5Ghz Amd & Intel (speculation topic)

geordiejc · 31 Oct 2017 at 22:03

If/when we see a 5Ghz part from AMD e.g. 6 core / 12 threads would performance be within the margin of error when comparing the two if using same ram etc... (apples vs apples)

Acceptance that cache size will differ...

Doesn't the future look promising for us all if that's the case !? (For those who can wait to upgrade that Is!)

jigger · 31 Oct 2017 at 22:24

Well it would depend to some extent, but a 5Ghz Ryzen would murder the current Intel designs.

geordiejc · 31 Oct 2017 at 22:34

jigger said:
Well it would depend to some extent, but a 5Ghz Ryzen would murder the current Intel designs.

Would you care to elaborate on your hypothesis please Mr jigger? In what situations ? I'm genuinely curious in yours and other people's opinions

sandys · 31 Oct 2017 at 22:37

high clockspeeds won't feature in any future designs, it's not the way forward, it's all about doing more in a clock cycle and on demand parallelism for efficiency.

You see it in most markets, mobile, server, graphics, desktop CPUs will be no different

geordiejc · 31 Oct 2017 at 22:40

I'm off to Google this demand parrelism jobbie brb...

So it turns out I'm none the wiser...

Is this IPC?

Terrorfirmer · 31 Oct 2017 at 23:02

jigger said:
Well it would depend to some extent, but a 5Ghz Ryzen would murder the current Intel designs.

What is the thinking here?

I believe the OP was asking that if an Intel part, and Ryzen part, were both at the same clockspeed and using the same ram, would they be comparable?

And you said at 5Ghz a Ryzen would somehow murder a 5Ghz Intel part? I'm not sure I'm following the logic there.

Eg i5-8600K v Ryzen 1600X both at 5Ghz (notwithstanding that just would not happen). The Intel would be better because better IPC. Not that's not quite a fair comparison because the 1600X is cheaper but they're both six core parts.

But not much different to comparing say a boosted i5-8400 vs a boosted 1600 on all cores, say they were 3.8Ghz each, the i5-8400 will still be faster.

BTVA · 31 Oct 2017 at 23:24

5Ghz is a bit of a reach. But let's say if we could pick a current ryzen CPU, a current Intel cpu and then make them both run at 3.5Ghz and 4.0Ghz after that just to see how they scale with clock, we can then make some prediction when at 4.5Ghz and further. That is my proposed method.

Aretak · 31 Oct 2017 at 23:24

Terrorfirmer said:
The Intel would be better because better IPC.

There's really not much of a gap on that front. Sure, Ryzen is directly comparable to Broadwell rather than Skylake (and its many rehashes), but the difference is under 5%. Of course, there's a gaping chasm in terms of the two architectures, which leads to some programs preferring (and presumably optimised for) Intel's ring bus setup. Skylake-X has the exact same problem when it comes to gaming.

sandys · 1 Nov 2017 at 01:01

What is interesting there is that the Ryzen core is the same for all chips, it spreads 11 points 136 - 147 and all that is changing is the cache.

gavinh87 · 1 Nov 2017 at 01:40

We've already been through this in other threads, even at the same clock speeds coffeelake is about 10% faster than ryzen. There is no murdering going on here.

Ryzen if currently 10% behind intel in IPC and is behind on clock speeds

humbug · 1 Nov 2017 at 02:51

You're exaggerating, its not 10% ^^^^

sandys said:
What is interesting there is that the Ryzen core is the same for all chips, it spreads 11 points 136 - 147 and all that is changing is the cache.

Actually the Ryzen 1300 and 1200 are not the same, they have less L3.

Aretak said:
There's really not much of a gap on that front. Sure, Ryzen is directly comparable to Broadwell rather than Skylake (and its many rehashes), but the difference is under 5%. Of course, there's a gaping chasm in terms of the two architectures, which leads to some programs preferring (and presumably optimised for) Intel's ring bus setup. Skylake-X has the exact same problem when it comes to gaming.

Love that the FX-8350 is in that.

Ryzen 5 and 7 are 70% faster than the FX-8350..... Womp Womp...

Oh and yes, it would be great. soon...

gavinh87 · 1 Nov 2017 at 03:48

humbug said:
You're exaggerating, its not 10% ^^^^

Actually the Ryzen 1300 and 1200 are not the same, they have less L3.

