AMD CPU speculation... and expert conjecture

Page 161 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
^^ intel's driver version is listed.
afaik, new drivers added opencl 1.2 support for both ivb and haswell.
ivb(hd4k) already had advantages in some opencl benches, new drivers improved on that. i dunno how gt3/3e performs... haven't read all the reviews yet.
trinity cpu bottlenecks it's igpu opencl(gpu) benches - shown couple of times. in tom's review, i noticed that fx8350 quite badly choked the titan in the opencl bench. i've had suspicions about this for a while, confirmed it today.
 

hcl123

Honorable
Mar 18, 2013
425
0
10,780


Hmm... its not intel its the users (don't complain of other for your choices). Intel only takes advantage of the preferences it induces by whatever mean(don't fall for it). I think Intel knows that amd medium long term only salvation is "detachment", that is, complete separation of platforms and what not(including much OpenCL apps and games like in never settle), that in amd could came with Hypertransport slots (HTX) for graphics adapters on APUs or CPUs platforms, and even the CPU uarch for AMD could transition to ARM scrapping x86 altogether(should/could be better).

But to do that right now is not possible, without a looot of pain that could prove to be fatal for AMD. Yet the HTX part is perfectly doable, meaning choosing Radeon you must choose FX or APU (which could came with "absolutely no settle" bundles (lol) of games and apps, of which we would not see any bench in the "popular" sites for a long time), it could prove to have tons of advantages, yet those same GPGPU discrete adapters could come in PCIe format for intel platforms for the non believers ...lol.

So i think the picture is on the wall, everything intel or none of it, and if its the first option expect prices to increase even more dramatically. I think intel agrees with me and is betting on the first option, that is why the unusual BGA push...

 

by the same idea (if i understood your post) i shouldn't 'fall for' amd either, right? if amd comes up with their own, closed hard/software ecosystem, won't it rob them off revenue real fast? i don't think introducing hypertransport adapters (i assume you mean it as a replacement bus to connect discreet gfx) will improve the drawbacks within the cpu design (not the buses). the reason amd even exists in the first place because ibm demanded a seperate supplier than intel.... ridding their core business like that would certainly hurt them a lot.
however, the closed ecosystem you seem to speculate may already be here. nvidia is bragging about 90% design wins (laptops) for haswell and amd is coupling their top mobile gpu with richland. i noticed a clear divide occuring between the two.
 


Yes it beats trinity but unlikely to beat Richland, another factor to consider is that for any resolution higher than 768 iris is about a frame faster than AMD's old timers now and then there is more, AMD richland mobile promised 45% over HD4000 so that about puts AMD's mobile on par with Iris while desktop will plunder in HD6670 performance.

I am told that leaked results of prototype Kaveri have its IGPU beating a HD7770 and rattling out more bandwidth.

Anyways I did say that Iris will cost you North of $500 basic and it seems its even higher, DT Iris will be over $600 and frankly its madness for no upgradeable. I am now waiting Mobile and DT trinity to redress the balance.

 

hcl123

Honorable
Mar 18, 2013
425
0
10,780


If games are alright the choke is not bandwidth which most possibly falls backs to drivers... which could be entire fault of AMD. Its true also that VLIW uarch was never much of a compute uarch, but the GFLOP potential is above none, so it should/could be much better.

And all this falls back exactly to my previous post, Radeon on AMD, by some tests floating around seems to be getting(edit) much better than Radeon on Intel, specially comparing the same with Geforce on Intel. Geforce on AMD then, seems to be getting worst and worst, Nvidia seems to be pushing the same trend with its drivers, it wants no competition from amd.

I think its logical that in future no mixes are possible or should be wise or advantageous. And Nvidia is playing a dangerous game, any time intel can shake then off, like they did when they killed their motherboard business.

 

radeon with amd is better than radeon with intel? toms own gaming (and some opencl) benches say otherwise. i saw in some f1 2012, dirt games, far cry etc games radeon cards doing better with intel. however, there are numerous settings, driver versions, game patches, cpu/gpu bottlenecks to consider... too many variables. but mostly intel seems to handle radeons better than amd does, single gpu as well as multi gpu.

