Try as you might but - new Intel 8-core still behind AMD

[Aereldor]

Intel needs to drop prices hard

[Daniel Z.]

Intel will be fine. If it's similar price and only slightly worse, their better brand recognition still puts them AMD for most prebuilts and beginner PC builders. Also, the gaming performance for any reasonably good, 6 or more core CPU since Ryzen 1000 released will perform almost the same in 4k and some 1440p games.

[S w a t s o n]

Due to the current stock situation, even some zen 2 chips are selling out in some places. Intel may not lower prices much because "it just sells"

[Bombastinator]

1 hour ago, Daniel Z. said:

Intel will be fine. If it's similar price and only slightly worse, their better brand recognition still puts them AMD for most prebuilts and beginner PC builders. Also, the gaming performance for any reasonably good, 6 or more core CPU since Ryzen 1000 released will perform almost the same in 4k and some 1440p games.

The 6 core thing is currently true.  For 8core to matter there have to be games that run the top off 6/12, and currently there aren’t any.  4/4 has had this happen already. 4/8 is reaching the end but hasn’t gotten there quite yet.  My personal prediction is 4/8 could hit that as early as Q2 2021.  I had previously been predicting Q1 2020 though so who knows how long that will go for.  Consoles are going from 6/6 directly to 8/16 though, so it is still outside possible 6/12 may simply get jumped. 

[spartaman64]

5 hours ago, HairlessMonkeyBoy said:

If Intel can't compete, it's bad news for everyone. Less competition always means less innovation and more stagnation.

 

4 hours ago, Moonzy said:

i'd start as soon as AMD gets greedy

and them raising MSRP and removing stock coolers is a good red flag

 

waiting for intel to be too weak to fight back is a bit too late

still good in the short term. intel absolutely dwarfs amd even if intel drops the ball for 5 years straight from now they probably still will be bigger than amd

[thechinchinsong]

As long as this is priced well and has very good availability, this should actually do pretty decently. I'm aware that the 5000 series will probably remain better in most metrics, but this might be a good option if Intel puts a decent price tag on it and has more on hand stock than AMD parts.

[leadeater]

Yea I wouldn't use these performance figures for gaming systems, HP bios imposes the Intel TDP limit and does not allow changing that and also does not have MCE. So the performance should be strictly as per Intel spec, not actually how gaming motherboard do it.

[Trik'Stari]

Do people actually expect good performance out of OEM machines? Because that makes me sad.

Spoiler

I've worked on Dell, Lenovo, Apple, and HP machines for a while. None of them use thermal paste that is what I would describe as "adequate", in either their application method, or consistency. ALL of them have the consistency, at best, of chewing gum. At least that's been my experience.

[Bombastinator]

5 minutes ago, Trik'Stari said:

Do people actually expect good performance out of OEM machines? Because that makes me sad.

  Reveal hidden contents

I've worked on Dell, Lenovo, Apple, and HP machines for a while. None of them use thermal paste that is what I would describe as "adequate", in either their application method, or consistency. ALL of them have the consistency, at best, of chewing gum. At least that's been my experience.

Goes to definition of good.

[Trik'Stari]

Just now, Bombastinator said:

Goes to definition of good.

"Not thermal throttling".

 

I've yet to see a factory Dell machine that wasn't running hot, IMHO. Although I've been off the repair bench almost 2 years now, maybe things have changed.

 

Maybe I just expect too much out of a laptop form factor? We didn't really do desktops when I was on the bench.

[Bombastinator]

1 hour ago, Trik'Stari said:

"Not thermal throttling".

 

I've yet to see a factory Dell machine that wasn't running hot, IMHO. Although I've been off the repair bench almost 2 years now, maybe things have changed.

 

Maybe I just expect too much out of a laptop form factor? We didn't really do desktops when I was on the bench.

I haven’t seen a laptop in general in years that didn’t run hot.  They want small so bad (still don’t get that one) and the result is hot machines

[Brooksie359]

7 hours ago, Techstorm970 said:

Holy shit!  Didn't know it was that big a disparity lol!

