Ryzen 2 IPC improvements measured

[Humbug]

These tests were done by comparing a 1st generation Ryzen 7 1700 to a Ryzen 7 2700x. A GTX 1080 was used and DDR4 @ 3200Mhz. Both CPUs were clocked at a base of 4Ghz for these tests. Tests were done on the old x370 chipset. Since both are 8 core, 16 thread parts this testing methodology allows us to attempt to isolate the instructions per clock improvements in various workloads.

 

The only problem with this testing methodology is that on the Ryzen+ chip a single core occasionally will boost to 4.3Ghz. So keep that in mind when viewing the numbers.

 

 

 

 

 

 

[bcredeur97]

Quote

The only problem with this testing methodology is that on the Ryzen+ chip a single core occasionally will boost to 4.3Ghz

so the test is kinda useless?

 

I mean that's 7%...

[FratStar]

4 minutes ago, Humbug said:

The only problem with this testing methodology is that on the Ryzen+ chip a single core occasionally will boost to 4.3Ghz. So keep that in mind when viewing the numbers.

I'm pretty sure you can turn off XFR and precision boost.

[Princess Luna]

1 minute ago, bcredeur97 said:

so the test is kinda useless?

Not at all, it still gives us a fair idea of raw gains one can expect... Certainly not enough to justify someone with a Ryzen 7 or an i5/i7 14nm Intel CPU to upgrade just yet but this refresh so far has been proving to be more meaningful than the Radeon RX 500 series was which is a gain to us all.

[Taf the Ghost]

I see someone else also saw that link floating around.

 

Fun part about these benchmarks is it's on a X370 board with the same top-line memory timings. The Overclockers are going to have some fun with the memory sub-timings on this. 

[Jurrunio]

That's something, but I wonder if that's done by hurting efficiency like coffee lake.

[Taf the Ghost]

1 minute ago, Princess Cadence said:

Not at all, it still gives us a fair idea of raw gains one can expect... Certainly not enough to justify someone with a Ryzen 7 or an i5/i7 14nm Intel CPU to upgrade just yet but this refresh so far has been proving to be more meaningful than the Radeon RX 500 series was which is a gain to us all.

Depends on the Ryzen 7. 1800X buyers probably not, but a stock 1700 is 3.0/3.7. (All-core is 3.4, I think?) For even certain buyers of the Ryzen 7s, the 2700X could be a >20% improvement as a "drop in" upgrade. For those that put a lot of work into OC'ing their CPUs, no, but most people don't do that.

[Taf the Ghost]

2 minutes ago, Jurrunio said:

That's something, but I wonder if that's done by hurting efficiency like coffee lake.

Coffee Lake's 14nm++ is actually a pretty big jump in efficiency over Kaby Lake's 14nm+ process. That's why the Thin/Light & Laptop spaces are seeing a big jump in raw performance with the 8th Gen Intel parts. I'm still not sure how Intel's engineers found that much of a lower-clock efficiency improvement within a node improvement, but it's actually quite impressive in that space. In the Desktop, the addition of 50% more cores was always going to be a little limiting, but the top SKU parts clock well at good efficiency. 

 

Now, Skylake-X is a different issue.

[The Benjamins]

15 minutes ago, Humbug said:

These tests were done by comparing a 1st generation Ryzen 7 1700 to a Ryzen 7 2700x. A GTX 1080 was used and DDR4 @ 3200Mhz. Both CPUs were clocked at a base of 4Ghz for these tests. Tests were done on the old x370 chipset. Since both are 8 core, 16 thread parts this testing methodology allows us to attempt to isolate the instructions per clock improvements in various workloads.

 

The only problem with this testing methodology is that on the Ryzen+ chip a single core occasionally will boost to 4.3Ghz. So keep that in mind when viewing the numbers.

 

How can it be clocked to 4Ghz, and still have Turbo boost active?

[Humbug]

3 minutes ago, The Benjamins said:

How can it be clocked to 4Ghz, and still have Turbo boost active?

Check the youtube video at the 3:14 mark where he talks about the boost and shows a graph. Unless he was just showing that as a indicator of stock behaviour unrelated to this test...

[Princess Luna]

@Taf the GhostIt makes better sense just overclock the CPU than buy a new one for the same frequency boost though... I don't think any one with any Ryzen 7 1700 will be really tempted to burn money on a Ryzen 7 2700x for a boost in performance that lets be honest we could all live without it...

 

That is the beauty on current line ups though, remember when actual useful 8 cores processors for good pricing were a myth? Hard to complain nowadays.

 

Also yes Coffee Lake is more efficient noticeably, the locked i7 7700 for instance would thermal throttle quite easily on its stock cooler... the i7 8700 now even carrying 2 extra cores thanks to its efficiency and improved TIM demands more "extreme" condition to thermal throttle with the stock cooler.

[The Benjamins]

5 minutes ago, Humbug said:

Check the youtube video at the 3:14 mark where he talks about the boost and shows a graph. Unless he was just showing that as a indicator of stock behaviour unrelated to this test...

That is unrelated to those graphs as far as I can tell. he mentions that the CPU will preform better then the graphs due to the 300Mhz boost (4.3) and showing a PC mark run where it boosts when its needed.

