Intel’s 11th Gen Core Tiger Lake SoC Detailed | Hacker News Reader

Intel’s 11th Gen Core Tiger Lake SoC Detailed

anandtech.com

118 pointsrbanffy154 comments

I won't be buying Intel CPUs for years to come. Haven't upgraded since 2016. Years of "incremental" improvements while still advertising $1000+ CPU prices has deterred me for quite some time. Intel fanboys might eat it up though.

Given their history of significant delays in the current roadmap, how does this compare to their competition who is quickly advancing on 10nm?

Also what is this internal naming schema?

> The 14nm process node has been Intel’s most profitable manufacturing node to date, and continuous intranode enhancements over the years (14+, 14++, 14+++, 14++++*)

This reads like a satire article.

Everyone I know, including myself, have gone AMD 39xx. Damn what a beast at a reasonable price.

I was very hesitant due to past experiences with AMD over 20 years, especially drivers.

I ended up going with a Ryzen 9 though & it's been flawless so far. Great performance. Wouldn't look at an Intel anytime in the near future for a CPU.

Right? I assembled a number crunching machine for $1200 a few months ago. The 3900X was under $400 at the time. I can’t imagine spending an extra 50% on a comparable machine just say I can say it has Intel Inside.

If I build a new computer to replace my Ivy Bridge hackintosh, the only thing that would keep me on Intel at this point is the desire to build one last hackintosh before Apple goes full ARM. At the middle level, say Ryzen 3700X vs i7-9700, they're pretty close.

Its quite possible to build an AMD-Hackintosh, its quite in stlye really.

depending on your workload you will have some troubles with an amd hackintosh (e.g. logic plugins)

I would absolutely love to do this but AMD is somewhat annoying when running MacOS.

1. Really bad VM support. No Hypervisor framwork, which means problems with Docker (apparently some workarounds for Docker specifically, but not 100% stable). Things like Parallels Desktop, Android VMs, etc won't work.

2. Problems with Adobe suite apps crashing on boot. Some patches exist but last I checked there was nothing for Lightroom (which I use a lot).

3. Things like iMessage, Facetime, Siri, etc won't work

4. Uses a custom kernel and there can be hassles related to power management.

etc.

My use case would have lots of problems, but it's definitely a great option for people who don't need VMs or Adobe.

iMessage works in my experience, not sure where you heard that.

I was upgrading a year or so ago, after the Spectre and Meltdown were discovered and patched by essentially slowing down CPUs. Went with AMD and never looked back.

At this point though intel has slashed their price/core to be competitive if not better than threadripper. They don't have anything that compete in core count, but if you look at the 9980/10980 series with 18 cores, its very competitive on a price basis per core with the 24/32/64 core threadrippers. If you want max cores then you still need to go with threadripper but if 18 is enough, i don't get all the hate 10980 gets. There are certain linear algebra related workloads where the intels excel and why I would probably go with the 10980x were I to upgrade (which im probably not, 4 year old pc doing fine)

Practically every review for the 3960X shows it's significantly faster at productivity applications than a 10980XE. Leaving six cores on the table is significant. If you have a primarily single-threaded task, then you should just go for a 10900K.

>They don't have anything that compete in core count, but if you look at the 9980/10980 series with 18 cores, its very competitive on a price basis per core with the 24/32/64 core threadrippers.

That's not a fair comparison, considering that price per core usually goes up the more cores there are. Also, if you check pcpartpicker you'll see all the i9s with greater than 12 cores are either sold out, or well above their MRSP.

I don't understand why the pricing of Intel's enthusiast CPUs should affect your buying decision if you're not in the market for an enthusiast cpu. That being said, if what you've got is working for you I certainly support the idea of sticking with it.

Being in the market for an enthusiast CPU is predicated on being enthusiastic about CPUs in the first place, and that's become somewhat difficult in the last couple of years.

Indeed. My old self-imposed rule was that upgrading was a waste of money unless I could get at least twice the performance. Manufacturers have generated high benchmarks by throwing "moar cores" at it but even the best chips have barely a 40% single-thread increase over my seven year old Haswell build.

are all your workloads single/lightly threaded? my 9700k compiles c++ two to three times faster than my old 4770k.

40% single thread increase, but most likely 100+% increase in core count. CPU manufacturers did not slap more cores on a CPU to get high benchmark scores, it's just one of the few viable options for increasing CPU performance.

Check top/task manager from time to time. It's pretty rare any desktop applications use more than 2 or 3 cores to the max.

That means those extra cores won't really help day to day performance.

single desktop app probably doesn't but you don't run just a single one. Chrome, IDE, database server, db client, web server, VMs, slack/teams/gitter, email client. Try provisioning a VM with single core and running them all at once inside...

EDIT: that said I wonder if you notice a difference between 4 and 8 cores in normal life

If you’re a dev (as I’m sure many here are), many big builds will max all cores for extended times. I had to up the cooling in my machine as building Chromium would regularly peg all 8 core for hours, causing thermal throttling to kick in and lower clock speeds.

Sometimes I manage to extract a little more performance by shutting down virtual cores to get lower thermal stress and better cache hit ratios. It's a very YMMV thing that depends on pretty much everything from the CPU microarchitecture and built-in power management all the way to the size of the fan, whether you polished the heat spreader on the CPU to the programs you are running and what the data they are processing looks like.

My workloads are more of an outlier when compared to the general PC user crowd, but I do keep an eye on my CPU usage and it is often between 25-100% CPU usage, depending on the situation.

Workloads, where more cores really help out, are:

* re-encoding video files to save disk space (it's amazing what speeds I can reach with my Ryzen 5 2600X here!)

* IntelliJ (startup, indexing)

* running tests (can be configured to be very parallel)

* running JS heavy sites, like Jira (you would be surprised at how resource intensive websites can be nowadays)

* running more than one Electron based app (Spotify, Slack are surprisingly CPU intensive, especially if you have poor/no GPU acceleration support)

Have you looked at what you wrote? 40% increase is MASSIVE. Not only that but multi threaded the increase is in the several hundred percent range.

