Zen 2 might not offer the IPC increase you expect

[mongeese]

Opinion: Across all sorts of tech forums, and particularly in TechSpot’s comment section, I’ve noticed a trend. Everyone loves to quote a 10-20% IPC (instructions per clock) boost for Zen 2, and, while I’m not ruling that out, I think it’s much less likely than what many people believe. That range has only come about because of a series of misunderstandings, errors, and easily refutable leaks, so I’m going to try to set the record straight.

Rumors kicked off six months ago when an Italian computer company with a 32-core Epyc sample told Bits and Chips that Zen 2 offered a 13% IPC performance boost in scientific tasks. For starters, scientific tasks benchmark wildly differently to games – AMD already beats Intel in scientific and productivity IPC, while falling ~3% behind in games clock for clock.

More importantly, using a 32-core server processor as a metric for an 8-core consumer processor is fraught with error. The server chip uses four 8-core CCX dies, meaning it incurs a latency penalty for communication between them. It deals with this penalty by incorporating a superior memory controller to facilitate the infinity fabric that is vastly different from what happens on consumer chips – just look at the 2990WX bandwidth debacle. Likewise, workloads for the server processor are much more optimized, it uses much more RAM, has a drastically larger cache and more PCIe lanes and many other things that may have impacted whatever workloads it was tested on.

Zen+ -> Zen2: +13% IPC (Average) in scientific tasks. Not bad.
P.S. No gaming data, atm.

— Bits And Chips - Eng (@BitsAndChipsEng) October 16, 2018

Confusion escalated earlier this year when Apisak spotted a 12-core AMD processor denoted as Ryzen in the Userbenchmark database. It provoked an enthralling debate on Reddit, out of which spurned some misinformed calculations that also spread to several tech news sites at the time. Comparing a 2700X’s single-core floating precision performance to that of the 12-core, Reddit found an 18% improvement, Tom’s Hardware found a 13% improvement and others somewhere in between.

Because the 12-core ran at an average clock speed of 3.6 GHz, and because very few 2700X’s have done the benchmark at 3.6 GHz, a common mistake was to use a faster 2700X. For example, one site used a 4.1 GHz 2700X and noted that it got the same score as the 12-core, implying that the IPC increase made up for a 500 MHz clock decrease. This is incorrect as the benchmark score actually stops improving after 3.8 GHz, leading to a false reading. Another site used a very underperforming 2700X.

To find a more accurate measure, I found five 2700X benchmark runs online that fitted the following criteria: an average speed of 3.6 GHz during the test, RAM speeds between 2133 MHz and 3200 MHz (consumer range), and a non-benchmark CPU utilization of less than 10%. The 12-core outperformed these five benchmark runs by an average of 5.2%, a much more realistic number.

Rumors started up again just last week, owing to an article claiming that “Ryzen 3000 CPUs allegedly feature 15% better IPC,” based on information found on Chinese social media. Unfortunately, the original source is unclear if the 15% improvement is regarding IPC or overall performance. While I believe it means the latter, look and decide for yourself.

There might be a lot of sources placing Zen 2’s IPC boost in the 10-20% range, but they’re simply not reliable enough to be used as indicators. The only thing we can really extrapolate on is the fact that Zen 2 processors with eight or fewer cores will not incur the latency penalty associated with the multiple CCX die configuration used in Zen.

I believe that a 10-20% increase is possible, but only as likely as 5-10%.

Further Reading

• 4GHz CPU Battle: AMD 2nd-Gen Ryzen vs. Intel 8th-Gen Core
• Radeon VII & GeForce RTX 2080 using Ryzen 7 2700X & Core i7-8700K

Permalink to story.

https://www.techspot.com/news/79942-zen-2-might-not-offer-ipc-increase-you.html

 

[Evernessince]

It should be noted out that AMD did have an official slide showing they achieved 1.25 times the performance from the 7nm node alone

https://hothardware.com/ContentImag...r-gains.jpg.ashx?maxwidth=1170&maxheight=1170

Performance isn't IPC of course and that figure likely includes clock gains and IPC gains (from the reduced distance the electricity has to travel).

 

[Vulcanproject]

Considering the ways that an 8 core Zen 2 would be sped up (latency, prime issue) I would be surprised to NOT see at least a 10 percent IPC improvement across the board. It's assumed we are now looking at native 8 core parts, 8 core per CCX.

Heck, it was about 3 percent between Zen and Zen+ 8 cores and AMD did nothing much at all. Tiny tweaks, finding the rest in the boost clock speed refinements. Hence the 'plus' and not the full blown '2' moniker, where you expect a proper improved sequel.....

If they can't find 10 percent from latency gains on native parts and on the infinity fabric, wider FPU registers, better cache, chipset enhancements and a few other refinements like a host of additional instructions I would be kind of disappointed! Zen+ has these obvious weaknesses that when focused upon you would expect good gains in those areas.

It's not easy to pull IPC out of the hat by any means, however AMD have gotten double the transistor density on 7nm TSMC versus 12nm Global Foundries. You should see the transistor budget used to find more single threaded IPC, added up to a decent overall gain on an 8 core part.