 

EDIT: it would be a poor review if he stats they are both at 4Ghz but lets the 2700x boost, that defeats the point of the both at 4Ghz tests.

[Humbug]

1 minute ago, The Benjamins said:

That is unrelated to those graphs as far as I can tell. he mentions that the CPU will preform better then the graphs due to the 300Mhz boost (4.3) and showing a PC mark run where it boosts when its needed.

oh ok, thanks

[Taf the Ghost]

5 minutes ago, Princess Cadence said:

@Taf the GhostIt makes better sense just overclock the CPU than buy a new one for the same frequency boost though... I don't think any one with any Ryzen 7 1700 will be really tempted to burn money on a Ryzen 7 2700x for a boost in performance that lets be honest we could all live without it...

 

That is the beauty on current line ups though, remember when actual useful 8 cores processors for good pricing were a myth? Hard to complain nowadays.

 

Also yes Coffee Lake is more efficient noticeably, the locked i7 7700 for instance would thermal throttle quite easily on its stock cooler... the i7 8700 now even carrying 2 extra cores thanks to its efficiency and improved TIM demands more "extreme" condition to thermal throttle with the stock cooler.

Well, "makes better sense" depends on context. If you have a weak IMC on your 1700, there's a case to be made for replacing the CPU with the new generation (and selling the old one).

[Sakkura]

I'm sure someone out there is annoyed that an improvement in cache and memory latency gets described as an IPC improvement. But whatever, the effect is the same.

 

If those numbers are legitimately the improvement at identical clocks, that's quite impressive. Along with a decent clock boost, Ryzen+ could be pretty awesome.

 

That old Ivy Bridge chip of mine is like...

 

Spoiler

 

[Humbug]

8 minutes ago, Sakkura said:

I'm sure someone out there is annoyed that an improvement in cache and memory latency gets described as an IPC improvement.

why?

Isn't it accepted that cache and memory system improvements directly contribute to IPC?

[DoctorWho1975]

3 hours ago, Taf the Ghost said:

Depends on the Ryzen 7. 1800X buyers probably not, but a stock 1700 is 3.0/3.7. (All-core is 3.4, I think?) For even certain buyers of the Ryzen 7s, the 2700X could be a >20% improvement as a "drop in" upgrade. For those that put a lot of work into OC'ing their CPUs, no, but most people don't do that.

 

My 1700 is at 3.9 across all cores.. I'll wait for the 3000 series but this refresh looks to be pretty damned compelling, again.

[Sakkura]

2 minutes ago, Humbug said:

why?

Because IPC can also be used in a narrower sense, where it defines exactly how many of a given instruction can be processed per clock cycle. That's purely a function of the design of the CPU core, independent of cache and memory controller.

 

The distinction just doesn't really matter in practice.

[ScratchCat]

5 minutes ago, Humbug said:

why?

Isn't it accepted that cache and memory system improvements directly contribute to IPC?

Cache and memory improvements don't increase the number of instructions per clock but do decrease the time the core is waiting for data. This results in a gain in performance.

[The Benjamins]

39 minutes ago, Humbug said:

why?

Isn't it accepted that cache and memory system improvements directly contribute to IPC?

The IPC gain is said to be around 3% (from what I heard), the reason these games and task improve is because they where bottle necked by the slower cache latecny of ryzen.

[Taf the Ghost]

11 minutes ago, The Benjamins said:

The IPC gain is said to be around 3% (from what I heard), the reason these games and task improve is because they where bottle necked by the slower cache latecny of ryzen.

From AMD's down slide that's currently floating around, it says 3%. (IPC is always dependent on the type of instruction, obviously, but as a general statement.) 

 

Gaming involves the GPU and the interaction with both the Memory system & CPU, which means there is always a lot more room for improvement.

[JoeCoke]

Really nice gains, but I'm still gonna wait for Ryzen 2. I've dealt with my 8320 this long, I'll survive another year or less. But yeah, 10% IPC improvement is pretty nice and if it can hit decent clock speeds (still kinda hoping for 4.5ghz even though it's looking REALLY unlikely) that'd be amazing. Gets me excited for what the 3000 series could offer. (Maybe another 10% IPC improvement @ 5ghz? That'd be the shit...)

[Mira Yurizaki]

8 minutes ago, 2Buck said:

Really nice gains, but I'm still gonna wait for Ryzen 2. I've dealt with my 8320 this long, I'll survive another year or less. But yeah, 10% IPC improvement is pretty nice and if it can hit decent clock speeds (still kinda hoping for 4.5ghz even though it's looking REALLY unlikely) that'd be amazing. Gets me excited for what the 3000 series could offer. (Maybe another 10% IPC improvement @ 5ghz? That'd be the shit...)

How is 10% "pretty nice"?

[JoeCoke]

4 minutes ago, M.Yurizaki said:

How is 10% "pretty nice"?

How is it not "pretty nice"?

[Curufinwe_wins]

This is much higher than I actually expect to be true. Much higher than anything AMD has claimed, and much higher than the 4.1 GHz leaked benchmarks values seem to imply.

 

Now there are lots of ways to make things seem apparently the same and them not actually be, but I guess we will wait to find out.