40% is pretty big, but over 7 years that's not much. It wasn't that long ago that 40% increases were happening every 2 or 3 years.

Gordon Moore would be very disappointed with 40% every 2-3 years. OTOH, he would probably like how GPUs are progressing.

Even though I'd say benchmarking fp16 GFLOPS against fp64 ones is cheating.

Moore's law I think is widely misunderstood as computer chips doubling in speed every two years (or something to that effect). In fact, what it says is that the number of transistors doubles every two years, which is still roughly true. In 2020, its hard to buy single-threaded CPU speed by just throwing more transistors on a chip, which is why most performance gains have been found in multithreaded workloads (because you can trivially get more CPU cores by putting more transistors on a chip).

I would love if I could find where my OS could collect informations on how long my CPUs stall for lack of proper execution ports, lacking parallelism in the code, L1-3 cache misses and so on.

If you're on Linux look at

  perf list

then

  perf top -e [...]

Maybe we should call these "Gaming CPUs", as it seems to be the most common usage they get, by far.

As someone who's really enthusiastic about CPUs, but not at all about gaming (we have things called "workstation" and "playstation" for a reason) I'm really pending towards AMD in the x86 space, because they have some really interesting parts in their EPYC line (and some pretty amazing deals in their APUs). Having said that, if x86 compatibility is not my goal, POWER9 also looks pretty cool too, as do all the ARM-based server chips that popped up recently. I also find the LinuxONE Community Cloud machines to be very interesting (I just wish I could have a zSeries of my own in a 19" server tower chassis). It's really sad the desktop or deskside workstation market these days is mostly x86. It's boring.

If my workflow didn’t depend on proprietary tools from stubborn companies I would be looking to buy high end ARM like ampere for my workstation. But I’m stuck on x86 for now :/

Same here. If I had a reason to justify it, I'd have gone with a POWER9 box from Raptor.

Unfortunately for my home datacenter, everything my Xeon deskside, my i5 Mac and my i3 Dell can't do, I can easily enough spin up a cloud instance and do there. ¯\_(ツ)_/¯

I never understood this idea that someone feels sleighted for something out of their price range that they perceive should be valued lower for their benefit while complaining that the company is setting up prices for their benefit too. Is it grounded on facts or mostly an emotional response?

A very succinct description of a common thought pattern. You see it frequently applied to premium brands like Apple and BMW.

"It's not FOR you," he sniffed, nose pointed precisely at 10 o'clock.

See every single thread about a luxury/clearly low run smartphone.

I see this all the time, in every area. Bicycle forums are filled with hate for "casual bikers" who buy $5,000 bicycles and all the expensive accessories. My feeling is "good for them! They help the industry and help finance the lower-end models."

Some people are just too petty to see the "enthusiast" market also drives the low-end.

> They help the industry and help finance the lower-end models

Do they though? From how most companies operate that I've seen from the inside, I would assume that such a market response would shift their focus towards high-end models, and low-end models will be neglected.

Sure, there is somewhat of a technology trickle down towards lower-end models over time, but that seems to be driven on an industry/supplier/competition level and mostly disconnected from luxury consumers. Especially "_financing_ the lower-end models" seems unlikely, as no manufacturer would lower the prices on the lower-end models, just because they make "enough" on the high-end ones. The only way this would apply would be through shared overhead, which given that we are talking about different product lines often tends to be minimal.

> The only way this would apply would be through shared overhead, which given that we are talking about different product lines often tends to be minimal.

My understanding is that in the CPU case, the lower product tiers are literally just higher tier chips that had faults in some cores/areas and had those features disabled.

> The only way this would apply would be through shared overhead, which given that we are talking about different product lines often tends to be minimal.

this is very much not minimal in the CPU business though. intel's entire line of (consumer) desktop parts is different bins of the same chip. they wouldn't be able to offer an i3 for $120 if they weren't also selling i9s for $500. and they probably couldn't sell consumer parts for <$1000 at all if they couldn't sell xeon parts with the same architecture for thousands of dollars.

> intel's entire line of (consumer) desktop parts is different bins of the same chip.

Is that true? Intel's 14nm yields are high enough that it wouldnt really make sense for the dirt cheap quad core processors to secretly be 8 core processors with half the cores (and cache) disabled.

I think for the current gen parts, the quad core parts are actually six core dies with two disabled. the overall point stands though.

This doesn't seem to apply to the CPU industry, where Intel has been charging these high prices for the enthusiast products and then not driving innovation in the midrange, due to having a monopoly position partly brought about by previous anticompetitive behaviour on their part. Is that similar to the bicycle industry?

I am a heavy gamer, both table and electronic.

In our market at least many companies gunning for the enthusiasts are completely screwing the normal users, not helping them in any way.

1. MTG fans been complaining a for years that some cards have not been reprinted and are literally essential to the game, new players need those cards. The manufacturer reaction was launch a product that does have those cards but in a sky high price point, when players complained they literally replies that the product wasn't for them, it was aimed at the luxury market. How that helps the low end? Not only just the possible buyers, but current players that can't have new play partners because new play partners can't afford to hunt on secondary market for basically mandatory game pieces.

2. Ana Android F2P game I play decided to go full pay to win while denying it, they launched a new piece of armor that makes your character drastically better on ranked matches, and then proceeded to make it cost a currency that the free players can get only half of the price of the piece of armor, to get the rest of the currency needed you need to pay some 200 USD.

When players complained the main dev wrote a extremely rude answer on the forums that also made clear all he cares about is getting money from big spenders, the currency to get the item in question is not available to even the vast majority of paying players, not even subscription players can get it, you need to buy that separate expensive pack. As for why people care about the ranking? Well, the reward for doing well is permanent character buffs, that there is no way to get in other manner, thus reinforcing the pay to win aspect.

