Move over 7700, i3 8350 almost matches in performance

[mr moose]

5 hours ago, Taja said:

Maybe not holding back developmet, but holding back the PRODUCT. Selling 8 core for 1000$, and than for 600$, and now for less than 400$... They had the product, they just overpriced it TO SHIT, and created a product line that only makes sense because they had no competition. now that they do, they can just change their strategies, they dont need to make heavy investment.

 

and about thew development part, they were focusing on other areas, because they launched "compelling" products every year (if you call 10% compelling), and in the market they were already kings, they didnt make a lot of effort.

 

And to finish, their complacency DID gave AMD space, but if intel does what they seen to be doing now (with coffee lake, not the shitlakes-x mess), they will be just fine.

 

Huh? It appears you have moved the goal posts to a position that is filled with just as many assumptions.

 

In your opinion it is too expensive and only makes sense with no competition, except that it is still expensive and they have competition.  You do realise that most of Intel's sales are more than just price/perf.   Enthusiasts are the dribble of the industry.  OEM's and data centers are the cash cow, they are more concerned with consistency, predictability and reliability. And just like most business, they are willing to pay more for a product that they know will perform within their expectations and within their current ecosystem.

 

Making claims about 10%  being compelling evidence for anything indicates that you still believe they can do better but choose not to.  You have no evidence for this, even though what we have seen over the last few months indicates otherwise.

 

You're still assuming they where complacent.  AMD did a sterling job on the ryzen in their own right.  It had nothing to do with your perception of Intel's complacency.

 

 

 

 

 

 

[ARikozuM]

10 hours ago, Sauron said:

Again, why would you buy a 400$ 6 core for that?

You need the fastest speed and the program uses multiple threads. Using more than one thread doesn't mean the program is reliant on multithreaded performance. It could very well be what @The Benjamins said. 

[Sauron]

3 minutes ago, ARikozuM said:

You need the fastest speed and the program uses multiple threads. Using more than one thread doesn't mean the program is reliant on multithreaded performance. It could very well be what @The Benjamins said. 

You either can use many cores, in which case ryzen is better, or you can't, in which case get an i3

[ARikozuM]

18 minutes ago, Sauron said:

You either can use many cores, in which case ryzen is better, or you can't, in which case get an i3

Look at benchmarks for Ryzen vs i7-7700K. The R7-1700, despite having eight physical cores, loses to the i7-7700K's eight logical cores. Why? Because programs that do use multiple threads, i.e. BF1 and CAD programs, are still designed to run sequentially. There's no getting around it even with stops and pauses in the functions themselves as that potentially adds latency to the engine. There is no multi-threaded load that wouldn't benefit from stronger single-thread performance. 

[Sauron]

5 minutes ago, ARikozuM said:

Look at benchmarks for Ryzen vs i7-7700K. The R7-1700, despite having eight physical cores, loses to the i7-7700K's eight logical cores. Why? Because programs that do use multiple threads, i.e. BF1 and CAD programs, are still designed to run sequentially. There's no getting around it even with stops and pauses in the functions themselves as that potentially adds latency to the engine. There is no multi-threaded load that wouldn't benefit from stronger single-thread performance. 

....again, if you need single core performance why would you be looking at a 6 core?? And separate threads can run at the same time by definition, otherwise there woul d be no point

[Drak3]

1 hour ago, Sauron said:

....again, if you need single core performance why would you be looking at a 6 core?? And separate threads can run at the same time by definition, otherwise there woul d be no point

Single core performance is a component of multicore performace. It's SCP times Core Count.

 

And separate threads are independent of each other, but a thread is still a sequential chain of instructions.

 

And some use cases involve concurrently running many single thread or few thread programs.

[Kamina]

It's definitely faster than my i5 now.

[Trik'Stari]

On 8/24/2017 at 6:31 PM, Daring said:

hahahaha try being on Z97/LGA1150.

That was my previous system lol

[Nowak]

Just now, Trik'Stari said:

That was my previous system lol

I'm still on Z97, meanwhile.

[Trik'Stari]

1 minute ago, Daring said:

I'm still on Z97, meanwhile.

That sucks. I mean it's not unusable by any means, but still.

 

It's a shame Intel can't/won't build in support for older CPU's that physically could fit into the socket. IIRC the only difference between Z97 and Z170/Z270 was basically artificially created by Intel to force people to upgrade.

