship

When an organization like Intel creates a CPU design, the method of producing brings about variation on the standard of the product. Some cores will solely attain a sure frequency, whereas others have surprisingly good voltage traits. Two targets of processor design are minimizing this variance, but additionally shifting the height increased, all whereas controlling how a lot of the silicon is definitely useable. That is a part of the magic of ‘binning’, the method of filtering the silicon into totally different ‘bins’ for applicability to a given product. It’s via this course of that the Core i9-10850K exists, albeit reluctantly.

Intel’s Core i9-10850K: Doing The Heavy Lifting

The Core i9-10850K is the entry member to Intel’s Core i9 lineup, with 10 unlocked cores and hyperthreading, and might turbo as much as 5.2 GHz beneath Thermal Velocity Enhance. The added bonus is that it is extra broadly obtainable out there than the Core i9-10900K.

Intel tenth Gen Comet Lake
Core i9 and Core i7
AnandTech Cores Base
Core i9
i9-10900K 10C/20T 3.7 5.1 4.8 5.2 5.3 4.9 125 630 $488
i9-10900KF 10C/20T 3.7 5.1 4.8 5.2 5.3 4.9 125 $472
i9-10900 10C/20T 2.8 5.0 4.5 5.1 5.2 4.6 65 630 $439
i9-10900F 10C/20T 2.8 5.0 4.5 5.1 5.2 4.6 65 $422
i9-10900T 10C/20T 1.9 4.5 3.7 4.6 35 630 $439
i9-10850K 10C/20T 3.6 5.0 4.7 5.1 5.2 4.8 125 630 $453
Core i7
i7-10700K 8C/16T 3.8 5.0 4.7 5.1 125 630 $374
i7-10700KF 8C/16T 3.8 5.0 4.7 5.1 125 $349
i7-10700 8C/16T 2.9 4.7 4.6 4.8 65 630 $323
i7-10700F 8C/16T 2.9 4.7 4.6 4.8 65 $298
i7-10700T 8C/16T 2.0 4.4 3.7 4.5 35 630 $325

As with the opposite 10th Gen Intel Core i9 processors, this one helps two channels of DDR4-2933, makes use of the LGA1200 socket on Intel 400-series motherboards, and has sixteen lanes of PCIe 3.0 for add-in {hardware}. Intel likes to level out it has one other 24 PCIe 3.0 lanes via the chipset, nonetheless that is restricted by the DMI/PCIe 3.0 x4 uplink to the processor.


With the Core i9-10850K, customers are primarily getting the Core i9-10900K high tier mannequin, however at 100 MHz decrease throughout the board. The height turbo is 5.2 GHz moderately than 5.3 GHz, the bottom frequency is 3.6 GHz moderately than 3.7 GHz, and each processors are set at a 125 W TDP. Saving 100 MHz additionally saves $35 from the majority pricing of the processor, however due to the shortage of availability within the 10900K, we’ve seen the distinction between the 2 differ from $50 to $200 in current months. This all feeds into the principle story of what’s going on right here.

Regardless of being a part of Intel 10th Era Core household, the Core i9-10850K was launched after the launch of the extra public and vocal members, such because the 10900K or 10700K. The e-mail concerning the 10850K dropped in our inbox on July 27th, two months after the official launch of the remainder of the household, and whereas Intel didn’t have samples prepared for that launch at present (to be trustworthy, it took us without warning), requests have been lodged and our arrived a couple of weeks later.

Whereas not utterly surreptitious, Intel pushed this processor onto the market with out a lot fanfare.

The Secret Artwork of Binning

Binning is a flowery phrase for high quality administration and filtering – when the silicon is manufactured, a few of it’s higher high quality than others, and by testing the standard every product will be filtered into the place it’s best suited, filtered into ‘bins’. The character of binning isn’t new within the trade by any stretch of the creativeness, as relying on the manufacturing course of high quality can differ wildly, and binning allows a semiconductor firm to take advantage of out of the mounted worth wafer prices. If a given wafer offers 50 processors, solely 10 meet the best high quality stage however one other 35 meet a decrease high quality stage, then the yield is 45 out of fifty, moderately than simply 10, enabling much less waste and arguably higher worth for the tip buyer. Usually when an organization talks about yield, they’re speaking about this 45 out of fifty quantity.