 

hcl123

Honorable
Mar 18, 2013
425
0
10,780


No, you should take full responsibility for your choices and expect no one to make the choices for you. No one to blame... no one to accuse...

You want my advice pay me first, pay me well... lol... Isn't that what is WRONG with the paradigma ??

Why would be evil then intel ask $1000 for a Celeron ?

There is no evil when evil can't sustain itself without the complacence of interventionists, specially for some one that knows biasing, that knows becnhmarketing, that should know that any choise make today in IT is *ridiculously* short lived in the assumptions... that then should know the dangers of monopolies, that should know that is how it is even outside IT.

If doubts, at least now you are alerted... what you choose what you care is up to you.

 

hcl123

Honorable
Mar 18, 2013
425
0
10,780


Radeon with Zambezi its truth, Radeon with Piledriver seems to be an inflexion with the very new drivers... gee... how fast assumptions change in IT... don't bet on SR the same...

And the fault is AMD alright, the state of flux, and doubt they would do anything to make me or anyone else right in this matter in any case... but all it takes is to make those drivers coded with XOP and FMA4 where ever possible and advantageous, to make a night and day difference ( intel doesn't support XOP/FMA4 natively, it would execute using microcode)
 

Cazalan

Distinguished
Sep 4, 2011
2,672
0
20,810


Crystalwell is impressive but at that price you'll probably only see them in the MacBooks that are already overpriced to begin with.

Good to see Intel taking graphics more seriously but Crystalwell probably won't be "affordable" until 14nm and 450mm wafers.
 
Failwell.

And given the power consumption numbers... Only high binned parts will be good in mobile. I'm kind of expecting a flop on Intel's CPUs this gen.

Now, to give Intel a little benefit of the doubt, their drivers are immature to say the least and the tests show some improvements in a handful of games used for the early reviews (expected) and the upcoming months will be better; it should at least.

So... When was the SR announcement? haha

Cheers!
 

hcl123

Honorable
Mar 18, 2013
425
0
10,780


(1) AMD has no problem in CPU design, actually IMO in many many features its designs were always better and more scalable than intel. Its implementations are the crux, simply it has not the resources to compete with intel on intel rules and time schedules. About Hypertransport it is the "bus" of HSA, if you didn't know... IOMMU v2 up, including the now HSA standard IOMMU v2.5 is based on Hypertransport, so HTX slots (identical to x16 PCIe slots but better) would only be a logic evolution that most logically could make everything better, most because *every* proccesor from AMD, APU or CPU, is already a Hypertransport controller (switch) in the Xbar (past work of Jim Keller) (dispensing that way the additional translation to PCIe packet, which should have another controller)... only missing HSA games then to make a hell of a difference.

(2) Wise IBM... but if you didn't know, the original 8080 for PCs was "bought" from IBM by intel, including the join development/specking or "advising" on the original 80286, perhaps "another" supplier was included in the deal then...

(3) more of the same and i fail to see any solution or improvement for AMD. AMD problem is not technical never was, its commercial. Right now is in a slow dead mode... HSA and ARM "detachment" might be exactly what the doctor ordered lol

(4) expect more of the trend and more pronouced.
 
HasBurst™ will likely steal most of PD's sales as far as non-APU x86 goes. I'm hoping AMD can roll out a 6-Core Kaveri that is supposedly (from those rumors) pumping out 7830-50/8770 (Better than 7770) iGPU performance and ~FX6350-6550 performance, for around the same price as the 6800K by the end of this year. That could make a killing in terms of iGPU and value for gaming. Remember, GT3e is mostly designed for laptops/ultrabooks with 768/900p screens.
 

jdwii

Splendid


^ Probably doesn't know that the Radeon 8000HD series hasn't came out yet to defeat his beloved titan. But hey if most people use it, it must be better yet just like Windows and Capitalization,

Also check your Tock haswell yet? Probably justifies it in some way.

Also i see nothing wrong with going off topic when their is no info it educates people and does not hurt anything.
 

jdwii

Splendid


At least when Bulldozer sucks i admit it these Intel fanboys won't admit that Haswell was one of the biggest disappoints in years at that company, then they justify it by saying oh its more efficient, funny seems like to me Amd is improving performance with lowering power consumption.
 


+1, and Haswell has 7W more TDP now, so no dice for the power consumption excuse.