[mr moose]

1 hour ago, Trik'Stari said:

That sucks. I mean it's not unusable by any means, but still.

 

It's a shame Intel can't/won't build in support for older CPU's that physically could fit into the socket. IIRC the only difference between Z97 and Z170/Z270 was basically artificially created by Intel to force people to upgrade.

Unless there is something I am missing, artificially preventing new CPU's from running in older boards would only cost them sales.  If they could make their CPU's work on Older boards then they could potentially sell more CPU's as some people who are financially limited might still upgrade the CPU rather than wait until they can afford a whole new system.   I have a i5 3550, if I could drop in a 7th gen 4 core CPU I would seriously have considered doing that last year rather than waiting till the end of this year to upgrade.   

[Sauron]

10 hours ago, Drak3 said:

Single core performance is a component of multicore performace. It's SCP times Core Count.

 

And separate threads are independent of each other, but a thread is still a sequential chain of instructions.

 

And some use cases involve concurrently running many single thread or few thread programs.

I don't understand where you're going with this... yes, of course, single core performance helps in multithreaded tasks too, but what does that have to do with anything? We have established that ryzen cores are fast enough that the difference in IPC doesn't make up for having 2 less cores (in heavily multithreaded workloads). If the program can take advantage of more than a couple of cores, ryzen is faster. If it can't, you might as well save some a lot of money and go for this 4GHz i3 - or whatever the i5 models will be.

[Drak3]

1 minute ago, Sauron said:

I don't understand where you're going with this... yes, of course, single core performance helps in multithreaded tasks too, but what does that have to do with anything? We have established that ryzen cores are fast enough that the difference in IPC doesn't make up for having 2 less cores (in heavily multithreaded workloads). If the program can take advantage of more than a couple of cores, ryzen is faster. If it can't, you might as well save some a lot of money and go for this 4GHz i3 - or whatever the i5 models will be.

The difference in IPC AND clockspeed, which determines single core performance, puts Ryzen as having, at best overclock, the same rough single core performance as the 4790K at stock.

 

And now we're looking at a 6 core part that is looking like it can hit around 5GHz. That's a theoretical total of 30GHz. On Ryzen, 4GHz x8, theoretical total of 32GHz. That would be assuming that all else is equal, but that isn't true. Intel has a signficantly better IMC, Ryzen is held back by its IMC. The monolithic die has lower intercore latency than the CCX design that has to move between CCX's.

 

The 7700K is doing well enough to keep up to the 1800X in most tasks that don't require concurrent threads. The 8700K will do even better than the 7700K, it should be able to match, or outmatch, multicore performance. The only thing the R7 chips will really be better at is concurrent multithread.

[Sauron]

4 minutes ago, Drak3 said:

The difference in IPC AND clockspeed, which determines single core performance, puts Ryzen as having, at best overclock, the same rough single core performance as the 4790K at stock.

 

And now we're looking at a 6 core part that is looking like it can hit around 5GHz. That's a theoretical total of 30GHz. On Ryzen, 4GHz x8, theoretical total of 32GHz. That would be assuming that all else is equal, but that isn't true. Intel has a signficantly better IMC, Ryzen is held back by its IMC. The monolithic die has lower intercore latency than the CCX design that has to move between CCX's.

 

The 7700K is doing well enough to keep up to the 1800X in most tasks that don't require concurrent threads. The 8700K will do even better than the 7700K, it should be able to match, or outmatch, multicore performance. The only thing the R7 chips will really be better at is concurrent multithread.

We have intel's OWN CLAIMS that the new part will be 51% faster than a 7700k at best. This places it below ryzen (in certain tasks of course). You can't automatically deduce performance by multiplying the clock frequency by the core count, even aside from ipc differences. We have the numbers, we've seen the actual performance of ryzen, and if intel themselves are claiming 51% more you can rest assured it won't be doing any better than that.

 

Again, if you don't need parallel performance neither of these products is for you.

[Drak3]

Just now, Sauron said:

We have intel's OWN CLAIMS that the new part will be 51% faster than a 7700k at best. This places it below ryzen (in certain tasks of course). You can't automatically deduce performance by multiplying the clock frequency by the core count, even aside from ipc differences. We have the numbers, we've seen the actual performance of ryzen, and if intel themselves are claiming 51% more you can rest assured it won't be doing any better than that.