3. There is a certain game publisher that found out profits can be made with DLC, now they launch then at breakneck speed AND lock essential features behind them. Whenever they update the game alongside DLC they balance the game in a way that it is not fun without the DLC, making them mandatory. I found out even UI buttons get shut down (example: one patch fixed AI notorious stupidity by letting players ask allies to follow them instead of wandering around in stupid manner. But the button that you click to do so, is part of the DLC. Or in other case sometimes you have a huge list of things to decide, you need to click on each item one by one, or but the DLC that adds a button on the interface to do the same action to the entire list). Another case with same publisher: updates literally removed features of the game, added new mechanics to compensate, but those are DLC.

Exorbitant prices on warmed-over revisions of yesterday's product in a market devoid of competition broadcasts "we've stopped trying because we own the market".

Yes, that can trigger an emotional response in people.

>we've stopped trying because we own the market

I always laugh at that take. Imagine working for Intel today and hear people complain you're not even trying.

Suddenly trying very hard after years of stagnation isn’t the same though. They have generated a lot of ill feeling and this will take years to overcome.

I read the comment to mean that Intel isn't worth the higher price compared to AMD these days. How does someone saying they won't buy Intel CPUs translate to a complaint that the company is setting up prices for their benefit?

There is a lot of frustration with Intel over their pricing structure. Intel has done an excellent job at maximizing cash extraction.

Whatever the logic, humans do not like the feeling of being taken advantage of, and when your counterparty is obviously maximizing how much they extract from you, it feels bad.

The funny part is lots of people like to make themselves feel better by condescendingly "educating" people who feel that way. It is poor emotional regulation all the way down.

Meanwhile amd has been cleaning their clock for decades at a better price point.

Literally there were people born that can legally drink now. Intel is the grey goose of chips. Amd gets the job done

My first CPU at age ~8 was a Pentium Pro, followed by a Pentium II. That was the last Intel CPU I ever owned. I stuck on an FX8350 for a while, and it was never the bottleneck for gaming performane compared to the GPU.

For a non-enthusiast, AMD has been a great choice almost for my whole life.

I've gone back and forth over the last 3 decades. My dad has been an employee at Intel since 1995, so I started with the original Pentium and basically upgraded my CPU every year from the 90s to the mid 2000s. I don't engage in fanboyism, I just like to have the best bang for my buck. So I've had a few opportunities to try AMD CPUs - mainly the Athlon series (Thunderbird, XP, and 64 series). My biggest issue with the AMD CPUs was always heat management; those chips got HOT. In laptops they were almost unusable after a few months because they would get so hot, components would start to melt. My first AMD laptop I had to send back after 3 months due to the heat issue. It cost me $99 to get it repaired and a 6-week wait time. That left a bitter taste in my mouth and put me off of AMD. Do you know if the heat issues are still prominent? I wouldn't mind giving AMD another chance

At various times in their history where AMD was behind in core design, manufacturing, or both, they cranked power (and thus heat) up to fairly obscene levels to attempt to compete with Intel. That's not the case now with Zen and TSMC 7nm - you cant really compare an AMD CPU now to one even 5 years old, much less 20 years old.

For laptops this generation (Intel Ice Lake vs AMD Renoir), there are now several reviews of the "same" laptop chassis with both versions:

Lenovo IdeaPad S540-13: https://next.lab501.ro/notebook/english-lenovo-ideapad-s540-...

Lenovo ThinkPad 14s: https://www.notebookcheck.net/Lenovo-ThinkPad-T14s-Review-Bu...

XMG Gaming Laptop comparison: https://www.techspot.com/review/2075-same-laptop-different-c...

TLDR: AMD significantly outperforms Intel in absolute performance and performance per watt; battery life is about equal - depends on implementation and how its measured (Intel is still a little better on idle, video playback, but much worse under load), temps usually limited by laptop's thermal solution, but under high TDP, favors AMD.

The Tiger Lake U chips should be pretty competitive with Renoir at 15W, especially as there are almost no AMD options paired with 16:10 or HiDPI displays - we'll have to see when they come out. Tiger Lake H isn't expected until Q1 2021 and will have to compete with Cezanne (Zen 3 core, which is expected to have some pretty big IPC gains).

Awesome, that's very useful info. Thanks for taking the time to link these and elaborate on the data

I actually like the P2 and working with smp clerons. But the AMD dx was a nice chip too

Intel's CPUs today are still king when it comes to single-thread workloads, like gaming. The main difference in the past two years has been that AMD has taken the lead in multi-thread performance, and that's scared Intel so much that they've significantly improved multi-thread performance and price across the board.

The difference is quite stark, in both directions [1]: AMD destroys Intel in anything that benefits from multiple cores at the same product tier (Ryzen 9 3900x vs i9 10900k), in some workloads as high as 50%. But, Intel wins when gaming; usually by around 5-10%.

The more measured position is: If you're just looking for the best value brand right now, AMD is the way to go, even if you are gaming. The 10900k is better than the 3900x for gaming, but its also a ~$120 up-sell: At similar pricing tiers, the single-thread performance difference is likely quite similar. But if you just want the best at any price, that's still Intel: More cores won't save every workload.

[1] https://www.pcworld.com/article/3543993/intel-10th-gen-revie...

The whole situation boils down to price.

AMD is a better value given the price. If Intel wants, they can just cut and match the prices.

Why is there so much callous fanboyism around/against a goddamn CPU supplier like they’re a statement of their lifestyle or such?

Its ok to do due diligence before purchasing but relentless obsession over my red team is better than your blue team screams of immaturity.

Same with the people that endlessly pixel peep and read MTF charts for days in photography. They spend more time talking about how great their camera and lens setup is than actually going out and taking pictures with it.