The final question will be: how fast can you run it? A totally unchanged 2700X at 4.5GHz would be a formidable opponent for Intel. AMD just couldn't find the clocks on GF's rather average 12nm process- certainly inferior to Intel's current 14nm++. AMD are using a superior process than Intel possess with Zen 2. Incredible.

If Zen 2 can bring that kind of clock speed to the table with a host of enhancements then it'll be a fantastic year for the CPU market.

 

[redhat]

I think it is not about IPC improvement but latency reducing and ST. If they will be able to omit that latency then it will be easy to see 10 - 15% increase in ST

 

[yRaz]

So while I agree it is possible that the hye train might be full steam ahead and that 5-10% is more realistic, I don't think the analysis in the article really makes sense.

 

[Mahone]

**For starters, scientific tasks benchmark wildly differently to games – AMD already beats Intel in scientific and productivity IPC, while falling ~3% behind in games clock for clock.**



That's like not true at all man lol

 

[pcnthuziast]

Nothing any informed and seasoned enthusiast didn't realize from jump.

 

[hahahanoobs]

When AMD is quiet, you should worry.

 

[Cycloid Torus]

I'm going to stick with my $42 6-core Xeon for a bit longer while speculation becomes reality.

 

[godrilla]

When AMD is quiet, you should worry.

There was that bold cinebench benchoff at ces though.

 

[looncraz]

Where in the world do you come off saying Zen stops scaling in Userbenchmark after 3.8GHz?

I have substantial evidence to the contrary, from hours of testing - I also happened to compare to the Zen 2 leak.

[Userbenchmark - 2700X Frequency Scaling](http://files.looncraz.net/zen/2/zen2leak_userbench_st_2700x_freqscale.png)

[Userbenchmark - 2700X vs Zen 2 leak - Floating Point](http://files.looncraz.net/zen/2/zen2leak_userbench_st_2700x_float_compare.png)

[Userbenchmark - 2700X vs Zen 2 leak - Integer](http://files.looncraz.net/zen/2/zen2leak_userbench_st_2700x_integer_compare.png)

[Userbenchmark - 2700X vs Zen 2 leak - Mixed](http://files.looncraz.net/zen/2/zen2leak_userbench_st_2700x_mixed_compare.png)

Userbenchmark's tests are extremely short and don't bottleneck from memory. They very likely fit with in the L1, but certainly within the L2, so there's really very little to prevent them from scaling with frequency... which is exactly why they scale very predictably with frequency.

Using samples from the wild is just a bad idea, in any event. I did all of those runs, three for every 200MHz bump in frequency, a restart with each bump, running DDR4-3200 CL14 (IIRC - I've changed things up quite frequently and was running 32GB 2400 for a file for a memory-heavy workload I had).

 

[Evernessince]

When AMD is quiet, you should worry.

AMD have been quiet ever since Lisa Sue took over the company. Even for the first Zen, which was huge, they had their own presentation and that's about it. All the media guys who used to be loud mouths for AMD now work for Intel's GPU division.

 

[Chadley75]

"We can't confirm nor deny the presence of a 15% ipc increase" should be the title to this article.


We all need to HOPE it is a 15% increase so that Intel is forced yet again into dumping money into RnD.

 

[MaikuTech]

We all need to HOPE it is a 15% increase so that Intel is forced yet again into dumping money into RnD.

They are putting more money and effort into RnD, as long as intel loses more market share to amd.
They will get much more serious and start building real chips that will completely chew out amd completely.
I thought intel would have really cranked the heat up upon the first ryzen 1 batch.

 

[Bullwinkle M]

Just ignore the multi die latency issues

Can an 8 core AMD CPU match Intel's 9900K @ 5Ghz ?

How about 4.5 Ghz ?

4 Ghz ?

Will a 7nm AMD CPU achieve the same performance level as a 14nm Intel CPU while using 1/2 the power ?

Even if AMD's 7nm chips match the performance of Intel's 14nm chips at the same power level, it can either mean that AMD's 7nm is equivalent to Intel's 14nm process or that AMD is way behind on a better process

If we wait several months for AMD to catch up to Intel's current technology, Intel will already have something better

I don't need 32 or 64 cores

I need maximum performance per core (4-8 cores)

AMD cannot compete on a per core basis!

 

[injurer]

This article is not less misleading and made on assumptions. How much IPC gain we will see can be only speculated. I won't be surprised if the gain is 5% but I won't be either if it is 15% or even more. AMD have no reason to be loud about this right now. Leaked information can hurt the sales of their current gen CPUs.

 

[jpuroila]

You're absolutely right in that we have no solid data at the moment, but given that it's a brand new architecture, I think it would be pretty surprising to see IPC improve by only 5%.

 

[Toju Mikie]

Just ignore the multi die latency issues

Can an 8 core AMD CPU match Intel's 9900K @ 5Ghz ?

How about 4.5 Ghz ?

4 Ghz ?

Will a 7nm AMD CPU achieve the same performance level as a 14nm Intel CPU while using 1/2 the power ?