 

But it makes sense at the same time.  Intel is selling things mostly based on reputation, not better real-world performance, pricing, thermals or efficiency.  Now, they are only preferred for specific use cases (MATLAB, Adobe Premier, maybe a few games) and Ryzen 5000 might take some of those away, too.

There are very few programs that are faster on Intel than they are on ryzen of the same price tier and even then it isn't a significant margin while the same cannot be said about the other way around. If you are an informed buyer then it doesn't make a lot of sense to buy intel over AMD currently. The only reason that I could think of why you would buy intel is that they are actually in stock so if you need something now then it is really your only option unless you want to go ryzen 3000 series. 

[porina]

10 hours ago, Aereldor said:

Intel needs to drop prices hard

Because?...

 

If it is some imaginary perf/price ratio, that's a very bad reason.

 

9 hours ago, S w a t s o n said:

Due to the current stock situation, even some zen 2 chips are selling out in some places. Intel may not lower prices much because "it just sells"

This is likely a factor. How many Zen 3 CPUs are out there? We don't know if OEMs got much of a shipment, but if you want to sell a system, you need to have a CPU in hand.

 

8 hours ago, leadeater said:

Yea I wouldn't use these performance figures for gaming systems, HP bios imposes the Intel TDP limit and does not allow changing that and also does not have MCE. So the performance should be strictly as per Intel spec, not actually how gaming motherboard do it.

While I agree this is not representative of potentially higher performance from a gaming/enthusiast type system, I think it worth recapping that running at or above TDP are both within Intel spec. As long as you do nothing other than adjust the power limits, this is NOT considered an overclock. You do not have to apply the TDP power limits. This is a choice Intel gives to system builders, so they can scale the peak system performance according to the cooling available, where a cooler meeting the TDP is the minimum required to give the specified performance. MCE is an overclock. Ref: https://www.anandtech.com/show/14582/talking-tdp-turbo-and-overclocking-an-interview-with-intel-fellow-guy-therien

 

Note AMD's policy is different. At stock, they do apply a power limit (PPT) which is a value higher than TDP. Changing that value is considered an overclock.

[LAwLz]

What the hell is up with all these comments? Did you guys only read the title and not look at the benchmarks or what?

 

15 hours ago, TempestCatto said:

But hopefully they get their shit together soon - we don't want AMD to turn into the greedy monster Intel once was.

This is them getting their shit together, somewhat. There is literally a 1-6% performance difference according to these benchmarks. 

Right now, Intel is behind. With this they seem to be even or maybe even get ahead depending on how performance will change outside of a prebuilt HP computer.

 

15 hours ago, GDRRiley said:

did people really think they could match the 5800x?

It is matching the 5800X. Did you not look at the benchmarks? It is 1% behind in single core score, and 6% behind in multi-core score. I would say that's matching it. Those scores might as well be margin of error.

 

14 hours ago, Techstorm970 said:

Maybe on 10nm, but this is still 14nm.  So no.

Did you just read the title and not look at the results? They are 1-6% behind in this benchmark and that is on 14nm.

 

14 hours ago, GDRRiley said:

even on 10nm its going to be hard.

Again, did you not look at the benchmarks? The performance difference between this Intel chip in a prebuilt HP computer, and a Ryzen 5800X is 1-6%.

That's with Intel being on 14nm.

[RejZoR]

Differences shouldn't be that high unless they were running single channel memory.

[porina]

1 hour ago, LAwLz said:

What the hell is up with all these comments?

It's not an AMD product, and therefore it must be rubbish.

 

More seriously, I'm not really familiar with geekbench in general, but I just poked around the website a bit. The result posted in OP is generally "a bit" behind other 5800X results, but I also had a look at 3800XT results, which get beaten.

 

So, Rocket lake will be somewhere between Zen 2 and Zen 3 performance overall, putting aside the lack of offerings above 8 cores. I think this is not an unexpected place for the CPU to land, and there is still some scope to move around depending on final clocks. We already know Zen 2 is, on average, ahead of Skylake architecture. Rocket Lake/Cypress Cove, drawing heavily on Ice Lake/Sunny Cove configuration, is a major update from Skylake, and it looks like clocks aren't suffering to implement it although we have to see where they end up in practice.