 

youcanDUit

Distinguished
Oct 27, 2011
203
0
18,680


the force seems strong with you. i'll ask you this: why/how are the 8-core AMD's, with a clock speed faster than intel's, slower than the lowest i5.
 

Cazalan

Distinguished
Sep 4, 2011
2,672
0
20,810


The 11W lower when idle makes up for that, unless you're F@H crunching 24/7 or something.
 

anxiousinfusion

Distinguished
Jul 1, 2011
1,035
0
19,360


I would gladly pay more than 140 dollars for a chip like that.
 


For desktop? I would challenge that, but since most of office computers are mostly IDLE or turned off, I don't know to be honest.

Now, on the other hand, that increase in power usage is due to the GT2.5 part, right? CPU wise, it doesn't seem to be much more power hungry. At least, it's pretty much equal to IB. OC is another topic though, but that area, for the folks who read Tom's and most other tech sites, will have to agree that it's a major blow. I mean, although AMD screw up with BD, it was at least a good OC'er (give or take on that assessment) and PD really improved in that area (AFAIK). IB to HW, it's backwards. Its like Intel is giving us the finger so hard, it's painful and infuriating. I will concede that there's the extreme(ly expensive) platform for the 1%, but come on...

And on that last point... Hell, looking at "features" on today's MoBos, I'm really wondering why no one has ever though about making smaller boards (hence cheaper) for gaming only. I mean, looking at most of the gaming hardware from friends and people around, they have:

- OC'ed CPU
- 8GB RAM (2 sticks being the norm; 16GB in 4)
- 1 video card (2 being very uncommon, unless big discount for the second one)
- USB things (mouse, keyboard, joysticks, etc)
- 2x 1080p 60Hz or 1x 1080p 120Hz monitor (yeah, most of my friends and fellow gamers I know have at least a second screen)
- RAID0 or SSD + big capacity HDD (or 2).

If you think about it, do we need the bazillion sATA adapters or the absurd amount of PCIe x16 adapters? AMD and Intel could cut down features so partners give us a more gaming oriented design (hence, smaller) that can be put in a smaller case. I really love my 932, but it has wheels. The thing weights a friggin TON and it even looks empty for crying out loud.

Oh well, sorry for the rant, but SFF FTW OMGBURGER.

Cheers!
 

anxiousinfusion

Distinguished
Jul 1, 2011
1,035
0
19,360


You bring up a very good point, here. Since the industry has decided it's going to get a hard-on for mobility, I've decided not to try and fight it. Rather, I will attempt to roll with the punches and leverage the advantages of this new focus. That means my emphasis in new builds will center around SFF chassis, low noise and minimalistic design.

But for this to be easy, manufacturers need to start rolling out better mITX boards just like you said. I would also like to see better standardization in SFF power supplies. Regardless, a mini ITX gaming desktop is much more realistic these days thanks to the developments brought about by the industry's newfound focus on mobile computing. ...But that is all I have to say about mobile that is good.
 

GOM3RPLY3R

Honorable
Mar 16, 2013
658
0
11,010


I just think its funny how you guys are ranting about how good this next series will be. It's probably going to be about the same as anything else. An 8970 would be a slightly improved 7970 and be about on par with the 780, at most getting just under the Titan. I'd be surprised if they made something stronger. The only problem is power consumption. You can run something with a 750w PSU, and with Intel parts that take less power you can get more performance. Yeah it may not be much, but you can probably overclock Intel parts even more and get no more than the wattage AMD does.
 

amdfangirl

Expert
Ambassador


The added TDP is from the integration of power circuitry. Haswell has lower idle but higher load power - attributed to the various Haswell improvements.

55329.png


55330.png


Given how much weighting I give to idle power consumption, I believe I'll be one of the few Haswell advocates. xD
 

amdfangirl

Expert
Ambassador


What I do understand is why mITX costs more than mATX. It's silly really. I wouldn't buying a stripped down mITX board without all those extras gamers don't need - Bluetooth, WiFi, Integrated display ports etc.

By all means have them for higher end mITX boards, but I want a high performance mITX board. With a nice mITX case - ie. not a Bitfenix Prodigy (which is ITX in name only). :p
 
Status
Not open for further replies.