The principle aspect to how this binning manifests is in two types of variability: the variability within the course of and the variability within the design. Manufacturing is vastly advanced, nonetheless the strategies and order of duties within the lithography course of, in addition to the velocity of manufacturing, can have an effect on this variability (it comes all the way down to a whole lot of R&D). Design variability is considerably totally different, because it requires engineers to construct mechanisms right into a processor design to reduce high quality variability, and this may come on the expense of energy or die space.

For a singular firm like Intel, technically they’ll manipulate each of those sources of variability, however for others who depend on foundry manufacturing, it’s a one sided affair and the businesses that pay essentially the most to TSMC (like Apple) get key particulars on the opposite aspect of the equation.

The top purpose is to reduce variability (make every processor off the road attain the identical goal), but additionally to maneuver that variability peak nearer a extra fascinating purpose, resembling efficiency, or energy. The entire course of is a conveyor belt of 10000 levers and switches, the place each can have an effect on the efficiency of a dozen others, and so discovering the very best configuration in a sea of choices will be very tough.

Instance shmoo plots of Samsung’s 5nm Take a look at SoC, displaying easy move/fail at voltage and frequency

Easy move/fail metrics are sometimes graphed in a shmoo plot, just like the one above. Past a move/fail metric, firms like Intel must additionally decide what proportion of the processors on a given wafer or batch meet these necessities. As a result of a graph of variability within the high quality of silicon will be so different, the place the processor firm defines its product binning targets is essential. An organization like Intel must determine what number of processors of every kind it is going to want, what its prospects will want, how that may change over time, and what it could possibly do to maximise gross sales $ per sq. millimeter of silicon. In a state of affairs the place prospects may need a cheaper product, Intel may take increased high quality silicon and label it as a decrease high quality product, so it doesn’t sit on a whole lot of unsold processors. However opposite, if prospects are demanding the next high quality product that manufacturing can’t ship, then it could possibly turn out to be a difficulty. It additionally issues on the subject of advertising as effectively.

A shmoo plot of Intel’s Centrino CPU, with the letters doubtlessly indicating
what % of processors move/fail at these factors.

A semiconductor firm like Intel can do themselves a favor by selecting binning targets and metrics that aren’t as aggressive. Nevertheless it’s actually the high-end halo merchandise that matter on the subject of selling the very best of the very best. Intel has a historical past of very aggressive binning, to the purpose the place its high quality necessities for the highest product are so strict that solely a handful of processors will ever meet that stage for each million produced.

Over the previous twenty years Intel has made tremendous not-so-secret ‘Everest’ or ‘BlackOps’ processor fashions. These are technically off-roadmap processors not for basic sale, due to the very strict high quality necessities. These items are chosen due to the super-high frequency attainable, normally at a very disregard for energy or cooling necessities. One of many first good examples of this was the special-order-only twin core Xeon X5698 rated for 4.4 GHz in Q1 2011, based mostly on Intel’s 32nm Westmere platform, and was constructed solely for high-frequency inventory market merchants who wanted the bottom latency no matter the associated fee. The idea of a microsecond for these merchants will be tens of millions of {dollars}, so throwing $20k+ every for the quickest processor obtainable is chump change (that features OEM markup). These have been 1000 MHz sooner than any of Intel’s usually binned processors for the open market.

Sandy Bridge additionally had an Everest mannequin, whereas for Ivy Bridge it was referred to as BlackOps, providing 6 cores at 4.6 GHz all-core and an enormous 250 W TDP. These have been once more destined for Wall Avenue, however got here with no product identifier, and so far as we are able to inform, no guarantee aside from dead-on-arrival (DOA). These have been such on the fringe of Intel’s manufacturing capabilities that should you wished one, you needed to settle for that it may not work past a few months. Once more, for these merchants, we’re speaking fractions of a p.c of price, in order that was nearly of zero concern.