 

When Ice Lake/Sunny Cove was originally released, Intel claimed an 18% IPC uplift relative to Skylake based on a mixed workload average. AMD's Zen 3 vs Zen 2 was claimed to be 19%. The test conditions may not be the same between them, but slotting in between Zen 2 and Zen 3 in IPC is still a good catch up move.

[RejZoR]

7 minutes ago, porina said:

It's not an AMD product, and therefore it must be rubbish.

 

More seriously, I'm not really familiar with geekbench in general, but I just poked around the website a bit. The result posted in OP is generally "a bit" behind other 5800X results, but I also had a look at 3800XT results, which get beaten.

 

So, Rocket lake will be somewhere between Zen 2 and Zen 3 performance overall, putting aside the lack of offerings above 8 cores. I think this is not an unexpected place for the CPU to land, and there is still some scope to move around depending on final clocks. We already know Zen 2 is, on average, ahead of Skylake architecture. Rocket Lake/Cypress Cove, drawing heavily on Ice Lake/Sunny Cove configuration, is a major update from Skylake, and it looks like clocks aren't suffering to implement it although we have to see where they end up in practice.

 

When Ice Lake/Sunny Cove was originally released, Intel claimed an 18% IPC uplift relative to Skylake based on a mixed workload average. AMD's Zen 3 vs Zen 2 was claimed to be 19%. The test conditions may not be the same between them, but slotting in between Zen 2 and Zen 3 in IPC is still a good catch up move.

Intel chips are really only a good option if you plan on doing great cooling and overlcocking it all core beyond 5GHz. All core is where Ryzen is hard to overclock and that's where you can gain the most. Otherwise you have to be stupid not to pick Ryzen 5000. It's just simply amazing chip.

[Taf the Ghost]

I'm not a fan of Geekbench as something that is all that comparative, especially since Intel really don't need that much for a 11900k to take the "best gaming CPU!" title. The issue for Intel, and why they're all but dead in DIY, is that the 500USD part only matters somewhat. What really matters is AMD doesn't scale down that far in terms of performance as you go down the stack, for single or lightly threaded. So Intel is going to continue to lose out in the 200-400USD range.

 

Frankly, I don't expect Intel to take back that much market share (remember, 10th gen was "better" for 6 months and barely shifted products... beyond it was never in stock anyway) until we get to Alder Lake at the end of 2021 and whatever the new platform is called. (LGA1700?)  But we'll see. 

[porina]

2 minutes ago, RejZoR said:

Intel chips are really only a good option if you plan on doing great cooling and overlcocking it all core beyond 5GHz. All core is where Ryzen is hard to overclock and that's where you can gain the most. Otherwise you have to be stupid not to pick Ryzen 5000. It's just simply amazing chip.

You'd be stupid to pick Ryzen 5000 unless you know why you need a Ryzen 5000. No doubt, if you want the best generally performing x86 CPU at a given core count today, that is where you go. How many people fall into that category? For sure there will be many, but as a proportion of the enthusiast market, it is still small. If you're not after "the best" then everything else, even Intel, remains a viable option using whatever biased value metric you pick. Zen 2 is still a great option. Zen 3 still needs lower models released before it can be considered a wider option.

 

I recently won an i5-10600k in a competition, and have got a new mobo to try it out, recycling parts from other systems so new spending is only the mobo. It performs fine, and runs similar temperature to my 3700X under same cooling. Literally, I moved the Noctua D15 over when I swapped mobos, same case too so same airflow. The power density of Zen 2 is still a pain even if it ran at 88W stock, compared to the observed ~130W of the 10600k under Prime95 (haven't checked lesser loads which will benefit Intel more there than AMD). I see no point in keeping the 3700X system parts now so I should sell it while it still has value.

[TempestCatto]

1 hour ago, LAwLz said:

There is literally a 1-6% performance difference

I'm sorry, but I just don't see this as near enough of an "improvement" to be competitive with AMD right now. Not unless the price speaks for the performance.

