I'm going for what ever is fastest (i.e. best in games) be that 8/12/16 core I don't mind. I want the fastest one for all I do now is game.Biggest choice is the sensible 8C16T or the interesting 12C24T... I don't "need" either but I like to make my hardware last a while
-
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 ***
- Thread starter Jamin280672
- Start date
Not sure I am entirely unbiased, but I like to put my money where the performance to value ratio is, then what I will likely use the system for. I do think Intel were taking the roll mick with their products and prices though. I would like to see them pushed again to bring another great chip to table.... So I can buy it lol.
Do you think we'll get reviews before they go on sale? I'd rather wait for reviews but I can also see there being very high demand especially for the fastest "for gaming" version.
Maybe not. But clock speeds will be available. And as the architecture is the same across all chips we can assume the fastest Ghz chip will be the best in games. I assume?
Of course, I was forgetting we'd get the specs up front. Going to be a preorder for me if the clocks look to be in the right ball park. I don't think I've ever bought any hardware on release so could be interesting!
Same, I've never bought anything on release but I'm really confident in this release. I'm sure Lisa will show us some benchmark ("3700x vs 9900k") at Computex that will be all we need to know. But yeah, even a marginal improvement on Ryzen 2000 and I am buying, I'm well due an upgrade and I've a fair amount saved up now not having bought into Nvidias latest offerings.
Yes I should have £800+ for an upgrade (cpu, ram, motherboard). Obviously happy to spend less than that.
The last CPU I bought new was AMD.
Then bought a used Q6600 in 2008 from the MM which lasted me until 2015.
What a chip that was.
Bought a used X5650 system again from the MM in 2015, still going strong.
So I'm another CPUFluid user.
Then bought a used Q6600 in 2008 from the MM which lasted me until 2015.
What a chip that was.
Bought a used X5650 system again from the MM in 2015, still going strong.
So I'm another CPUFluid user.
You should have bought an AM3 Phenom X4 which you could have upgraded to the six-core Phenom II X6 1100T. It's almost double the performance of a Q6600.
It's not going to be worse than Ryzen 2000. That much is certain. And I would be perfectly happy for Ryzen 2000 performance. So I am pretty safe to pre-order.
Caporegime
- Joined
- 8 Jan 2004
- Posts
- 32,622
- Location
- Rutland
Yup same thoughts here. Sit back and see what the reviews show. Especially if the 16c32t part isn't available at release.
Might go all out on Threadripper 2 if that's interesting.
The slight latency that you get from the separate IO die isn't going to mater to games, the GPU is at the other end of the board not even attached to the CPU's PCB.
Ryzen 1000 / 2000 has "inter-core latency" which if didn't exist should improve gaming performance, despite the IPC on Ryzen 2000 at least being about the same as Coffeelake and about 10% higher than SkyLake-X which also suffers from inter-core latency.
Ryzen 3000 shouldn't have any of that inter-core latency, not on the 8 core one as the CPU Cores and Cache are in a Ring Bus on the same die.
Ryzen 1000 / 2000 has "inter-core latency" which if didn't exist should improve gaming performance, despite the IPC on Ryzen 2000 at least being about the same as Coffeelake and about 10% higher than SkyLake-X which also suffers from inter-core latency.
Ryzen 3000 shouldn't have any of that inter-core latency, not on the 8 core one as the CPU Cores and Cache are in a Ring Bus on the same die.
It's a reasonable assumption, but as d_brennen said, we don't know the full details of how the chiplet + I/O die latency is going to pan out, nor if a chiplet is now a single 8-core CCX. It's not outside the realms of possibility that single chiplet SKUs would perform faster than double chiplet SKUs in certain circumstances.
8-core using a single CCX will be great, but if the higher cored CPUs clock higher, then you'd no doubt benefit more from clock speed. I hope the 8-core isn't gimped though, that'd suck.
I can't see AMD shooting themselves in the foot by gimping anything, although I don't see the Ryzen 3s having the best silicon out of the bins. Even if they do end up 6 cores.
Associate
- Joined
- 2 Jan 2019
- Posts
- 617
6c at current gen clocks, with improved IPC and similar/reduced latency, combined with PCIe4, will be the best performance/price CPUs, and will definitely be high demand irrespective of whether the gaming benchmarks are still behind (marginally) a 9900K.
