AMD Ryzen 5 5600X takes the PassMark 1T CPU crown

[1/2Genius1/2Wit]

The MT scores don't seem to have a massive improvement. So it says to me most of the improvement in single core scores is down to clockspeed,and the reduction in latency moving to a unified 8 core CCX/CCD.

The clock speed improvements aren't particularly dramatic, though - not enough to change the score by that much. It's possible that smaller improvements in the multi-thread results are due to the chips hitting power limits.

Not long until we get some reviews and get to find out!

[kompukare]

The MT scores don't seem to have a massive improvement. So it says to me most of the improvement in single core scores is down to clockspeed,and the reduction in latency moving to a unified 8 core CCX/CCD.

BTW,will Hexus test some older games when they get Zen3?? I have a feeling its mostly the games based on older engines which will see the biggest improvements overall.

But isn't part of the rational of SMT that it uses parts of the core which would otherwise be idle when there's a holdup like a miss-predict somewhere?
In which case, the better the main thread gets in terms of branches predicted, pre-fetch etc. then there are less idle resources to give to a second thread.
In which case, for Zen4 they might add more duplicated resources so the secondary thread isn't stalled by the primary thread being to efficient.

[1/2Genius1/2Wit]

So the issue here is when both are tweaked,I can't see the Ryzen 5 5600X being really that much ahead. Also as AMD downgraded the CPU cooler with the Ryzen 5 5600X,I assume you will need to budget for a better cooler to tweak it.

In the end it probably makes more sense to see how the Ryzen 5 5600 non-X and Ryzen 7 5700X pan out next year,but again if the £ gets weaker,it might not help either.

The 5600X is a 65W chip, so while that Wraith Stealth cooler sucks, there's really no need to tweak the CPU at all to get the advertised performance - whereas you'd have to get a decent cooler for the Intel chip and tweak to reach equivalent performance. You'd also be looking at nearly twice the power consumption for that performance.

The non-X Ryzen 5000s will definitely be the better value options. Unless something changes dramatically from previous ranges, you'd be looking at performance within 5% of the X models that can be gained back with tweaking, for ~£40-50 less outlay.

[DanceswithUnix]

But isn't part of the rational of SMT that it uses parts of the core which would otherwise be idle when there's a holdup like a miss-predict somewhere?

Yes, but it neither knows nor cares which of the two threads is a "main" thread and schedules the two evenly. It has to do things like that or else you get odd things happening, such as thread priority inversions.

In which case, the better the main thread gets in terms of branches predicted, pre-fetch etc. then there are less idle resources to give to a second thread.

Nah, even if one thread were given priority there are still places where a thread is held up on outstanding memory reads, dependent register operations etc.

In which case, for Zen4 they might add more duplicated resources so the secondary thread isn't stalled by the primary thread being to efficient.

Careful, you're designing Bulldozer there and we know how that ended up

The thrust with current (non construction core) threading is to make the CPU so wide that lightly threaded tasks can go really fast. The threading stops it being a waste of silicon on heavily threaded tasks.

[CAT-THE-FIFTH]

The clock speed improvements aren't particularly dramatic, though - not enough to change the score by that much. It's possible that smaller improvements in the multi-thread results are due to the chips hitting power limits.

Not long until we get some reviews and get to find out!

Not if Passmark is more latency dependent,since AMD Zen2 has higher IPC/PPC than current Intel CPUs. Most of these changes seem to affect lightly threaded performance,and noise is these CPUs might be boosting closer to 5GHZ.

The 5600X is a 65W chip, so while that Wraith Stealth cooler sucks, there's really no need to tweak the CPU at all to get the advertised performance - whereas you'd have to get a decent cooler for the Intel chip and tweak to reach equivalent performance. You'd also be looking at nearly twice the power consumption for that performance.

The non-X Ryzen 5000s will definitely be the better value options. Unless something changes dramatically from previous ranges, you'd be looking at performance within 5% of the X models that can be gained back with tweaking, for ~£40-50 less outlay.

Not if you ever used a Wraith Stealth. It runs very hot,and a Wraith Spire is nearly 20C cooler in my own testing,and after more extended usage you could see throttling and that was on a 65W TDP CPU. AMD is most likely giving numbers using aftermarket cooling not its stock cooler IMHO.

Also the Core i5 lacks thermal velocity boost,so actually gains the most from manual overclocking of all the Intel K/KF series SKUs. The problem is the Core i5 10600KF is now £230,and that is between £50 to £60 cheaper than a Ryzen 7 5600X. That will more than cover a decent cooler,and GN showed the Core i5 itself gained alot from cache and memory tweaks.

The issue is the Ryzen 5 5600X and Ryzen 7 5800X pricing sucks,and is more per core than the already overpriced Intel 6C and 8C CPUs. To put this in context,the Ryzen 9 5900X is cheaper per core to a significant degree.

If the Ryzen 7 5800X matched the same per core pricing it would be £350. If the Ryzen 5 matched the same per core pricing it would around £260.

But isn't part of the rational of SMT that it uses parts of the core which would otherwise be idle when there's a holdup like a miss-predict somewhere?
In which case, the better the main thread gets in terms of branches predicted, pre-fetch etc. then there are less idle resources to give to a second thread.
In which case, for Zen4 they might add more duplicated resources so the secondary thread isn't stalled by the primary thread being to efficient.

It might be interesting to see the SMT "yield" for both Zen2 and Zen3,to see whether scaling is different??

[edmundhonda]

noise is these CPUs might be boosting closer to 5GHZ

AMD's own spec page says 4.6Ghz, even for specifically single threaded 'bursty' workloads. This Passmark score is 21% over the 3800XT (4.7ghz boost) so it's not clockspeed. Unless someone's playing silly buggers with LN2.

Won't comment on pricing, beyond agreeing that the current numbers look bad, but without knowing real world performance and pricing it's pointless to try and make a conclusion.

[CAT-THE-FIFTH]

AMD's own spec page says 4.6Ghz, even for specifically single threaded 'bursty' workloads. This Passmark score is 21% over the 3800XT (4.7ghz boost) so it's not clockspeed. Unless someone's playing silly buggers with LN2.

Won't comment on pricing, beyond agreeing that the current numbers look bad, but without knowing real world performance and pricing it's pointless to try and make a conclusion.

Some of the accurate leakers on Twitter have hinted the Zen3 CPUs are boosting closer to 5GHZ in some of the leaked benchmarks(one of the models hit 5GHZ). There was a tweet which showed the purported clockspeeds.

Zen2 by extension rarely gets to its maximum stated boost for long and AMD got a ton of flack for this. Zen3 stated boost clocks appear to be conservative. So around 10% higher clockspeed,higher IPC,etc would make up that 20% relatively easily.

Pricing has already been released by AMD,and pre-orders of the Ryzen 5 5600X are around £280~£290,which fits in with the currency conversion.

AMD already gave out gaming numbers(hidden as a price/performance metric). They state the Ryzen 5 5600X is 13% better gaming performance value than a Core i5 10600K when both are priced equally. So that means 13% better gaming performance,for £50~£60 more money. In the end Intel for years had faster and faster six core CPUs.Yet despite this they were overpriced. AMD is now making a faster six core CPU,but the deciding to charge more. Its the same problem.

We are in danger of going back to the years of stagnation we had with quad cores - Intel for years was just releasing faster and faster quad cores. A Core i7 7700K would wipe out a Core i5 2600K in a number of modern games and applications.

So what happens if Intel next year makes a six core CPU faster than a Ryzen 5 5600X?? Will it mean they will charge over £300?? Soon we will be back to October 2017 pricing! Just look at how GPU prices have gone.