It is a hobby in itself. It’s irritating to others because the true hobby is hidden under the guise of another hobby.

The irritation comes from a mismatch of expectations around the end goal between those who are fanboys and those who just want to get stuff done.

As a film camera guy, I can say that I am unashamedly more of a camera hobbyist than a photography hobbyist. I mostly take pictures to keep the mechanisms operational.

I think a lot of the AMD hype for enthusiasts comes from the fact that without AMD being a meaningful player in the bulldozer years, desktop CPUs became VERY boring. 4-cores and single digit performance increases from Intel at enthusiast prices for approaching a decade. All you have to do is look at a youtube channel like Linus Tech Tips to see that a lot of people find it interesting when there is competition.

Encouraging competition can be done without hurling callous statements and consumer entitlement.

As a retail business owner, I find 80% of our consumers sound and reasonable. The 20% - we never want their business. We keep our best to not anger them and most of our energy goes into maintaining the public image of our company.

There are some real asshole customers that do things you never want to hear - stuff that makes me lose hope in humanity.

By the time you subtract all the sub-prime performance from their vulnerabilities, this isn't actually true on properly patched operating systems. They took a sub-prime loan on validation, sacrificing security in the process, and the result is chips marketed and sold as something they are not. There really should be government-level lawsuit and restructuring at intel right now.

I looked for but couldn’t find recent benchmarks showing performance degradation on windows and macOS related to mitigating Spectre and Zombieload / Z2.

Can you point at recent information on this?

It's worse, isn't it?

From what I've heard, Intel likes to run their benchmarks with many (most?) mitigations disabled... but Windows ships with all of them enabled.

Any modern (ie 2020) third-party benchmarks ran against these processors have the vulnerability patches applied. They're factored in.

Yes, Intel will often cite performance characteristics without the patches applied, and shame on them for that. And, Yes, it would be interesting to know how much of Intel's performance lead pre-Ryzen was due to this bad speculative execution microcode; it seems likely that AMD wasn't as far behind as everyone thought, but the past is the past.

But the debt is now paid, and Intel still leads in single-core performance. Really, they've never not led in single-core performance (maybe going back to the late-90s Athlon days they were trailing). That's something an Intel shill would tell you, but it's also something anyone can find out just by looking at the benchmarks or doing their own testing. Marketing is mostly lies, but those lies stop when I hold the silicon in my hands.

I have a feeling we will see a bunch of AMD vulnerabilities revealed in the coming years. Researchers haven’t really had time to study AMD’s micro architecture yet and it was not really a priority because of low penetration in the server market. In theory intel should have less bugs since they have massive design verification teams.

Yes, the idea AMD is more secure than Intel is mostly a myth. AMD was less aggressively optimised than Intel but AMD didn't know about side channel attacks either and would have eventually ended up with the same techniques.

Meanwhile, we can compare the security track record of SEV vs SGX to see some real meat. SEV has been cracked completely and repeatedly by simple C or cryptography programming errors like buffer overflows or not checking for points being on the curve. SGX has had no such errors, every single attack on it has been an exotic statistical side channel attack. The last round weren't even practical. Researchers studied it for a year and couldn't actually make it work against real enclaves, but Intel issued CVEs and mitigations for it anyway, just in case. And the root level of SGX was never cracked, which is why it's always been software patchable/renewable without needing new silicon. AMD SEV has needed new silicon more than once.

SGX has held up pretty well given it was born just before the discovery of an entire class of CPU design vulnerabilities.

> But, Intel wins when gaming; usually by around 5-10%.

10% more CPU performance is going to provide overall gains that are difficult to measure. That's even more true for games, which nowadays are seldom CPU-bound.

It's a complex equation, but in general, if you have an RTX 2080 Ti, you'll get a higher average FPS out of an i9 10900k than a Ryzen 9 3900x across most games.

Of course, that "more performance" often means "Running at 300fps instead of 270 fps", and it really doesn't matter. And, the higher the resolution, the more GPU stress you'll see, so at 4K-class resolutions its far closer to trading between CPU-bound and GPU-bound.

But, then, DLSS2 complicates things further by amplifying what the GPU is capable of at even very high resolutions, and the RTX 3080 Ti is releasing in a couple weeks.

So its only possible to speak in generalities about facts which are true in most situations. In the past decade, there has not been any competitively positioned products which would make an Intel/Nvidia PC not the generally-best option for PC gaming. AMD deserves a ton of credit for what they've done with Ryzen, and outside of gaming, BUY RYZEN. But, within gaming, it's a far more complex equation which depends on how much you're spending, what you're using the PC for other than gaming, if you're streaming, etc: In short, its such a complex equation that its better to just not think about it and flip a coin.

> But, Intel wins when gaming; usually by around 5-10%.

It's measurable but is it significant ? True cpu bottleneck is a thing of the past for 99.9% of the games and 99.9% of people. By the time you're cpu limited you're in triple digits fps territory in 1440p. Even 5 years old cpus aren't that far from today's best: https://www.guru3d.com/articles_pages/amd_ryzen_7_3700x_ryze...

Yup, I play Overwatch at ~100fps at 1440p on an intel 2500k from 2011. CPU might matter for other games, but not the ones I play.

> The 10900k is better than the 3900x for gaming, but its also a ~$120 up-sell

Is it correct to think that compared by work-done-by-watt AMD is currently better than Intel? (in this example Intel's 10900k uses 250-300W [1], AMD's 3900x uses 140-170W [2]?)

In my case I use often many threads/processes (generated by databases and/or my Python programs that usually spawn as many subprocesses as possible without overloading the CPU) and as my PC/server is next to my desk then if the performance difference is not too great my priority is usually for a CPU that gets less hot/is more efficient (because the fans have to work less therefore the box is less noisy, additionally the room heats up less which makes me suffer less during summer).

[1]: https://www.tomshardware.com/reviews/intel-core-i9-10900k-cp...

