• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

Sargatanas2511 said:
It's funny to me that Intel have had 2 years of Ryzen, plus the hype for ~6 months before that to try and counter it and still nothing is being shown other than more cores more cores more cores.
Click to expand...

CPU development plans are set in motion years in advance. Intel will have made predictions on how they thought Ryzen was going to perform and what configurations it was going to be available in. Then they would have set their development cadence to these assessments.

Unfortunately a triple whammy of Intel not predicting the massive leap in IPC and scalability of Ryzen, along with their 10nm fab troubles, and AMD building Ryzen to compete against the 'threat' of Icelake (which has been pushed back numerous times) means that they are massively on the back foot with Zen 2 looming.

Hence the scrabbling around to re-purpose anything they can get their hands until they can make a true answer to Zen 2.
 
HeX said:
CPU development plans are set in motion years in advance. Intel will have made predictions on how they thought Ryzen was going to perform and what configurations it was going to be available in. Then they would have set their development cadence to these assessments.

Unfortunately a triple whammy of Intel not predicting the massive leap in IPC and scalability of Ryzen, along with their 10nm fab troubles, and AMD building Ryzen to compete against the 'threat' of Icelake (which has been pushed back numerous times) means that they are massively on the back foot with Zen 2 looming.

Hence the scrabbling around to re-purpose anything they can get their hands until they can make a true answer to Zen 2.
Click to expand...

AMD weren't quiet about how excited they were over Zen and stated many times their goal was a 40% uplift in IPC in gen 1 (which admittedly they exceeded). They stated this many times before launch or even tech demoes so Intel were duly warned.

I understand it takes a long time to redesign cpus at an achitectural level but I kind of hoped 2 1/2 years down the line we would have some kind of information as to what Intel will show next. They got very lazy and happy to milk what they had and it's evident to everyone now.
 
HeX said:
Hence the scrabbling around to re-purpose anything they can get their hands until they can make a true answer to Zen 2.
Click to expand...

Hell, Intel could've played up the perceived arrogance and gone "bitch please, we're Intel. This is just a bump in the road." and then actually not humiliate themselves by repurposing kit that's DOA or pushing things past sensible levels. Literally just shut up, knuckle down and make something.

I know plans are put years in advance and you can't bang out a new arch overnight (something which I continually point out to people crying about AMD's graphics woes), but with the money Intel have it'd probably be more cost-effective and efficient to radically change those plans and start something else.

Christ, maybe even go to Samsung and say "don't suppose we can use your 7nm fabs, can we?". What's the point in continually throwing money at your bespoke 10nm process because you're the "technology leader" when AMD have taken that title away from you?
 
Lots of very large companies have been complacent whilst milking their customers, not just Intel.

We're looking at a 9% single thread performance increase based off that one benchmark, if my numbers are correct.
Let's then look back at the original +40% Ryzen IPC target, and how it looks now.
Assuming they were at 100 (used for illustrative purposes only) to start with, the jump to Ryzen would have put them at 140, which would still be behind by some margin. Now that's 2 years ago, and Intel haven't had any ST IPC gains themselves in that time (or at least only marginal).
Instead, what we got was 52% with Zen, another 3% with Zen+, and around 9% with Zen 2.
100x1.52x1.03x1.09 is approximately 171, so a 71% increase on where they originally came from.
Intel were not 71% ahead to begin with, though not too far off that in all honesty.
However, what AMD also brought to the table was huge increases in core count, huge increases in power efficiency, and huge reductions in relative pricing. Even the most cautious Intel planners couldn't have predicted that.
It's kudos to AMD for pulling it off.
I have no doubt that Intel will be back with a vengeance, but the next two years at least they won't be top dog in performance.
 
52% IPC was vs Excavator, that CPU was never on the desktop, it found its way into some laptops but that was it.

