Thuban to Ryzen IPC uplift?

[Chicken76]

So, almost six months after Ryzen's launch, what is the consensus regarding the IPC difference compared to Thuban? (in general workloads, nothing compiled specifically for the new instructions in Ryzen)

And another question, is there still an IPC boost to be had when disabling SMT on Ryzen or was that ironed out in microcode updates?

 

[moinmoin]

And another question, is there still an IPC boost to be had when disabling SMT on Ryzen or was that ironed out in microcode updates?

That was/should be ironed out in AGESA updates indeed.

 

[Ancalagon44]

From the benchmarks I have seen, it looks like Kaveri finally equaled Thuban in IPC.

Therefore, when AMD says that Ryzen has 52% higher IPC than Kaveri, it also has 52% higher IPC than Thuban - at minimum.

Bear in mind that certain workloads in Bulldozer CPUs can use newer instructions that are a lot faster. I know cryptographic functions are typically affected. For those, the IPC gain could be well over 52%.

 

[WhoBeDaPlaya]

Thuban ~ Core 2. Since Ryzen ~ Haswell (thereabouts), a good estimate would be :

1 (Core 2)
1.1 (Nehalem)
1.1 (Sandy)
1.05 (Ivy)
1.1 (Haswell)

So ~40%+ IPC increase.

 

[Chicken76]

From the benchmarks I have seen, it looks like Kaveri finally equaled Thuban in IPC.

Therefore, when AMD says that Ryzen has 52% higher IPC than Kaveri, it also has 52% higher IPC than Thuban - at minimum.

Bear in mind that certain workloads in Bulldozer CPUs can use newer instructions that are a lot faster. I know cryptographic functions are typically affected. For those, the IPC gain could be well over 52%.

Didn't AMD say the 52% was from the last construction core architecture, and Kaveri was the one before that.
Anyway, I should see at least a 40% uplift from my Thuban, right? Guess it's time to upgrade from AM3+. If only RAM prices were a bit more down to earth... I might as well wait until the end of the year for some better offers, since this system has served me so well until now.

Thuban ~ Core 2. Since Ryzen ~ Haswell (thereabouts), a good estimate would be :

1 (Core 2)
1.1 (Nehalem)
1.1 (Sandy)
1.05 (Ivy)
1.1 (Haswell)

So ~40%+ IPC increase.

Wasn't Sandy a 1.2 - 1.25 step up?

 

[formulav8]

Since Ryzen is Broadwell-ish IPC (depending on workload), you would have a very solid 50+% IPC uplift.

 

[Ancalagon44]

Didn't AMD say the 52% was from the last construction core architecture, and Kaveri was the one before that.
Anyway, I should see at least a 40% uplift from my Thuban, right? Guess it's time to upgrade from AM3+. If only RAM prices were a bit more down to earth... I might as well wait until the end of the year for some better offers, since this system has served me so well until now.

I thought Kaveri was the last one, but yeah, I meant the last one, whichever it is. They had a 52% increase, and from benchmarks that I found, Thuban is approximately equal to the last construction core in IPC.

Took them something like 4 years to get IPC back to where it was!

I'm in the same boat - I also have a Thuban. Phenom II X6 1055T to be exact. But, I'm starting to wonder if I shouldn't just wait for Zen 2 or Zen 14nm+ to be released. Reason being, I won't have the money to upgrade this year. I need to spend some money on renovations to my flat first. And, when those are done, I am thinking of buying a Vega GPU so that I can mine coins in order to pay for the Ryzen upgrade. So, by the time that is done, it might be time for Zen 2 or Zen+ or whatever.

 

[Chicken76]

Since Ryzen ~ Haswell (thereabouts)...

Since Ryzen is Broadwell-ish IPC (depending on workload)

Are you guys referring to the same thing (integer, floating point, etc.)? From what I've read online, opinions are divided, some say it's Haswell level of IPC, others that it's Broadwell level (especially in floating point), while others say it's between Ivy Bridge and Haswell but closer to Ivy. Is there no consensus yet?
By the way, which is Ryzen's strongpoint, integer or floating point?

Took them something like 4 years to get IPC back to where it was!