[2]: https://www.tomshardware.com/reviews/ryzen-9-3900x-7-3700x-r...

Thx

Yes: In terms of performance per watt, AMD generally beats Intel. It's not really even close in many situations. What this means for consumers is more complex, as there are many variables you can try to optimize for.

If you're after minimal power usage in order to optimize for electricity cost, while still keeping good performance, actually: mobile Intel CPUs tend to be the best pick. AMD has some strong mobile CPU options, but they're essentially impossible to build into a desktop PC, whereas mobile Intel CPUs are available in products like the Ghost Canyon NUCs (ok, well, "available" may not be the right word, because they're freakin' impossible to find right now, but Officially, anyone can buy them).

If you're after lowering heat output for the purposes of reducing noise or increasing overclocking performance, generally its better to solve this via better cooling. Noise is more of a function of bad cooling, rather than super-high heat output from the processor. Many people view liquid cooling as "super high overclocking performance", but actually: The biggest benefit to liquid cooling is that it reduces "air churn" inside the case by removing the typical CPU radiator fan just sitting in the middle of the case spilling air in every direction. Instead, you've got dedicated intake/output fans on defined sides of the case, resulting in clear air paths, which increases fan efficiency, reducing the number of fans you need or reducing the amount of work they need to do.

But: the laws of thermodynamics state that heat can't just disappear. If you're optimizing for actual heat reduction for the sheer purpose of heat reduction, then an Intel mobile CPU or Ryzen desktop CPU is better (with Intel mobile CPUs being better by far).

Check out the Ghost Canyon NUCs. If you can find one, they're more expensive than just DIY, but would likely be a great option for you.

Interesting - thank you

Gaming has not been "single-thread" for a while, tbh. Multi-thread workloads are going to be increasingly relevant in the future, as even low-to-middle end cpu's will ship with a sizeable number of cores and support for GPU compute acceleration, e.g. via Vulkan.

Nice background. Thank you. What's interesting to me is that the world's "leading" chipmaker got caught off-guard by lacking multithreading performance. You'd think that these world experts wouldn't have that blind spot.

> single-thread workloads, like gaming.

Most games these days have multi-thread heavy workloads. Gaming hasn't been single-thread dominant for AAA titles for a while.

Just a note for anyone reading through, AMD Ryzen 9 3900X and Intel i9 10900K are both near the top end of the respective companies' ranges of processors. These are not required for modern AAA gaming.

The current guidance for "good" gaming CPUs is in the AMD Ryzen 5 3600 and Intel i5 10600 range (+/- "X" or "K"). The 3900X and 10900K mentioned above will outperform these two, but it is unlikely to be a significant limiting factor for most gamers. At this level, Intel still beats AMD for gaming performance.

Finally, if you're looking to go for a more budget build, the AMD Ryzen 3 3300X is a very solid entry-level gaming CPU. Based on my observations, most guidance suggests this CPU for lower-cost builds than can support an i5 or Ryzen 5.

If you're looking to make some educated decisions between CPUs, always look for benchmarks. Gamers Nexus [0] does very good CPU reviews, with a mix of gaming and productivity workloads. If you're leaning more production workloads, Phoronix [1] has one of the most comprehensive test suites out there.

[0]: https://www.youtube.com/user/GamersNexus

[1]: https://www.phoronix.com/scan.php?page=category&item=Process...

Traditionally Intel's "nm" measurements have always deviated a bit from their competitors, ala TSMC and Samsung. In short Intel 10nm ~ TSMC 7nm and Intel 7nm ~ TSMC 5nm. But with the recent announcement that the 7nm will be delayed, it means TSMC will for sure take the node crown. https://en.wikichip.org/wiki/7_nm_lithography_process

I think the most notable is part is that Tiger Lake will be a minimal CPU improvement (Sunny Cove ~ Willow Cove cores), and the main selling point will be the Xe graphics and support for Thunderbolt 4, LPDDR5, etc. Superfin seems to be marketing hype but will see when benchmarks come out.

The main question is, will Tiger Lake will enough for next year? AMD's Renoir follow up is reportedly retaining the Vega graphics but will further improve the CPU microarchitecture from Zen2 to Zen3. Techspot just published an apples-to-apples comparison that shows, if all things equal except the CPU, the AMD 4800H is 10-60% faster than the i7 10750H...that's a huge chasm. https://www.techspot.com/review/2075-same-laptop-different-c...

>'Superfin seems to be marketing hype' Superfin is the main reason why Tiger Lake is getting a 10-20% frequency boost over Ice Lake. That SuperMIM cap isn't just for show.

Oh snap you're Ian Cutress from Anandtech!

Point taken, mine is that SuperFin isn't some node improvement so TBD on what performance gains it accounts for.

Analysis by SemiAccurate on SuperFin: https://semiaccurate.com/2020/08/13/intel-talks-about-10nm-s...

I have a Intel i5 2500 that I paid $209 for in April 2011. I'm looking at new i9 desktop chips and they're like 50% faster. 10 years? WTF? The laptop chips are barely faster.

https://cpu.userbenchmark.com/Compare/Intel-Core-i9-9900-vs-...

It would be interesting to see when these benchmarks were taken. I have a feeling that most of the i5 Benchmarks are from prior to the Spectre and Meltdown mitigations.

Things have indeed slowed down but you're cherry picking numbers. +50% single-thread but +200% in multi-threaded!

Agreed but I dont think I do much that would keep more than 4 cores maxxed out.

I consistently recommend Hennessy and Patterson's Turing Award Talk[1] as a good overview of the past, present and future of chip performance.

[1] - This will take you directly to the meat of the discussion: https://www.youtube.com/watch?v=3LVeEjsn8Ts&t=28m10s

We've reached a point where you really can't get that much more frequency out of chips, and there's only so much you can do to attempt to get more done per clock cycle. Especially now that its made clear that much of these attempts lead to the various speculative execution bugs like spectre that have been discovered over the past few years.