Love that the FX-8350 is in that.

Ryzen 5 and 7 are 70% faster than the FX-8350..... Womp Womp...

Oh and yes, it would be great. soon...

You need more than 1 test to work out ipc.
Soon tm. It's all we heard from the fx series ....

jaffveyda · 1 Nov 2017 at 08:23

gavinh87 said:
We've already been through this in other threads, even at the same clock speeds coffeelake is about 10% faster than ryzen. There is no murdering going on here.

Ryzen if currently 10% behind intel in IPC and is behind on clock speeds

10% for some of us is like night and day.

David Bisset · 1 Nov 2017 at 08:45

That graph shows only cinebench - IPC is not an objective easily comparable measure like some keep treating it (and it varies significantly based on other factors such as ram speed etc with different scaling for different architectures). It varies heavily by workload and in some situations Ryzen trails Coffeelake by more than the quoted 10% while in others it can be ahead.

jigger · 1 Nov 2017 at 10:33

geordiejc said:
Would you care to elaborate on your hypothesis please Mr jigger? In what situations ? I'm genuinely curious in yours and other people's opinions

It's probably a little too complicated to get into here, but say AMD was building on a process that was capable of a 5Ghz clock speed then Ryzen would be tuned around x amount of cores with x amount of performance each. It's a safe bet that Ryzen in its current form would benefit from a hike in clock speed.

sandys · 1 Nov 2017 at 11:13

humbug said:
Actually the Ryzen 1300 and 1200 are not the same, they have less L3.

I did say they are all the same bar cache

the chart shows the reliance on cache for keeping the chip fed as they all have the same core processor across the range bar that difference.

It is something you see when you start tuning your system, the cache, memory bandwidth and latency help this chip perform.

jigger · 1 Nov 2017 at 11:27

sandys said:
I did say they are all the same bar cache the chart shows the reliance on cache for keeping the chip fed as they all have the same core processor across the range bar that difference.

It is something you see when you start tuning your system, the cache, memory bandwidth and latency help this chip perform.

It's not the same when it comes to measuring performance across multiple cores. The performance gets even more difficult to measure when you consider SMT.

gavinh87 · 1 Nov 2017 at 11:55

jigger said:
It's probably a little too complicated to get into here, but say AMD was building on a process that was capable of a 5Ghz clock speed then Ryzen would be tuned around x amount of cores with x amount of performance each. It's a safe bet that Ryzen in its current form would benefit from a hike in clock speed.

Of course it's going to be faster than it's current form, it has a higher clockspeed.
Both cpus at the same clockspeed and intel would still win due to the ipc. The stilt has done Many tests and had shown that kaby is 10% ahead on average using a multitude of tests. Add the clockspeed advantage and your looking at 15-20% faster in some scenarios.
For some that's a gpu upgrade.
How you come to the conclusion that ryzen murders intel at the same clockspeed has me intrigued. Feel free to pm explaining what you can not go into here, you're almost making me regret getting rid of the 1700!

jigger · 1 Nov 2017 at 12:13

You can't take a single cores IPC and apply that to a CPU's multi core performance. AMD have a better implmentation of hyperthreding and better system for inter comunication. Both of which will improve with clock speed. Intel's current designs have essentially either stopped scaling with clock speed or have switched focus from performance per core all together. AMD have a better design and a lot more room to leverage metrics like improved frequency.

A 5Ghz Ryzen design would be a monster. All those those performance advantages seen with the AMD design would scale.

gavinh87 · 1 Nov 2017 at 12:37

jigger said:
You can't take a single cores IPC and apply that to a CPU's multi core performance. AMD have a better implmentation of hyperthreding and better system for inter comunication. Both of which will improve with clock speed. Intel's current designs have essentially either stopped scaling with clock speed or have switched focus from performance per core all together. AMD have a better design and a lot more room to leverage metrics like improved frequency.

A 5Ghz Ryzen design would be a monster. All those those performance advantages seen with the AMD design would scale.

I agree with some if what you said however having a better interconnect is wrong. It's better for cost but it adversely affects performance. There are core to core latency tests that show this. It is even worse with epyc.
Then there is the memory controller, I'm not sure how easy it is to improve this. A 5ghz ryzen design would be nice but I don't think it would easily best an 8 core 5ghz ring bus arc either.