Even if AMD's 7nm chips match the performance of Intel's 14nm chips at the same power level, it can either mean that AMD's 7nm is equivalent to Intel's 14nm process or that AMD is way behind on a better process

If we wait several months for AMD to catch up to Intel's current technology, Intel will already have something better

I don't need 32 or 64 cores

I need maximum performance per core (4-8 cores)

AMD cannot compete on a per core basis!

There was a demo back in January that AMD did and it slightly beat the 9900k and with improved efficiency, but that's all we really know. If I were to guess, they compared both CPUs at stock, and the AMD CPU was still at 4.3 GHz, which means 13%-14% improved IPC over the 2700x. This is what I'm leaning more towards. I don't think AMD will try to win the clock race. I think AMD will put more weight toward efficiency.

 

[ZolaIII]

Thing is that they all calculate the average IPC which means average increase of instructions per clock on both integer & floating point instructions. I doubt integer will see more than 6~7% increase but FP is getting a 2x theoretic boost which translates in real code in some 10~12% increase, 20~25% in good optimized one and around 30~35% in things really written for massive SMP and hard optimized code (best case servers only). So users will see average 13~15% increase while servers could see 20~25% but when you also add duble number of cores & figures groves up to the 2.5x then things get more serious. Still CISC hit its limits long time ago (regarding both design and architecture) and real performance gains are on the other side, to make things worse both Intel and AMD know only for brute force approach (pushing clock's higher utilising high performance lithography libs) so no wonder what ARM archives 4x IPC/W (2x architecturally & 2x from high density libs).

 

[mosu]

First of all, IPC is not about overall gain even it contributes to that and it's agnostic to frequencies used. Second to that the article looks like damage control on Intel's behalf and adds credibility to the 15% gain rumor. Just waiting for the real thing, probably at the end of June or the beginning of July.

 

[LogiGaming]

As I always say, AMD fanboys are the ones responsible for the biggest letdowns coming from AMD. They put unrealistic expectations on the new releases, and then when the products are out everyone gets disapointed.

ADoredTV also predicted Zen+ Would reach 4,5ghz and 5% IPC improvement. It barely clocks at 4,2ghz and IPC improvement was more like 3%.

ADoredTV also predicted VEGA VII would compete with the 2080ti, and there we go, it uses 300w+, barely beats a 1080 on a lot of games, offers similar performance to a 1080ti in other games.

You can keep believing AMD will offer you 12 core CPUs at 5ghz for 400€. Just don´t be disapointed afterwards.

 

[bluetooth fairy]

while falling ~3% behind in games clock for clock

I'm not lazy to check the numbers, and here they are for 6/12 CPUs, directly from the link:

Ashes of the Singularity
2600X still a whopping 11% slower than the 8700K

Assassin's Creed Origins, ultra quality
8700K is a further 14% faster [than 2600X]

Assassin's Creed Origins, high quality
8700K is 12% faster than the 2600X

Battlefield 1, ultra quality
2600X is still 7% slower than the 8700K

Battlefield 1, medium quality
2600X now 10% slower than the 8700K

Far Cry 5
2600X is still 8% slower than the 8700K

Italic font is used for quotation here.

Now that we see the picture in gaming with 1080Ti GPU, we can say, that AMD needs 10-15% IPC increase to be on par with the Intel 6/12 part. Given that Intel CPUs have higher boost clocks, it's not unusual that the audience like to see this margin could have been finally covered.

Verdict:
10-20% is what users actually want to get from Zen 2 because it's the actual margin in gaming. That's why when we talk on forums we mean these numbers, and when we see them we feel it's close to what is just expected from AMD to catch up with Intel.

 

[hahahanoobs]

There was that bold cinebench benchoff at ces though.

Someone needs to tell AMD consumers don't care about Cinebench scores. Just like we didn't care about AoTS scores.

 

[Mindbreaker]

while falling ~3% behind in games clock for clock

I'm not lazy to check the numbers, and here they are for 6/12 CPUs, directly from the link:

Ashes of the Singularity
2600X still a whopping 11% slower than the 8700K

Assassin's Creed Origins, ultra quality
8700K is a further 14% faster [than 2600X]

Assassin's Creed Origins, high quality
8700K is 12% faster than the 2600X

Battlefield 1, ultra quality
2600X is still 7% slower than the 8700K

Battlefield 1, medium quality
2600X now 10% slower than the 8700K

Far Cry 5
2600X is still 8% slower than the 8700K

Italic font is used for quotation here.

Now that we see the picture in gaming with 1080Ti GPU, we can say, that AMD needs 10-15% IPC increase to be on par with the Intel 6/12 part. Given that Intel CPUs have higher boost clocks, it's not unusual that the audience like to see this margin could have been finally covered.

Verdict:
10-20% is what users actually want to get from Zen 2 because it's the actual margin in gaming. That's why when we talk on forums we mean these numbers, and when we see them we feel it's close to what is just expected from AMD to catch up with Intel.

Why are you comparing 8700K to 2600X? 8700K is $370 while 2600X is $200. 2700X is $295 so that would be a better comparison. Or there is the Ryzen 7 2700X AMD50 Gold Edition for $330.

 

[Joseph100]

I made an account and bookmarked this article, just to come back and laugh at author if they are wrong..! (Because they probably are!)