I supposed you preordered a mid range (6 core) Bulldozer chip, as it couldn't be any worse than Thuban or Sandy
I have no reason to believe Zen 2 will be worse in any metric but it would be foolhardy to claim otherwise until we have numbersI have been doing some research into the HTT/SMT affect on gaming and ended up finding this thread here which has links to other threads, its a gpu discussion, but what is interesting is the affect of AMD and Nvidia policies on CPU usage, and it has a lot of answers to the questions I had.
https://forums.guru3d.com/threads/discussion-on-async-compute-from-amd-prospective.423079/
So to summarise.
On DX11, nvidia look at hardware threads available (so logical threaded cpu's have more of them), and then offload draw calls from the main cpu core to those threads to boost performance (this has been primarily whats been holding back AMD on DX11 as well).
Also AMD prefer developers to do parallel execution, whilst nvidia prefer concurrent. Concurrent seems much less efficient when reading how it works. Some of you on here may remember I have commented a few times I have observed that when games are loaded that 2/3 of my cpu utilisation is on context switches. Yet when I did things like limit games to a single core to get rid of that overhead performance often dropped. After reading up on geforce CMDLIST I now have all the answers to this.
So basically nvidia drivers are very wasteful with cpu resources, but ultimately in DX11 it often gains performance as it bypasses a bottleneck, at the cost of needing more cpu power. This explains why in certain games logical threaded cpu's have seen gains as the amount nvidia offload from the main cpu core, is directly determined by the amount of cpu threads available to it. It isnt because SMT/HTT cpu's are faster (which I already knew they not), its because it offloads more as a result of more threads been available, basically there is smaller slices on the pie as a result of there been more slices and the slice allocated to the main core/thread is smaller.
I recommend reading that thread as well as many of the spin off links, is some very interesting information there. Mainly on gpu's and the way they interact with cpu's for gaming performance.
How much of this applies to older DX9 I have no idea, but of course we know vulkan and DX12 are engineered for high threads from the ground up, so those will benefit from higher cpu cores pretty much all the time.
https://forums.guru3d.com/threads/discussion-on-async-compute-from-amd-prospective.423079/
So to summarise.
On DX11, nvidia look at hardware threads available (so logical threaded cpu's have more of them), and then offload draw calls from the main cpu core to those threads to boost performance (this has been primarily whats been holding back AMD on DX11 as well).
Also AMD prefer developers to do parallel execution, whilst nvidia prefer concurrent. Concurrent seems much less efficient when reading how it works. Some of you on here may remember I have commented a few times I have observed that when games are loaded that 2/3 of my cpu utilisation is on context switches. Yet when I did things like limit games to a single core to get rid of that overhead performance often dropped. After reading up on geforce CMDLIST I now have all the answers to this.
So basically nvidia drivers are very wasteful with cpu resources, but ultimately in DX11 it often gains performance as it bypasses a bottleneck, at the cost of needing more cpu power. This explains why in certain games logical threaded cpu's have seen gains as the amount nvidia offload from the main cpu core, is directly determined by the amount of cpu threads available to it. It isnt because SMT/HTT cpu's are faster (which I already knew they not), its because it offloads more as a result of more threads been available, basically there is smaller slices on the pie as a result of there been more slices and the slice allocated to the main core/thread is smaller.
I recommend reading that thread as well as many of the spin off links, is some very interesting information there. Mainly on gpu's and the way they interact with cpu's for gaming performance.
How much of this applies to older DX9 I have no idea, but of course we know vulkan and DX12 are engineered for high threads from the ground up, so those will benefit from higher cpu cores pretty much all the time.
Last edited:
Soldato
- Joined
- 1 Dec 2015
- Posts
- 18,512
Problem is if you Dont Pre-order when the FIRST batch is up for Order - they will firstly sell out fast regardless of people waiting for reviews; then once people digest reviews they will still flock to preorder , specially if it beats their 4 core i7s or first gen Ryzens by a lot .
once stock is gone... this will allow resellers to UP pricing due to demand and supply . We've seen it happen with EVERY intel and Nvidia products in the last 3-4 years . Only Distributors are limited to selling at set price , even at pre-order and 10th product batch, resellers are able to inflate prices to match demand .
I have a feeling boards will follow the same if a lot of Intel owners jump ship !
personally have a feeling 6/8 core will sell like hotcakes if it hits 5ghz and matches or is lower then current pricing with lower heat . PS5 details now signalling 8 cores is the high sweet spot for future gaming .
X570 Aorus Pro Wifi + 8 core 5ghz for me at hopefully same price as 9900k chip