The days where clock speeds on processors was doubling every year or two are long gone, and its not because chip makers have gotten lazy.

What about this benchmark?

https://www.cpubenchmark.net/compare/Intel-i9-9900-vs-Intel-...

Aside from the pressure from AMD, I think there's another force at play here.

There are quite a lot of people now that just use a web browser on their desktop or laptop. For them, a 5 year old (or older?) CPU is good enough that they don't know they are missing anything. Especially if they have enough memory and an SSD.

> There are quite a lot of people now that just use a web browser on their desktop or laptop. For them, a 5 year old (or older?) CPU is good enough that they don't know they are missing anything. Especially if they have enough memory and an SSD.

My 2015 laptop is still more than capable of doing .net development, or I should say, was when I last tried in 2019.

At work I've got a 2015 macbook pro and it's also more than enough for devops (namely, terraform, docker and the odd script in bash or python)

I should add that neither feel sluggish though no doubt new kit would be faster

I agree that most computing can be done easily with 5 year old hardware, or 10 or 15 if you pick your software carefully.

But I do know that whenever my 2019 mbp is heating up, it's more likely than not to be the browser (with a rogue tab).

I built a desktop rig for my dad back in 2013 with an i3-3220. Seven years later it still works just fine for his needs. The only thing I changed over the years was to add a bit more RAM. I don't plan on updating it unless the CPU/mobo dies.

Yeah - it's great. I'm rocking a 2008 Nehalem at home, having since doubled the memory to 12GB, upgraded the GPU and replaced a failed PSU. It runs like a champ and my non-techie friends think I'm kidding when I say it's more than 10 years old.

It’s because an equally priced 2020 intel cpu is at most twice as fast as your Nehalem (now you can get a 6core for the same money you got your 4core). Per core performance improvement is not even close to twice as fast.

Yeah this absolutely makes a huge difference. The upgrade cycle for laptops and desktops has gotten long. When things have been slow for me at least it almost always memory bound and not cpu bound at this point as well.

I think you can argue that may be caused by performance stagnation. When the CPU market is not advancing in performance, software is not going to advance in computational requirements.

You can see a counter-example in mobile phones. Their processors are continuing to make considerable strides in power-constrained performance, and 2-3 year old phones regularly wind up getting perceptually slower as the software advances in performance requirements.

> There are quite a lot of people now that just use a web browser on their desktop or laptop.

As opposed to the people who used to use their computer for email and spread sheets?

I would argue that the resource needs of a web browser has absolutely been increasing over time. The 4GB of RAM in that old laptop just does not work for the modern web. And considering how many web site now require the client to run tons and tons of JS just to work, I can see a need to refresh your CPU regularly.

4GB of RAM on a machine (regardless of OS) is absolutely fine for browsing the "modern web".

That is the opposite of what I see on a daily basis managing several 100+ SMB Windows based computer networks. I'm curious what your summary is to back up 4GB being absolutely fine.

> Windows

There's your problem!

Windows alone uses like 2GB, open some tabs and you are already out of ram. I think 8 is the new minimum.

I have 30 tabs open in Chrome, real tabs, on youtube, PDFs, and code/API docs, and I just cycled through them all to make sure they are loaded, and Chrome is at 600MB.

The phase "open some tabs and you are already out of ram" is a bit of an exaggeration.

Maybe, if you have only one tab open

Spreadsheets and some other business software can get really gnarly though (especially on RAM usage). Outlook itself can be quite heavy.

But I also agree, browsers and web apps aren't exactly lightweight either. Maybe WASM adoption can help on this front?

I don't believe I claimed that 4GB RAM was enough. Just that a 5 year old (and perhaps older) CPU was good enough.

I mentioned web browser usage as opposed to more demanding loads like gaming, content creation, software development, multiple office tools being used at the same time alongside the browser, etc.

> The 4GB of RAM in that old laptop just does not work for the modern web

Completely false, I'm on 4GB RAM and Q9300. I haven't had any issues. 99% of websites run smooth. I even code on this machine. See my comment here: https://news.ycombinator.com/item?id=24145436

EDIT: I have 10 tabs open right now in Firefox. From bbc.com, HackerNews, and Zillow. I also have AnyDesk, Slack, Signal, VSCode, and a couple of other programs open right now. No performance issues.

EDIT2: I opened up Chrome with Youtube in addition to the above, here's proof everything is still doing well: https://imgur.com/a/SlxT4yE

EDIT3: Ok now i'm just having fun. I added a Netflix tab to Chrome. Opened up a big project in VSCode. Added Vivaldi browser, Notepad++. It stuttered a little (5-10s) loading up the programs but once they were up, it's fine again: https://imgur.com/a/BNqPKU9

If only 4GB of RAM would suffice on my Windows 10 SOE. That thing has 20GB of RAM And a 2018 i5 and still struggles. MS SCCM and all the other 'protection' destroys any ability to be productive.

Software tends to fill the space it lives in. I have seen websites that are more taxing on my CPU than video games.

I have no issues using my 2013 Macbook Pro for 95% of the tasks. For everything else I can run it on cloud. This is also why older Macbooks retain value with a solid price floor even after seven years.

Checking in. My main desktop - which is the machine I use 95% of the time for personal use (other than my phone) - is a Q9300 w/ 4GB of RAM. It's about 13 years old. I use it mostly for browsing and coding w/ VS Code; both of which it's still perfectly good at doing. I game on the PS4, so I haven't needed to upgrade my desktop in a long time. That said, I've been eyeing VR-ready desktops for the last couple of years, but those graphics card prices are hard to justify.

I have several laptops that are newer (within the last 2-5 years), but they're used primarily by my kids or for work.

When i read comments here i'm wondering how and why people are either for or against intel?

IF i need to upgrade, i will check the current situation and will buy the best cpu in comparison to my needs and price.

