@Minstadave You're proberly right I'd not really thought about it in that way.
-
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 3 (5000 Series), rumored 17% IPC gain.
- Thread starter AStaley
- Start date
- Status
Now we know the 5800X is more powerful than the 5600X, would you prefer that the 5800X was the ~£300 replacement for the 3700X, or that the 5600X was the £300 replacement for the 3700X?
This isn't hard. It's a question of AMD giving you less for your money this time around. The argument does not hinge on the 5600X being faster in some metrics than the 3700X (some metrics). The 8 core 5800X exists and it could have been £350 instead of £450.
Instead the 5600X is the £300 part. People are saying it's the successor to the 3700X based on its price. Ideally the price would be lower and the 5800X would be the successor.
There is no doubt that the 5800X (8 core part) is more powerful than the 5600X (6 core part).
Think about this. Had the 5600X been a £250 part, it would have been described as the successor to the 3600X... instead because the price is higher it is called the successor to the 3700X. Even tho it is mostly only faster in gaming tasks, and not universally faster in multi-core tasks.
Associate
- Joined
- 14 Aug 2017
- Posts
- 1,196
More pcie lanes on the threadripper platform* IIRC. Maybe better heat dissipation on the larger package too?
(*For the AM4 processors there are 16 for GPU, 4nvme, 4 for chipset, which on x570 provides another 16 multiplexed off that, so 36 usable lanes. There are 88 on the upcoming threadrippers)
Last edited:
possibly more cache too, amd puts more cache on its other Threadrippers
What if they kept the inner workings of the chips a secret and only revealed the performance of the CPU?
Could you buy the faster of two different $300 CPUs and be happy?
Would you be unhappy if you learned the faster CPU had less cores/clock speed/cache/whatever...?
Do you want *performance* or do you want to collect cores for the sake of cores?
I would buy a single-core, 1ghz chip, *if it was faster* than my 8-core 4.5 ghz chip, and not give a damn.
The point is, as a gen-on-gen upgrade, 3700X -> 5600X is worse than 2700 -> 3700X.
You don't need to know the internal workings of the chips to draw that conclusion. You don't even need to know the number of course.
You're simply getting a smaller upgrade than you might have expected.
3700X -> 5800X should have been the upgrade we are all talking about, but the 3700X was ~£300 and the 5800X is £440.
Turns out the 5600X is the "upgrade" for 3700X users. And it's not all roses. Single thread decent upgrade, but you're losing two cores which takes away a considerable amount from multi-threaded workloads.
If AMD pursue this and give us a 6300X (4 core) for £300, you'd lose another two cores and gain a bit more single-threaded perf. And then I hope people would start to say, "No, this is ridiculous. We've gone from 8 core to 4 core at the £300 price point. And all we get in return is a bit more single-threaded perf."
OK I might not be explaining this very well so let's look at a more extreme example.
Let's assume that the 5600 is a £250 part and the <£200 5000 series part is a 4-core.
The previous <£200 part was the 6 core 3600.
Now let's say the "upgrade" for the 3600 is the 5400X, a 4-core part.
You will gain ~20% single-thread perf and lose 33% of your cores.
Would that make sense as an "upgrade" for 3600 users? You tell me.
Let's assume that the 5600 is a £250 part and the <£200 5000 series part is a 4-core.
The previous <£200 part was the 6 core 3600.
Now let's say the "upgrade" for the 3600 is the 5400X, a 4-core part.
You will gain ~20% single-thread perf and lose 33% of your cores.
Would that make sense as an "upgrade" for 3600 users? You tell me.
You state a 33% reduction in cores without stating any performance gain or loss in multi-thread performance. Why is that? Are just assuming that more old cores will always outperform fewer new cores?
This may be where the communication is breaking down. Without even looking it up, I'm sure there are old 8-core parts that get beat in multithreaded tasks by newer 6-core parts.
I bet a 5600X will outperform a 1700X at...well...everything.
I asked a very specific question.
How many people would happily trade ~20% better single-thread performance for a 33% reduction in cores... from one generation to the next.
I.e. a 3600 to a 5400X.
I'm not comparing a Bulldozer FX-8 to Zen 3, that should be quite clear.
e: Just to be clear, this isn't a trick question. You can assume that ST performance is the same across all cores (it mostly is, with some cores being marginally better than others).
Pretty much every gamer, or in fact anyone not spending a lot of time encoding, rendering, compiling etc. The 5600X is the better chip than the 3700X for many/most users.
So gamers would happily trade their 3600 for a quad core 5400X that had ~20% better ST perf?
Really?
I'm a gamer and I sure as hell wouldn't do that.
It's a red herring.
How fast is it?
How much does it cost?
This is all that matters. The fact that the 5600X can't quite beat a 3700X in multithreaded tasks is an issue, but it's the performance at issue. not the core count.
A 7-core 5650 (if AMD were to make one) would easily outperform a 3700X in everything.
It would have less cores...and it would be faster....at everything.
You can actually be faster in multithreaded tasks with fewer cores if the fewer cores are strong enough.
Bingo! The only caveat is "how fast is it?" totally depends on what you are doing. Gamers and encoders give different answers.
If you go back a few posts you'll see there is a thread of a conversation here. We moved on from the 3700X -> 5600X to a 3600 user "upgrading" to a 5400X quad. It's a counterpoint to the people who say, "cores don't matter at all."
Same price, give or take. 20% faster ST perf (doesn't matter how, IPC improvements or clock bumps).
33% loss of cores.
That's all the information you need to make the judgement based on your specific use cases.
And that you're starting with a 6 core 3600, and moving to a 4 core 5400X.
Again, it's not a trick question.
So for gaming you can give up 25%, 33% or even 50% of your cores for a 20% ST improvement?
That is clearly not the case.
Let's say you start with a quad core and move to a dual core with a +20% ST perf increase. Good deal, for gamers?
The idea that "cores don't matter" is just wrong.
Now you're just being silly. The point is "cores don't matter", performance does. Treat the CPU like a black box, don't even consider the internal architecture. Performance in the task at hand is all that matters. I couldn't care less if my processor had 2 cores or 64 if delivered 150fps in the game I was playing or encoded the video in the time it took to take a leak.
Under what circumstances will you get more performance moving from a 6 core to a 4 core with 20% better ST perf per core?
Apart from ST apps, which is a given?
Do you think most games in 2020 are still unable to use more than 4 cores? I mean that's fine if you do think that, but I'm fairly sure it's not true for all games today.
- Status