The one motive we find out about these is as a result of over time some have filtered into the fingers of lovers and collectors. Extra not too long ago, Intel’s newest high-frequency buying and selling processor was a little bit of a doozy. We overheard (and confirmed) at an occasion that Intel was planning to launch an auction-only OEM-only high-performance processor the place it couldn’t assure inventory nor would it not provide any guarantee.

Particular auction-only CPU

This was a 14-core Core i9-9990XE processor, with all cores working at 5.0 GHz on a regular basis, constructed for Intel’s high-end desktop platform (quad channel reminiscence, 48 PCIe lanes). We truly bought one into evaluate via our trade contacts, in a particular single socket server unit and customized cooling. The unit had a peak energy loading someplace north of 600 W. It went effectively above Intel’s 28-core high-end Xeon merchandise. Single threaded efficiency was loopy.

Sadly the Core i9-9990XE didn’t achieve this effectively commercially. The concept was that the OEMs that ended up ‘successful’ these CPUs at Intel’s auctions would construct methods for high-frequency merchants and promote them on that means. A lot of the firms that ‘received’ the processors at Intel’s auctions ended up promoting them as CPU-only fashions as an alternative, after it couldn’t get enough curiosity. Intel has since stopped making these CPUs obtainable, aside from the one firm that’s nonetheless promoting to HFT.

Since then, Intel has launched its shopper grade 8-core Core i9-9900KS, which additionally had all eight cores at 5.0 GHz, albeit with solely twin channel reminiscence and 16 PCIe lanes. This was nonetheless a really restricted version product, and was solely on sale for 4 months earlier than being discontinued. Intel additionally gave away 400 particular binned variations of its 10900K as a part of an influencer giveaway, centered round its new thermoelectric/liquid all-in-one cooler designed to deliver temperatures beneath ambient throughout idle operation. These should not a particular numbered model you should buy, nonetheless.

When Intel (or anybody else) designs a processor, there are a selection of instruments on the firm’s disposal to be able to assist the efficacy and variability of the manufacturing to assist information its merchandise into given high quality bins. Ought to Intel deal with making a really particular HFT processor, then there are 1000 design choices that may be made which might make no sense for a shopper product, and so you’ll be able to see the complexity of attempting to extract the very best efficiency out of the silicon if the majority of the gross sales for that design are going to be for a extra mainstream product. For an instance of a processor constructed for excessive duties, then I level in the direction of IBM’s z-series {hardware}, which begins at 12 cores working 5.2 GHz, and 4 processors in a server share a 960 MB L4 cache chip. It’s a loopy (however enjoyable) rabbit gap of processor design to go down.

Binning: Core i9-10850K

So why spend a lot speaking about particular 1-in-a-million processors? The purpose is that when Intel decides the place to attract the road on silicon high quality, the place it attracts the road for its industrial flagship is essential. It wants to attract the road at a spot the place it has a aggressive product, however also can produce sufficient to fulfill the wants of the market. There’s no level drawing that line for shopper flagship on the 1-in-a-million stage should you can solely promote 10 a month. If reviewers lead with that 1-in-a-million evaluate on launch day however no-one should buy it, ever, then as a model you’ve misdirected your buyer base. At the least with a BlackOps processor most customers perceive they may by no means see one, however for the highest Core i9 product, it must be broadly obtainable – it must be someplace in that 10000-in-a-million stage a minimum of (one may think).

Intel’s 10-core Flagship Comet Lake Household: 10900K, 10850K, W-1290P

Which brings me to the only real motive why the Core i9-10850K exists and why it was launched two months after the principle product line, after the Core i9-10900K. Merely put, Intel drew that high quality line for its high bin processor too aggressive to the purpose the place the corporate couldn’t meet the demand. We’ve by no means had a x50K processor in earlier generations as a result of that drawn line was so aggressive, we’ve solely ever had the road drawn additional away (resembling Satan’s Canyon on Haswell). That is very a lot the primary occasion in current reminiscence I’ve seen a vendor must introduce a processor with extra lax necessities as a result of it couldn’t construct sufficient of the top-end chip.

Within the first six months because the launch of the Core i9-10900K, inventory has been scarce is well-liked markets, and non-existent in secondary markets, with a really sluggish trickle via. That stage has considerably normalized now, however customers who couldn’t buy the 10900K had two choices: (1) wait till they may get one, or (2) purchase one thing else, maybe from a competitor.