[leadeater]

1 hour ago, porina said:

While I agree this is not representative of potentially higher performance from a gaming/enthusiast type system, I think it worth recapping that running at or above TDP are both within Intel spec. As long as you do nothing other than adjust the power limits, this is NOT considered an overclock. You do not have to apply the TDP power limits.

Well Intel has been a bit on and off about that, but strictly to Intel spec was referring to the Intel ARK page and the specs on that and also their technical documentation on their CPU architectures and how the boost and power limits are derived. None of the gaming motherboard adhere to any of this documentation at all, that is because Intel allows configuration of these settings but that is still what is considered outside of Intel spec because you will not be able to match up any power or boost durations while using these boards with any Intel documentation.

 

Running outside of spec doesn't necessarily mean overclock either, it's not like I think of MCE as an overclock and neither the default higher than normal/required vCore gaming motherboards use.

 

But honestly saying CPU running in gamer motherboards is to spec is really not correct, it's just not a violation of warranty.

[porina]

2 minutes ago, leadeater said:

But honestly saying CPU running in gamer motherboards is to spec is really not correct, it's just not a violation of warranty.

We might be arguing over words here. While the "at TDP" spec is part of it, it is not mandatory, so to say it is out of spec when running unlimited power limit in enthusiast mobos is not correct. It is within a different interpretation of spec, since it is still operating within Intel allowed parameters. My understanding of Intel's definition of overclocking is if you directly manipulate the clock ratios or voltage, so MCE would clearly fall in that. Adjusting the power limit by itself does not directly manipulate those, as it is still under CPU control.

[leadeater]

7 minutes ago, porina said:

it is not mandatory, so to say it is out of spec when running unlimited power limit in enthusiast mobos is not correct.

Sure it is correct to say, Intel has documented figures on everything from PL1, PL1, Tau etc etc, those are the specs of the CPU and the architecture. Anything other than those is not to spec. Yes Intel allows you to configure these values, to something different to the defined values in the documentation, hence no longer to spec.

 

If PL2 is not 1.25 times the PL1 (TDP) and Tau is not for example 8 seconds then it is not configured to Intel spec, it's configured to how I want it or the OEM wants it. Intel allows this but the CPU is operating outside of Intel's defined spec.

 

I'm not at all talking about overclocking, that has nothing to do with it. The literal point is Intel has defined specification for their architectures, you can look them up and read them. In an HP system you can match the CPU behavior exactly to the Intel documented spec, it is running at Intel spec. You will find this impossible to do the same on an Asus or Gigabyte gaming motherboard, for that you have to go to the vendor documentation if they have it documented at all.

[leadeater]

50 minutes ago, Taf the Ghost said:

The issue for Intel, and why they're all but dead in DIY, is that the 500USD part only matters somewhat. What really matters is AMD doesn't scale down that far in terms of performance as you go down the stack, for single or lightly threaded. So Intel is going to continue to lose out in the 200-400USD range.

That's what I like most about Ryzen so far, they are giving you the choice to buy a CPU of your needs without penalizing you if you only need 4 or or 8 cores to a significant degree. Everything is unlocked, cores will boost as much as they can within power budget etc. None of this messy big differences in base or boost, or 2 core vs 1 core boost, locked vs unlocked blah blah. Simple often does equal good, here it does, at least in my eyes.

 

Just need to kill off those X SKUs or do something more meaningful with them, like I don't know... decently higher TDP maybe? Something more than now anyway.

[porina]

8 minutes ago, leadeater said:

 

I get what you're saying, but I guess my point is that, the spec defines operation at one (or more) points. Unless operating at those points are mandatory, operating at different points is not in itself out of spec.

 

To pick a car analogy, typically you'll be given a fuel economy at a given speed. If you then travel at a different speed, you don't consider that out of spec do you?

 

1, run at defined TDP settings, get expected performance - in spec

2, run at defined TDP settings, don't get expected performance - out of spec

3, run other allowable settings: not out of spec