If it means that i need tons of core because of my workload i might choose a threadripper. Do i need to compile often and i know a specific intel cpu will save me ton of time waiting but it costs more? Fine.

Do i wanna play games? lets see what i can get for ~150$ bucks range.

/shrug

There's no one particular reason, I think, it's just yet another thing to show brand loyalty for.

I'm fairly certain every single person has used or currently uses an Intel powered machine, so choosing AMD has a bit of caché to it (ha ha get it), because you're supporting an underdog, and as it so happens, that underdog has been kicking Intel's butt as of late.

When you're a business, you can't make turn-on-a-dime decisions like that. You commit to business-equipment purchase/leasing/upgrade agreements with a particular OEM; who in turn usually commits (for economy-of-scale reasons) to one CPU supplier or the other. So, if you're stuck with an OEM who's in turn stuck with "the wrong" CPU supplier (right now Intel), that can make you cross, at the potential performance that's being left on the table.

But even for the individual, unless you're buying sealed non-upgradable appliance devices, you've still gotta consider the fact that in building a PC, you're choosing a motherboard socket, and thus potentially making it cheap to upgrade to later-gen CPUs that stay compatible with that socket.

In that mindset, it makes sense to be happy with the CPU maker you're "stuck with" for a while when they provide good, high-ROI upgrade options; and to be angry with them when they don't.

>in building a PC, you're choosing a motherboard socket, and thus potentially making it cheap to upgrade to later-gen CPUs that stay compatible with that socket.

I'm curious how common it is for people to upgrade just the CPU or motherboard.

I tend to upgrade / replace my personal PC's every 4-6 years. At that pace, it's always seemed worthwhile to upgrade both motherboard and CPU at the same time, because of platform improvements. But maybe I'm an outlier.

One oft-stated reasoning is that people who currently don't have much money (e.g. college students) can build a PC with a motherboard socket targeted by both low-end and high-end CPUs; and then select, for now, the (cheap) low-end CPU. Then, a few years later, when they're working and have more money, they can replace it with the high-end version.

I'm not sure how common this actually is in practice, but it seems logical.

Probably not too common. The only people I‘ve seen do this were enthusiasts that had high end gaming rigs and spent a lot of money on their computers. It probably would’ve been cheaper for them to replace everything every 3-4 years. But it‘s their hobby and who am I to judge?

it's great if you decide to upgrade your CPU a couple years later and you motherboard is still compatible, but imo this is not a good reason to choose a particular platform in the first place. you never know when the CPU manufacturer is going to drop support for that chipset. even if the new CPU does support the old chipset, you might be sacrificing some new features or leaving some performance on the table. plus if I'm going to buy a high-end cpu, I'd like to pair it with a high quality power delivery also, which you probably aren't going to find in a college student's budget build.

They usually don't drop support for chipsets. They move to a new socket.

AMD is pretty good about sticking with motherboard sockets and maintaining firmware compatibility. The motherboard I got with my 1st generation Ryzen is handling my 3rd generation chip just fine, though admittedly I'm missing out on PCIe 4.

I had a time in my life where the price of a motherboard mattered. But you can't tell me, that most comments here on HN are from poor students which will not be able to afford a new CPU+Mainboard after 4 years or so.

I have not had this issue to be honest, i just switch both if required. Especially with PCI4 etc. its often enough critical to upgrade after a while and i did sell my old motherboard and cpu combi. That should reduce the 'in my opinion overstated risk' of wasting your old mainboard.

Asides from 486 to Pentium III, I've been using AMD since the K7. Intel has been the 800lb gorilla with lots of questionable anti competitive business practices. For the most part Intel already has market dominance and better performance so engaging in such to such underhanded behavior to secure their position doesn't sit well with me.

I'm also an ATi fan and since they merged with AMD it doubles me down on loyalty.

With Ryzen and Vega/Navi it's a lot easier to recommend AMD products. While they might not have the peak absolute performance their value is much better than the other alternatives. Unlike when AMD went with the Bulldozer architecture, I always died a little inside when I had to recommended an Intel Core i3/i5 over AMD's offerings.

Does anyone know where Intel's codenames come from? They're... supremely weird.

"Ivy Bridge" and "Skylake" evoke images of fantastical locations—chasms transversal only by green bridges woven of ivy, or serene bodies of water that float among the clouds. "Willow Cove" is similarly suggestive, if somewhat more mundane.

But, Tiger Lake? Coffee Lake? SuperFin? Is there a formula here?

> "We have a formal process for what types of things are lakes vs. bays vs. peaks," Tripp told us. However, he didn't elaborate on what that process was. "Different components are named after different geographic areas."

https://www.tomshardware.com/news/decoding-intel-code-names

Just about every product since 2015 has been a lake. I'd guess "lake" is the code word for "just overclock last year's chip and call it a day"

Yeah, each subsequent "lake" was just "Skylake, maybe with a few uArch tweaks, and on a more-optimized 14nm node, and/or sold pre-overclocked as high as we can make them run."

The CPUs have been lakes. Their memory and storage division uses stream, pass and -dale. Server chipsets are -burg, Ethernet controllers are -ville. All of these suffixes are still in active use for new or upcoming products.

Intel advertised a bunch of fancy transistors and process improvements well after it was clear 10nm as a whole was a bust (COAG, Single Diffusion Break, probably more)

Color me skeptical.

Charlie, the arch Intel skeptic, was convinced and if he thinks they managed to fix their problems with 10nm I'm willing to give them the benefit of the doubt.

https://semiaccurate.com/2020/08/13/intel-talks-about-10nm-s...

I'll believe it when I can finally buy 10nm at retail in volume.

https://asia.nikkei.com/Business/China-tech/China-hires-over...

"As a company, TSMC competes to our fullest within the law, but we do not slander our competitors and we respect the intellectual property rights of others. Similarly, we expect our suppliers and other companies to respect TSMC's intellectual property rights and will take appropriate protective actions."