I'm in the same boat - I also have a Thuban. Phenom II X6 1055T to be exact. But, I'm starting to wonder if I shouldn't just wait for Zen 2 or Zen 14nm+ to be released. Reason being, I won't have the money to upgrade this year. I need to spend some money on renovations to my flat first. And, when those are done, I am thinking of buying a Vega GPU so that I can mine coins in order to pay for the Ryzen upgrade. So, by the time that is done, it might be time for Zen 2 or Zen+ or whatever.

Indeed, it's been disappointing to watch AMD take a step backwards on so many levels with the construction cores.

When the next Zen will be released does not seem very important to me. I think they've (if not straight out promised) hinted that the next generation will work on existing AM4 motherboards, so any investment in DDR4, cooler and Mobo won't be lost when next gen arrives, and the CPU itself will probably hold a good portion of it's value on the second-hand market. If only RAM were not so expensive! I really don't want to buy 4GB sticks again.

 

[WhoBeDaPlaya]

Are you guys referring to the same thing (integer, floating point, etc.)? From what I've read online, opinions are divided, some say it's Haswell level of IPC, others that it's Broadwell level (especially in floating point), while others say it's between Ivy Bridge and Haswell but closer to Ivy. Is there no consensus yet?
By the way, which is Ryzen's strongpoint, integer or floating point?

Of course it varies from workload to workload, but on the average, Ryzen's IPC is in the Haswell / Broadwell neighborhood

Another point that is sometimes overlooked is that AMD also finally caught up to Intel platform-wise (eg. SATA performance, etc.)

 

[Ancalagon44]

Indeed, it's been disappointing to watch AMD take a step backwards on so many levels with the construction cores.

The construction core era was very disappointing but thankfully it is over now.

 

[formulav8]

Are you guys referring to the same thing (integer, floating point, etc.)? From what I've read online, opinions are divided, some say it's Haswell level of IPC, others that it's Broadwell level (especially in floating point)

That's why it's said depends on workload.

 

[rainy]

Therefore, when AMD says that Ryzen has 52% higher IPC than Kaveri, it also has 52% higher IPC than Thuban - at minimum.

You mean Carrizo/Bristol Ridge because they're using Excavator cores when Kaveri is based on Steamroller.
https://en.wikipedia.org/wiki/List_of_AMD_accelerated_processing_unit_microprocessors#Desktop_APUs

 

[itsmydamnation]

Are you guys referring to the same thing (integer, floating point, etc.)? From what I've read online, opinions are divided, some say it's Haswell level of IPC, others that it's Broadwell level (especially in floating point), while others say it's between Ivy Bridge and Haswell but closer to Ivy. Is there no consensus yet?
By the way, which is Ryzen's strongpoint, integer or floating point?

There is only one muppet who claims IB level IPC. You have to be careful, look what the workload is bound by, you might compare a IB with 1866 DDR3 memory to a Zen review running 2133 DDR4 and go "look at the IVB IPC, ROFL fail". While ignoring that middle of the road 1866 (9-9-9-27) has a access latency of 9.65ns while the 2133 DDR4 has an access latency of 13.2ns.

Now give the Zen SOC 3200mhz (15-15-15-X) which is 9.33ns and compare again, getting a double uplift because it brings up the entire Infinity fabric clock rate.

i did this test on ivb using dragon age:

speed    timings    Memory latency NS    AIDA latency    FPS
1333    7-7-7-24-T1       10.5    62.2    122
1333    8-8-8-27-T1       12    66.1    116
1333    9-9-9-30-T1       13.5    72.8    74
1333    10-10-10-33-T1    15    75.4    70
1600    8-8-8-27-T1       10    60.5    122
1600    9-9-9-30-T1       11.25    64.4    106
1600    10-10-10-27-T1    12.5    61.5    107
1600    12-12-12-39-T1    15    76.1    75
1866    9-10-9-33-T2*     9.6    60.2    108
1866    10-11-10-36-T2    10.7    58.1    105
1866    12-13-12-39-T2    12.8    66.9    97
1866    15-16-15-40-T2    16    FAIL TO POST  
1866    11-12-11-36-T1    11.7    62.9    111

