Future PC Upgrade (Skylake or Zen)

[mysticjbyrd]

Hey guys, I am new here!

I am considering upgrading my PC in the near future, and was looking for some advice from people that follow the tech industry more than myself.

My current setup is an Intel 930 & HD7950 Boost. I don't buy top of the line components, as I have seen/found in the past that you are merely paying for the R&D. I am looking for just the high-midrange stuff that will last quite a few years. If I bought a CPU I would certainly not want to spend more than $250. I don't do any video editing or anything else with the PC.

I don't feel I need an upgrade atm, though I am annoyed at the features of the old MB, but I am really curious as to what you guys think my best options in the near future are.

There are several concerns I have about ensuring a future proof build.
#1 DX12 superior threading capabilities
This new API two should potentially see an influx in the current standing of AMD's performance. Current AMD chips won't outright outperform Intel, but it should at least lessen the gap. Particularly, in the cost / performance scale. Right?

#2 AMD's Zen
The potential of this new CPU intrigues me, despite knowing AMD is completely shitting the bed in the CPU market as of late. Perhaps, it's all hype, but I have a notion to wait and see how things pan out. Is there any hope AMD can come back into the market, and at least lower Intel's pricejacking?

#3 Simultaneous multithreading prevalence in gaming
This seems to becoming far more prevalent in gaming, and AMD is supposedly going to finally utilize this technology in their new zen arch. It's seeming an expensive feature in Intel CPUs as well. Perhaps this too could help with AMDs price/performance.

#4 14 nm tech lithography
Intel is already using this technology in it's Skylake arch. However, Samsung is doing a deal with AMD for 14nm wafers. I heard that Zen wouldn't use 14nm wafers, but it was just some random post. Is it even possible for AMD's Zen to catch up if it's not using the same scale?

#5 Cannon lake (or w/e is after intel's skylake)
I realize Skylake just released, but will Intel release another within like a year? I saw a projection that showed this arch coming in 2016, but I believe that was just for mobile devices. When it does drop, is there any reason to believe that it will be anything other than another 10% increase?

So, what do you guys think? Should I wait for Zen, potentially cannon, or just wait for new skylake models and just upgrade then?

 

[R0H1T]

In terms of core count, Zen will have an advantage so unless you meant Skylake-E Zen would be the one to get, especially if you want more cores. As for IPC & per core performance, that's anyone guess atm however I do expect Zen to be a great overclocker so there's that to keep in mind.

 

[desprado]

A smart or a normal person will never buy AMD processors.

Insulting other members is not allowed.
Markfw900

 

[NTMBK]

Forget about getting a new platform, upgrade your current one! There are some awesome Xeons for LGA1366, which go really cheap on eBay at the moment- take a look at this thread (over 100 pages of people overclocking Xeons): http://forums.anandtech.com/showthread.php?t=2335636

A Xeon X5670 has 6 cores, a higher clock speed than your 930, 50% more cache, and runs cooler. (It's a 32nm chip, while your 930 is a 45nm chip.) Drop one of those into your motherboard, overclock it, and forget about upgrading for a few years.

EDIT: And you can get it for $100 on eBay: http://www.ebay.com/itm/INTEL-XEON-...BV7-2-93GHz-/221896421559?hash=item33aa0eb0b7

 

[Arachnotronic]

Forget about getting a new platform, upgrade your current one! There are some awesome Xeons for LGA1366, which go really cheap on eBay at the moment- take a look at this thread (over 100 pages of people overclocking Xeons): http://forums.anandtech.com/showthread.php?t=2335636

A Xeon X5670 has 6 cores, a higher clock speed than your 930, 50% more cache, and runs cooler. (It's a 32nm chip, while your 930 is a 45nm chip.) Drop one of those into your motherboard, overclock it, and forget about upgrading for a few years.

EDIT: And you can get it for $100 on eBay: http://www.ebay.com/itm/INTEL-XEON-...BV7-2-93GHz-/221896421559?hash=item33aa0eb0b7

Bad plan considering that one of the reasons the OP wants to upgrade is dissatisfaction w/ motherboard

@ OP

Just buy a Skylake. Intel won't be putting out Cannonlake until 2H 2017. Forget about Zen; it will not be released until the end of 2016 and although there are certain forum members who think that AMD will deliver the the world's best, no-compromise CPU with Zen (high IPC! 8 cores! Amazing overclocking), I can't help but be reminded of the same "wait for Bulldozer" nonsense that the AMD loyalists used to talk people out of buying perfectly good Sandy Bridge chips back in the day.

