'

AMD's Zen 3 architecture is expected to deliver an 8+% IPC boost

Boosted IPC plus higher clocks will make for a compelling next-gen product

AMD's Zen 3 architecture is expected to deliver an 8+% IPC boost

AMD's Zen 3 architecture is expected to deliver an 8+% IPC boost

10 days ago, were were one of of the first websites to report on AMD's Zen 3 architecture and the radically redesigned aspects of the company's next-generation processor designs. Let's be clear about this again; this information isn't a mere rumour. Tt came from an AMD representative, Martin Hilgeman, and from an official AMD presentation at the HPC AI Advisory Council's 2019 UK Conference. AMD may not want this information in the public domain, but that doesn't mean it isn't there. 

Now, a new report has claimed that AMD's next-generation Zen 3 architecture will deliver an 8+% IPC boost over Zen 2 alongside a 100-200MHz boost in clock speeds. When combined, this will provide a single-threaded performance boost of over 10% across many workloads. Even so, it must be remembered that IPC boosts are always application-specific. Even if 8% is AMD's average IPC boost for their next-gen architecture, Zen 3 could offer much larger, or lower, performance boosts in specific workloads. These IPC boost claims are little more than a rumour at this time. 

As we have said previously, AMD's Zen 3 architecture fundamentally changes the core topology of AMD's 8-core Zen CPU dies. Instead of relying on quad-core CCX's in each 8-core CPU die, Zen 3 appears to be moving to a unified core design that keeps all eight CPU cores within a single group on each die. This allows AMD to combine each chip's split L3 caches into a singular unit, granting each core direct access to a larger bank of L3 cache. 

This design change will decrease many of the internal latencies of each of AMD's Zen 3 CPU dies. This design shift potentially accounts for some of AMD's IPC boost in specific workloads. AMD's slides also hint at an increase in L3 cache size, which could lead to further performance gains. While this change will not impact the latencies between separate CPU dies, this reduction in internal latencies will have enormous implications for many workloads, especially in consumer-systems that use single 8-core Zen CPU dies. 

PC gamers should also see this potential reduction in CPU latencies and increase in cache size as a good thing, given how AMD's improved L3 Cache design/ "GameCache" boosts Zen 2's gaming performance. 

AMD reveals early Zen 3/Milan architecture details and Zen 4/Genoa plans  

While 7nm+ isn't expected to offer insane performance/efficiency improvements as AMD's previous node shift, it looks like AMD plans to make Zen 3 another substantial release thanks to smart processor design and refinements that target many of Zen's current shortcomings. 

Given what AMD has shown us regarding Zen 3, we find it easy to believe that AMD has an 8+% IPC boost in store for us. The question now is how long will it take for Zen 3 to come to market, and whether or not AMD can deliver much of a clock speed boost over Zen 2. While AMD has confirmed that Zen 3 designs complete, it remains too early to know anything definitive about core clock speeds, as these are decided very late in a processor's development cycle.  

You can join the discussion on AMD's Zen 3 core design and rumoured IPC boost on the OC3D Forums

«Prev 1 Next»

Most Recent Comments

14-10-2019, 12:22:46

Master&Puppet
It doesn't sound like much but a further 8% and a 100-200mhz boost on top of the ryzen 3000 will look pretty good vs Intel.

Clock for clock ryzen 3000 is what, 5-9% faster than Intel coffee refresh? Add this news together and Intel are in real trouble. My guess would be that the Ryzen 4900X would look something like 4.7-4.9Ghz, +8% IPC and 12C/24T for 449-499$ and it will probably be the last generation of chips on the existing AM4 platform to boot.

To clarify it means Intel will loose the performance per core title, be essentially non-existent in most multicore environment and will be less power efficient. Interesting times.Quote

14-10-2019, 13:03:59

NeverBackDown
I wouldn't expect anything more than 4.3-4.5ghz. Ryzen can barely hit those clocks reliably now.Quote

14-10-2019, 13:32:10

tgrech
Depends a lot on how well TSMC's EUV node matures I think, AMD can clearly make chips Ryzen that can hit 4.7Ghz, but seemingly only on like one core per 16 core CPU with current 7nm. If the node matures well then we might see like, one part hitting 4.9Ghz boosts, but yeah it's probably going to be closer to ~4.1Ghz for base clocks. Worth noting though TSMC's 7nm+ doesn't use EUV on the bulk of the chip and that's not what will deliver the benefits here but the general maturing of their 7nm tech will help.Quote

14-10-2019, 17:02:51

demonking
Quote:
Originally Posted by tgrech View Post
Depends a lot on how well TSMC's EUV node matures I think, AMD can clearly make chips Ryzen that can hit 4.7Ghz, but seemingly only on like one core per 16 core CPU with current 7nm. If the node matures well then we might see like, one part hitting 4.9Ghz boosts, but yeah it's probably going to be closer to ~4.1Ghz for base clocks. Worth noting though TSMC's 7nm+ doesn't use EUV on the bulk of the chip and that's not what will deliver the benefits here but the general maturing of their 7nm tech will help.
Agreed on all but i think that another part is heat. Whilst overall these chips produce less heat (compared to previous nodes), because of there smaller size the heat is compacted into a smaller area. Dissipating a similar amount of heat on a smaller surface area is much harder, takes longer and requires more energy. There is no easy way round this either, its basic physics, getting the heat transferred to the IHS faster is the problem in this case not the cooler on top of it. We may well be pushing the limits of silicon transistors here and an alternative material or more exotic cooling methods or transistor shapes may be needed to push much further.Quote

14-10-2019, 19:07:21

NeverBackDown
Everyone talks about heat density but these CPUs are running just as cool. Even with the stock heatsinks they are not much different than the 1 and 2 series. It doesn't make much sense to me why everyone keeps bringing it up. If you had everything compacted in the center sure, but these CPUs are using the entire surface area as they are each in their own modules. Which reduces complexity, which reduces the number of transistors in one given area, which reduces heat, etc.

Intel and their architecture, given they keep decreasing overall CPU die size area and increasing overall GPU die size, they will more than likely run into that issue. Unless they can reduce power draw so significantly that it doesn't output much heat to begin with and goes unnoticed until pushing to the extreme.Quote
Reply
x

Register for the OC3D Newsletter

Subscribing to the OC3D newsletter will keep you up-to-date on the latest technology reviews, competitions and goings-on at Overclock3D. We won't share your email address with ANYONE, and we will only email you with updates on site news, reviews, and competitions and you can unsubscribe easily at any time.

Simply enter your name and email address into the box below and be sure to click on the links in the confirmation emails that will arrive in your e-mail shortly after to complete the registration.

If you run into any problems, just drop us a message on the forums.