Here it is again sorted by latency

speed    timings    Memory latency NS    AIDA latency    FPS
1866    9-10-9-33-T2*    9.6    60.2    108
1600    8-8-8-27-T1       10    60.5    122
1333    7-7-7-24-T1       10.5    62.2    122
1866    10-11-10-36-T2    10.7    58.1    105
1600    9-9-9-30-T1       11.25    64.4    106
1866    11-12-11-36-T1    11.7    62.9    111
1333    8-8-8-27-T1       12    66.1    116
1600    10-10-10-27-T1    12.5    61.5    107
1866    12-13-12-39-T2    12.8    66.9    97
1333    9-9-9-30-T1       13.5    72.8    74
1333    10-10-10-33-T1    15    75.4    70
1600    12-12-12-39-T1    15    76.1    75
1866    15-16-15-40-T2    16    FAIL TO POST

*9-9-9-30 unstable 9-10-9-33 maybe not working well?

The latency trend is clear.
Anything that has serial sections of code where maximum ILP cant be extracted ( even if it can run lots of threads) gets good gains from reducing memory latency.

 

[Markfw]

All I can add is that Ryzen LOVES fast low CL memory. I am running 3200 CL14, and it screams compared to benchmarks using slower memory. My Threadripper runs 3600 CL16

 

[DrMrLordX]

Thuban -> Ryzen? Hmm!

Are we comparing on a per-core or per-thread basis? Thuban doesn't have SMT or CMT, so per-core it does less in workloads that scale up to a large number of threads. For example:

https://hwbot.org/submission/2432626_bones_cinebench___r15_phenom_ii_x6_1100t_be_640_cb/

That 6c/6t chip scored a paltry 640 @ 4118 MHz. Seriously? Normalized to a speed of 4 GHz, that chip would get ~621.66. Scale it up to a hypothetical 8c/8t "x8" chip and it gets ~828.88.

My 8c/16t R7 1800x @ 4 GHz can do 1803 in the same benchmark (check the CB R15 thread if you don't believe me). So I see a %117 increase in IPC moving from Thuban to Summit Ridge.

And that bench doesn't use anything newer than SSE4.

edit: oooh ooh y-cruncher time!

http://www.numberworld.org/y-cruncher/benchmarks/charts/500m.html

R7 1700X @ 3.8 GHz: 41.612s

x6 1055T @ 4.06 GHz: 178.526s

Hypothetical "x8" @ 3.8 GHz: ~143.056s

In this bench Summit Ridge has ~244% higher IPC. Oh my.

 

[rvborgh]

i can compare K10 i guess since i run the equivalent of 8 Thubans/Istanbuls on my quad Opty. 48 K10 cores at 3.2 GHZ = 16 Zen cores at 4.1 on Cinebench R15 (both score around 3200cb).

 

[SPBHM]

well, you have Ryzen vs PII both with 4c/4t here
http://www.hardware.fr/articles/965-2/performances-applicatives.html
PII 980 (3.7GHz) and R3 1300X (3.5-3.85GHz), Ryzen is 64% faster on their applications test and 67% for gaming
given the load I would think Ryzen is not running at max turbo most of the time, is Ryzen at full load typically running at base clock or how high? I'm guessing the average clock since most there is heavy MT is going to look more like 3.6-3.7Ghz? so pretty close, I guess around "70%" would be more or less correct, and when you add SMT is a lot better.

 

[Chicken76]

Thanks guys for all the data, I appreciate it.

Now give the Zen SOC 3200mhz (15-15-15-X) which is 9.33ns and compare again, getting a double uplift because it brings up the entire Infinity fabric clock rate.

The Infinity Fabric clock being tied to the memory clock is an interesting aspect. Are there any measurements for the effect on power draw when raising memory/IF clock and keeping core clock and voltage fixed (@stock preferably)? Does the memory controller voltage go up when running higher frequency RAM?
I also wonder if running high memory/IF clocks for a prolonged period is going to damage the chip and in time force the clocks lower in order to maintain stability. Officially they only support up to 2666MHz memory, right?