Just get a 6600K and a good Z170 motherboard and you will be set for years. I'd honestly recommend the 6700K because you get more cache, much higher base clock, and hyperthreading but if you are dead set on spending no more than $250 the 6600K is a great choice.

 

[Cerb]

#1 DX12 superior threading capabilities
This new API two should potentially see an influx in the current standing of AMD's performance. Current AMD chips won't outright outperform Intel, but it should at least lessen the gap. Particularly, in the cost / performance scale. Right?

Depends on driver implementation quality, and whether or not the game needs one or more fast single threads. Performance of the CPUs will lessen or broaden the gap, in the end. Anything DX12 brings will be brought to both of them equally well, as it concerns the CPUs.

#3 Simultaneous multithreading prevalence in gaming
This seems to becoming far more prevalent in gaming, and AMD is supposedly going to finally utilize this technology in their new zen arch. It's seeming an expensive (ed: ???) feature in Intel CPUs as well. Perhaps this too could help with AMDs price/performance.

SMT is not prevalent (or not) in gaming. It's prevalent (or not) in the CPUs. The games don't know, nor do they care. CMT worked perfectly well in BD, but I would assume that when ironing out Zen, they did not see it as something they wanted to repeat. Since both CMT and SMT are affected by how caches are implemented, no-SMP v. CMT v. SMT was just one of many features that had to be decided on based on many technical details, and we're just going to have to wait it out, to see how it turns out. It's not just the feature, but the implementation of it, that matters.

 

[cytg111]

I don't feel I need an upgrade atm, though I am annoyed at the features of the old MB, but I am really curious as to what you guys think my best options in the near future are.

I dont understand your question.
You dont require an upgrade atm but will so in the near feature. How come? What software or peripheral do you expect to aquire that is gonna move you out of the "atm" zone and into "the near future"?
That being said, IMO, Zen is wild card, a joker, which means that if you're a gambling man, then yes, go ahead and put those factors into your decision making. We will know what Zen is when it lands, not a moment sooner.


#1 DX12 superior threading capabilities
- Make CPU overhead smaller, thus weaker cores appear less weak. Still weaker.

#2 AMD's Zen
- Allready covered

#3 Simultaneous multithreading prevalence in gaming
- I see no direct evidence that SMT rocks gaming, remember the i7's is typically clocked higher and have a larger L2

#4 14 nm tech lithography
- same as your number #2 question sortof?.. other than that, why do I care what vegetables my cpu is cut from.

#5 Cannon lake (or w/e is after intel's skylake)
I'd say that 10% would be pretty damn optimistic .

 

[el etro]

I think you don't need to wait, just get a i7 6700k now. You will love it.

 

[tential]

If you're waiting for zen when you already could use a upgrade I feel bad for you. I wouldn't buy into zen hype. If it delivers great. Amd has yet to deliver on cpu side for years, why you would put any time behind the for cpu is beyond me. I doubt the processor will even come out in 2016.

 

[Arachnotronic]

If you're waiting for zen when you already could use a upgrade I feel bad for you. I wouldn't buy into zen hype. If it delivers great. Amd has yet to deliver on cpu side for years, why you would put any time behind the for cpu is beyond me. I doubt the processor will even come out in 2016.

The only time we should be discussing Zen as a potential buy is when it is out and on shelves.

 

[Yuriman]

Expect Zen in early 2017. Expect Kaby Lake in mid-late 2016, and Cannonlake in mid 2017.

Doing some back-of-the-envelope math, if AMD delivers 40% more IPC than Piledriver (which is what they've stated), we're looking at somewhere around Sandy Bridge to just shy of Haswell performance (depending on the task) at a given number of cores and at given clocks. Whether AMD can hit those IPC numbers while maintaining similar-to-Intel clockspeeds is something to consider.

Here's Skylake next to Sandy and Haswell:

Bear in mind that if AMD can offer similar performance, they're not going to give it away for cheap.

 

[tential]

So I'm expecting AMD to beat it's own performance numbers by a large margin?

No thanks.

 

[UnholyCarcass]

If you are willing to wait then it may be better to wait for Zen but if you want to upgrade now then I would suggest going for skylake

 

[frozentundra123456]

If you are only willing to pay 250.00 for the cpu, then go with a Haswell or Skylake i5. I dont expect Zen in any kind of decent availability until 2017, and *if* it measures up even close to the hype, AMD would be fools to sell it for 250.00. More like 350 to 400 would be what I expect.

 

[frozentundra123456]

In terms of core count, Zen will have an advantage so unless you meant Skylake-E Zen would be the one to get, especially if you want more cores. As for IPC & per core performance, that's anyone guess atm however I do expect Zen to be a great overclocker so there's that to keep in mind.