I remember a Taiwanese news said TSMC ask their suppliers not to sell TSMC's custom machines to others.

a while back on Hacker news someone said TSMC have no choice but to build a fab in the U.S. because TSMC needs suppliers like Applied Materials but it doesn't seem that way because everyone can have the same suppliers but Intel and Samsung is struggling while TSMC charge ahead.

https://wccftech.com/samsung-lost-5nm-snapdragon-x60-snapdra....

Is Intel deliberately making its branding and specs as confusing as possible now?

What do you mean "now"?

So this CPU is about to compete with AMD Ryzen 4000 laptop series?

It will be competing against the AMD Ryzen 4000, and most likely losing that competition.

I'd take the other side of that bet. According to the article and other related ones, both the Xe graphics architecture and the 10++ nodes are substantial bumps.

But it'll be a close race, which is a piece of awesomeness we've never had in laptops before.

The Ryzen 4000 is a 2019 vintage. Slightly losing against Intel's 2021 design is no shame. Comparing Tiger Lake against the Ryzen 5000 is a fairer comparison.

And then come the questions of volume -- if Intel can't manufacture in volume they can win on paper and lose on the ground.

AMD has little stopping them from moving to 5nm or 6nm TSMC next year, possibly sooner than Intel can ship this. AMD could do a small run to win on both paper and ground.

The real factor here is the third competitor, Apple, walking in with what is expected to be a 5nm 'no cost limits' CPU that they will ship on every machine based on A14 architecture. Because they own the design and just have TSMC do Fab it's likely that Apple will intentionally put what would be a $500-$1000+ monster core-count cpu spec in every base product, stealing the war. and possibly deflating the market for victors here.

Zen 3 will likely ship before Tiger Lake at least on EPYC, but it's TSMC N7+, Zen 4 is 5nm Samsung and/or TSMC but 2022.

I do agree that Apple will likely take a hit to their margin if that's necessary to ensure that their new ARM laptops are clearly more performant per $ than their outgoing x86 ones. Whether that will be a significant spoiler is an interesting question.

I don't see that happening. Apple needs to make money and they do so by having huge margins. They can certainly increase the CPU budget by taking out the Intel cut (though really they've just swapped that for the TSMC cut), but they're not going to chuck a $1000+ CPU in a ~$1200 laptop.

This may be my lack of knowledge in CPU manufacturing, but from what I understand larger dies get you more performance not more efficiency, so considering their inability to build adequate cooling solutions and fondness for reducing thickness even if they chucked a 32core bohemeth in there it won't be running very fast.

On top of that, since AMD are using the same nodes I don't really see how they wouldn't be able to compete. Especially considering Apple probably won't be selling their CPUs separately, leaving out large parts of the market.

So by Mega-core-count I mean more than 10 and I'm including their big.little architecture (so maybe 6+ high performance, 8 efficiency, remember that Apple can use all cores grouped however, asymmetrical).

The chip design and margin advantages here are massive. Apple sell's phones at $399 with the A13 (7nm) that outperforms (by any metric we currently have), some of Intels > $300 notebook processors, building using iPhone cores as a base means your margins are subsidized by one of the largest selling products and most massive R&D budgets in the world. Then you have the efficiency of all of this: sure, larger Dies won't get better efficiency but Apple's chips use massively modular power operations. They can bring cores on and offline dynamically, you can chop the GPU, core by core, when you don't need it. They could (important: could) ship chips that use 4, smartphone power-level cores in normal operations but can flash on 64 core performance for a second and lose it just as fast. Thats how the iPhone's A13 works right now. Notebook (unsustained) performance in a smartphone with no fan. It's why the iPad Pro outperforms the 13 inch MacBook Pro at video rendering. It's the advantages of ARM chip design in 2020 and I imagine they will play to every last one of their strengths.

All the real money is in pre-built systems. AMD sells a ton of units but nothing like Qualcomm or even Intel with their notebook stronghold. I'm sure theres space, just that I believe the market will be decreased in size and if Apple is over there frying everyone, then it's gonna be hard to advertise your hard work beating Intel. I'm not saying all of this will happen or even will happen next year, it's just that the forces here are somewhat massive advantages.

I ask because Intel's cache comparison is vs. Zen2. So my assumption is they (Intel) target AMD 4000 with slightly more elements/accelerators to get a better look than Zen2. I'd expect AMD 5000 to be released at the same time as this *lake CPU.

Unfortunately the ryzen 4000 series already compares fairly unfavorably against the ice lake 10th gen mobile cpus on arguably the two most important mobile metrics: per-core performance and overall battery use. It's really unfortunate how AMD just isn't dedicating as many resources to the mobile as it deserves.

That said, by the time Tiger Lake comes out, Zen3 mobile will likely already be close to shipping, hopefully there will be significant improvements there.

AMD Ryzen has a track record of shipping on time, whereas novel process Intel CPUs have not shipped on time for nearly half a decade.

They will certainly compete at some point but I'd expect Ryzen 4000/5000 mobile processors to be out when AMD says they will.

Laptops are already shipping with Ryzen 4000 series APU-s.

Can I get this particular Intel rig without the Intel ME backdoors courtesy of the Talpiot program?

No? OK then, I'll stick with Ryzen.

...Which itself has similar "backdoors"

Not according to security researchers.

This is nonsense. AMD has it's equivalent of the ME and it's under a lot less scrutiny. If it didn't have "backdoors" it would be a miracle.

What researchers?

https://www.google.com/search?q=AMD+backdoor&rlz=1C1GCEA_enA...

[citation needed]

https://www.google.com/search?q=AMD+backdoor&rlz=1C1GCEA_enA...

Too many links over the years to specifically cite just ONE

Then you just have a system with AMD PSP security issues

Sweet title.

Reply on news.ycombinator.com