Vs Piledriver, the previous AMD CPU we all know, the FX-#3## series the IPC is up 70%+
 
NinjaCool said:
Possible 12-Core 24-Thread Third-Gen Ryzen ES CPU spotted at UserBenchmark!
https://twitter.com/TUM_APISAK/status/1088264161191022593?s=09
https://www.tomshardware.co.uk/amd-matisse-third-gen-ryzen-benchmark,news-59832.html
Click to expand...

as a heads up, this is what roughly AMD send Board vendors in terms of first Gen Engineers samples . a lot LOWER speed clocks and default ram speed .

retail version caught all of the vendors off guard with final clock speeds, ram scaling performance and ability to house faster ram ( though this is part IMC and part board memory design and bios)

hoping AMD follows the trend of Retail being a lot faster , and vendors like X470/B450 being a lot better supported/Designed.

safe to say , X570 VRMs are beefed up , don't think they'll be to many 4+3 phase designs and more 8+1 plus :D
 
Potatowithearsontheside said:
https://cpugrade.com/articles/cinebench-r15-ipc-comparison-graphs/
This article suggests that Intel had a larger ST IPC lead to begin with, though also that AMD made larger gains than my above post suggests.
The conclusion remains unaltered though.
Click to expand...

Yeah, @ 3Ghz 2700X = 129 vs 132 8700K it amounts to about 3% to the 8700K

It matches the chart i like to use... tho it also shows MT and in that Zen+ is about 4% ahead of Intel, somewhat better SMT on the AMD side.

dcQvZib.png
 
That's a really good chart.. never seen Bulldozer / Excavator compared to all others like that... it's good to see it.
 
Potatowithearsontheside said:
The lower end of my previous IPC estimations used when calculating potential CES ES CB MT scores was 9%, which IIRC suggested around 4.3GHz for that ES.
I think now it is only a matter of how high the clocks can go.
Click to expand...

I have a feeling the same chip will reach the 2200 cine' points mark when released, i think there's between 5 and 10% more in it vs the CES demo. the 9900K would have to hit 5.2Ghz all core to match 2200 points.
 
humbug said:
I have a feeling the same chip will reach the 2200 cine' points mark when released, i think there's between 5 and 10% more in it vs the CES demo. the 9900K would have to hit 5.2Ghz all core to match 2200 points.
Click to expand...
I pretty much agree.
2200 represents another 6-7% clockspeed over the ES, so maybe 4.6GHz on all cores as an upper limit.
Doesn't tell us about single core clocks though unfortunately. AMD will likely be needing to hit 4.9-5.0GHz on at least one core to be able to claim any sort of a lead, and unless latency is improved then they may need even more than that for gaming.
 
Potatowithearsontheside said:
I pretty much agree.
2200 represents another 6-7% clockspeed over the ES, so maybe 4.6GHz on all cores as an upper limit.
Doesn't tell us about single core clocks though unfortunately. AMD will likely be needing to hit 4.9-5.0GHz on at least one core to be able to claim any sort of a lead, and unless latency is improved then they may need even more than that for gaming.
Click to expand...

I'm not worried about gaming, i think from a performance perspective having the cores clustered together on a single die actually already fixes the "Inter core Latency" given that those core are operating in a Ring Bus, not 2x 4 core CCX's like Zen+, the dual chiplet version maybe different but its not going to be any worse than Zen+ Threadripper and they do just as well as the 2700X in games, at least up to the 2950 given that Windows gets confused with 24 and 32 cores.

At stock (4.025Ghz) all core boost the 2700X is around 15% slower than the all core boost 4.7Ghz 9900K. i have laid that out with a side by side video review some pages back, the 3600X only needs to gain that in gaming performance over the 2700X to match the 9900K, i think AMD can manage that, Zen+ is the same amount faster in games vs Zen as it is in Cinebench.

No worries there IMO :)
 
Even something like a 4 core 5Ghz boost/8 core 4.8 /16 core 4.7 would be great overall.

Hoping they can hit the 5Ghz mark with 4 cores for the gaming crowd and still provide huge multicore performance.
 
Sargatanas2511 said:
Even something like a 4 core 5Ghz boost/8 core 4.8 /16 core 4.7 would be great overall.

Hoping they can hit the 5Ghz mark with 4 cores for the gaming crowd and still provide huge multicore performance.
Click to expand...

Generally, the high clocked chips are binned for the high priced CPUs. More likely to see 5Ghz on the R7 tier than the R3/5 tier.
 
You must log in or register to reply here.
Back
Top Bottom