Any documentation of that great overclocking? Well, it might overclock a lot if they release it stock at 3ghz. But personally, I am expecting AMD to have trouble reaching high clockspeeds with Samsung 14nm, much less great overclocking.

 

[Arachnotronic]

Any documentation of that great overclocking? Well, it might overclock a lot if they release it stock at 3ghz. But personally, I am expecting AMD to have trouble reaching high clockspeeds with Samsung 14nm, much less great overclocking.

It is wishful thinking.

 

[R0H1T]

Any documentation of that great overclocking? Well, it might overclock a lot if they release it stock at 3ghz. But personally, I am expecting AMD to have trouble reaching high clockspeeds with Samsung 14nm, much less great overclocking.

Yes that's what I'm basing the assumption on, Intel HEDT platform (6 cores & more) is not nearly as (overclock) friendly as compared to the FX 8 cores especially the the 8320E & 8370E. Add to that the fact that traditionally AMD has always been a good overclocker, even for the last decade or so, as long as thermals are kept in check. If Zen doesn't run as hot as Bulldozer & it's derivatives, then it's a fair assumption that it would still be good for overclocking & if not overly TDP limited, using the term pretty loosely here, then it can very well do better than Intel HEDT in that particular department.

The final performance, thus, will not just be limited to something like an IPC or per core/thread performance as power users will almost certainly overclock it to the max (with appropriate cooling) & that's the market segment it's targeting.

 

[Arachnotronic]

Yes that's what I'm basing the assumption on, Intel HEDT platform (6 cores & more) is not nearly as (overclock) friendly as compared to the FX 8 cores especially the the 8320E & 8370E. Add to that the fact that traditionally AMD has always been a good overclocker, even for the last decade or so, as long as thermals are kept in check. If Zen doesn't run as hot as Bulldozer & it's derivatives, then it's a fair assumption that it would still be good for overclocking & if not overly TDP limited, using the term pretty loosely here, then it can very well do better than Intel HEDT in that particular department.

The final performance, thus, will not just be limited to something like an IPC or per core/thread performance as power users will almost certainly overclock it to the max (with appropriate cooling) & that's the market segment it's targeting.

Keep in mind that the FX chips were built on a custom-commissioned, high-performance SOI process. Zen will be built on a standard foundry process optimized for mobile processors.

 

[PPB]

Zen node, at least from gf, is already custom tuned. 28nm bulk on kaveri was too.

 

[Arachnotronic]

Zen node, at least from gf, is already custom tuned.

Uh, no. The node was licensed from Samsung -- I doubt that Samsung really had AMD's CPUs/APUs in mind when it was designing it.

Further, AMD said all the way back in 2012 that it would no longer commission "custom" processes and would instead use standard, off-the-shelf foundry nodes in order to keep its costs in check.

 

[PPB]

You can license a node, then tune it within possibilities to your contractor needs.

Off the shelf doesnt mean "not tuned to our best needs". Otherwise Kaveri wouldnt have even reached the clocks it did on the 28nm bulk node GF produces.

Your train of tought is too structured if you think process nodes are unmodificable blackboxes. We dont even know the contract conditions between GF and samsung for licensing to determine if GF is refrained from the slightest modification to suit their own contractor needs.

The licensed node from samsung, if left as is, wont give Zen the clocks it needs (at least 3ghz, ideally 3.5ghz or above), and by the looks of it between iphone's comparisons, it is underperforming compared to TSMC 16FF+. That is why I even said "if done by GF". Zen CPU may as well be made on TSMC. We dont know for sure.

 

[ShintaiDK]

TSMC isn't even using 16FF+ yet, just 16FF that is a 20nm with FF.

If AMD depends solely on Samsungs node, it looks really bad. The performance/watt on the CPU alone looks to be almost 50% better with TSMCs 16FF over Samsungs 14nm.

 

[Arachnotronic]

TSMC isn't even using 16FF+ yet, just 16FF that is a 20nm with FF.

If AMD depends solely on Samsungs node, it looks really bad. The performance/watt on the CPU alone looks to be almost 50% better with TSMCs 16FF over Samsungs 14nm.

The 16nm node used to build the A9 is 16FF+.

 

[Fjodor2001]

The performance/watt on the CPU alone looks to be almost 50% better with TSMCs 16FF over Samsungs 14nm.

Clarification and source?

 

[ShintaiDK]

Clarification and source?

Just look in the Apple thread. Battery run test on Samsung and TSMC. 6 vs 8 hours. (Then add the static usage from the rest of the phone)