So as to bridge that hole, the Core i9-10850K got here out onto the market, the place the binning line was ever so barely extra lax in comparison with the 10900K, and extra silicon may obtain the targets. We’ve seen within the SKU checklist that it seems to be a 100 MHz drop throughout the board, however as we’ll see on this evaluate, the voltage additionally appears to be increased, so it’s burning extra energy to get a decrease frequency.

The Core i9-10850K has been broadly obtainable since launch. In our month-to-month CPU Guides, I’ve seen enough portions of 10850K items with costs hovering across the Intel anticipated pricing (typically cheaper with OEM-only non-packaged 1-year seller-only variations), particularly at a time the place inventory of the 10900K is so restricted that sellers have been charging +$200 premiums for the inventory they did have. At that worth, customers have been 16-core AMD fashions, particularly for a high-performance market the place multi-core efficiency continues to be a key metric.

If Intel had drawn the road on the 10850K stage within the first occasion, the idea of binning wouldn’t be a dialogue at the moment, and there would have been enough inventory on the cabinets since launch. The query on the time was all concerning the message that Intel wished to mission with its top-tier 10-core overclockable mannequin, and the 5.3 GHz mark supplied that message, even when amount of accessible processors was restricted. There’s no level having a BlackOps fashion processor because the flagship if no-one should buy it. For this reason the title of this evaluate is that the Core i9-10850K is the true flagship – it’s the one which has been actively obtainable to buy.

This Evaluation

Intel supplied a Core i9-10850K retail processor for evaluate, and because of this we’ve put it via our benchmark suite. Over the subsequent few pages we’ll cowl CPU efficiency, gaming efficiency, and a few of our microbenchmarks for energy and frequency ramping response. The Core i9-10850K was launched earlier than the Ryzen 5000 processors, nonetheless as each can be found at present, each units of numbers are included.

Take a look at Setup

As per our processor testing coverage, we take a premium class motherboard appropriate for the socket, and equip the system with an appropriate quantity of reminiscence working on the producer’s most supported frequency. That is additionally usually run at JEDEC subtimings the place attainable. It’s famous that some customers should not eager on this coverage, stating that typically the utmost supported frequency is kind of low, or sooner reminiscence is obtainable at an analogous worth, or that the JEDEC speeds will be prohibitive for efficiency. Whereas these feedback make sense, finally only a few customers apply reminiscence profiles (both XMP or different) as they require interplay with the BIOS, and most customers will fall again on JEDEC supported speeds – this consists of house customers in addition to trade who may need to shave off a cent or two from the associated fee or keep inside the margins set by the producer.

Take a look at Setup
Intel LGA1200 Core i9-10900K
Core i9-10850K
Core i7-10700K
ASRock Z490
PG Velocita
Corsair DomRGB
4×8 GB
AMD AM4 Ryzen 9 5900X
Ryzen 7 5800X
Ryzen 5 5600X
X570 Godlike
2×32 GB
GPU Sapphire RX 460 2GB (CPU Exams)
NVIDIA RTX 2080 Ti FE (Gaming Exams)
PSU Corsair AX860i
Corsair AX1200i
Silverstone SST-ST1000-P
SSD Essential MX500 2TB
Extra Cooling supplied by SST-FHP141-VF 173 CFM followers

Many due to…

We should thank the next firms for kindly offering {hardware} for our a number of check beds. A few of this {hardware} isn’t on this check mattress particularly, however is utilized in different testing.

An enormous due to ADATA for the ​AD4U3200716G22-SGN modules for this evaluate. They’re presently the spine of our AMD testing.

Customers within the particulars of our present CPU benchmark suite can discuss with our #CPUOverload article which covers the matters of benchmark automation in addition to what our suite runs and why. We additionally benchmark rather more knowledge than is proven in a typical evaluate, all of which you’ll be able to see in our benchmark database. We name it ‘Bench’, and there’s additionally a hyperlink on the highest of the web site in case you want it for processor comparability sooner or later.

Leave a Reply

Your email address will not be published. Required fields are marked *