 

Again, if you don't need parallel performance neither of these products is for you.

Intel's own claims of 51% improvement over the 7700K is on multicore. 50% more cores, same boost clocks, better IMC. Makes sense.

 

8700K and 7700K have identical IPC. Ryzen is roughly behind 5%. Close enough that I can talk just clock speed for rough estimations.

 

And again, not all parallelized tasks benefit from just throwing cores at it. Nor are highly parallelized programs the only use case for higher core count CPUs.

[Sauron]

1 minute ago, Drak3 said:

And again, not all parallelized tasks benefit from just throwing cores at it. Nor are highly parallelized programs the only use case for higher core count CPUs.

No, in fact all parallelized tasks do benefit from more cores if they're sufficiently parallelized. I can't think of a task that uses exactly 12 threads and no more. Usually parallelized workloads use a few hundred threads and the operating system distributes them among the available logical cores, and having more does equate to a benefit. As for not being the only use case, what are the others? I'm curious. Is it virtual machines? Those definitely do benefit from having more available logical cores. Is it streaming? Encoding does benefit from more cores, as seen in the various ryzen streaming benchmarks.

8 minutes ago, Drak3 said:

Intel's own claims of 51% improvement over the 7700K is on multicore. 50% more cores, same boost clocks, better IMC. Makes sense.

 

8700K and 7700K have identical IPC. Ryzen is roughly behind 5%. Close enough that I can talk just clock speed for rough estimations.

Rough estimates are pointless when we have the numbers. It turns out that in ideal conditions, a ryzen 1700x is about 57% faster than a 7700k, placing it ahead of the 8700k according to intel's own numbers. Even if they performed the exact same, the pricing makes the ryzen offer more compelling since you can get a 1700x on amazon for 350$ right now, and presumably for less as time goes on.

[Trik'Stari]

7 hours ago, mr moose said:

Unless there is something I am missing, artificially preventing new CPU's from running in older boards would only cost them sales.  If they could make their CPU's work on Older boards then they could potentially sell more CPU's as some people who are financially limited might still upgrade the CPU rather than wait until they can afford a whole new system.   I have a i5 3550, if I could drop in a 7th gen 4 core CPU I would seriously have considered doing that last year rather than waiting till the end of this year to upgrade.   

That would be nice.

[yian88]

Unless i3-8350k is ~100-120$ max this is irrelevant and intel i3/i5/ series are dead.

Until they actually release 4 core i3, 6 core i5, and 6-8 core i7 with HT for the same price range the current i3(2c/4t)~100$,  i5/i7 4c/8t (200-400$) etc, nothing intel does matters.

They are keeping off the real release of desktop cpu's with increased core count as much as they can because they have insanely huge profit margins on i3 dual core at <200$ 4ghz + that's close to i7, they have small cheap silicon for more than double the cost, i would roughly estimate any i3 processor leaves the factory at a manufacture price of 50-70$ the rest is taxes and profit up to 200$.

[mr moose]

8 hours ago, yian88 said:

Unless i3-8350k is ~100-120$ max this is irrelevant and intel i3/i5/ series are dead.

Until they actually release 4 core i3, 6 core i5, and 6-8 core i7 with HT for the same price range the current i3(2c/4t)~100$,  i5/i7 4c/8t (200-400$) etc, nothing intel does matters.

They are keeping off the real release of desktop cpu's with increased core count as much as they can because they have insanely huge profit margins on i3 dual core at <200$ 4ghz + that's close to i7, they have small cheap silicon for more than double the cost, i would roughly estimate any i3 processor leaves the factory at a manufacture price of 50-70$ the rest is taxes and profit up to 200$.

 

You forgot logistics and handling, warehousing, retail markup and in some cases GST/VAT.  And this is before we even consider that Intel, like any company, has to make an RoI on every product or there will be no next product.

[ARikozuM]

2 minutes ago, mr moose said:

 

You forgot logistics and handling, warehousing, retail markup and in some cases GST/VAT.  And this is before we even consider that Intel, like any company, has to make an RoI on every product or there will be no next product.

People never account for R&D, marketing, product loss, RMA's, et cetera. They only think of the bill of materials. 

[mr moose]

7 minutes ago, ARikozuM said:

People never account for R&D, marketing, product loss, RMA's, et cetera. They only think of the bill of materials. 

And they think because AMD are cheaper that Intel's prices must be inflated.