News AMD opens Kaveri A10-7850K pre-orders

This is remarkable marketing-bullcrap they're spilling around there.

The A10-7850k will nowhere come close to the performance of an i5-4670k... not even when saturn, mars and the moon are aligned at midnight. You simply can't count the iGPU-performance, as there's basically no software around that makes any use of it besides the stupid benchmarks.
And those who really need a GPU for their software simply buy a €75 HD7750 1GB GDDR5 and beat the iGPU of the A10-7500k yet again by a mile.

Seriously AMD, your marketing-stunts get worse and worse to make people believe that you're still up to par with intel.

And while we're at it... if you want to count the iGPU than compare yourselves against the Iris Pro chips from intel and see where this goes. The 21" iMac for example has the Iris Pro iGPU.

 

@ jrs77 - i see your point but you need to compare like for like based on pricing not top end against top end.

AMD-7850K will likely be much cheaper than the Iris Pro CPUs. Depending on launch price it HAS to be compared with whatever intel part is in the same bracket.

 

ladies and gentleman - the intel PR rep has spoken :roll:


many productivity programmes now use OCL , and the amd apu`s have already been shown to be faster than all the i7`s for homes use, using those programmes with OCL.

oh so it has a good igpu? lets not use it then shall we , because intel cannot compete!

AMD are correct - the APU`s are a match for anything intel has , when using the igpu. Look at the projected cost of the Gigabyte BRIX with the i5-4750R as an example.

its like saying FREESync isn't worth is because AMD can do it already at no extra cost - but please go any pay extra for G-Sync....

 

The i5-4570R with IrisPro costs exactly $7 more than the i5-4670k and is built into every entry-level 21" iMac.

As stated above, the i5-4570R costs the same as the i5-4670k.

I'm not an intel PR rep at all, but aslong as AMD doesn't manage to deliver a CPU/APU that can deliver the CPU-performance of intels i5-offerings, there's simply no reason for most people like me to buy an AMD APU. Singlethread-performance of the APUs is laughable, which is pretty much critical for all current office-software and rendering-performance in Vray, Cycles, C4D, LuxRender, etc is aswell nowhere comparable with a $160 i5.

So what software would that be that makes an APU that interesting for John Doe?

 

why are you using the cpu in C4D or luxrenderer anyway? GPU only mode is much much faster and that's where the APU`s shine??

and again - VRay RT settings can enable OCL for gpu rendering....

 

There are some machines you can't just pop a HD7750 in, my HTPC for one, it has one expansion slot for freeviewHD tuner, it runs well on an A10-5700, iGPU performance here is king, it already has enough CPU power that I down clock and undervoltage it to 2.5Ghz for power/thermal considerations and so I can up GPU clock speed in preference, it can run a number of games on TV at a reasonable lick and I'd definitely drop one of these Kaveri in if a 512 core <65w chip drops.

The entire upgrade from my old x2 on HD3200 chipset which included CPU/MEM/Mobo was a bargain at ~ £160 I already had the cooler admittedly, don't think Intel could match the price/performance.

 

V-Ray RT is not for final rendering tho, but for live-preview. Currently I don't know of any render-engine used for final rendering that uses the GPU, not C4Ds built-in engine, nor LuxRender, nor V-Ray, as they cannot render complex materials, but only simple matte materials.

But feel free to show me one that I can use with my GPU for final renderings.

OFC there's some scenarios, where these APUs make sense, but usually a small i5 paired with a HD7750 is better and not that much more expensive.

 

I could be getting in a muddle, but im not sure AMD is targeting office or rendering type work loads. From my perspective they are targeting the low to mid range market, the vast majority of the market that want a cheaper do it all system with the possibility of adding a discreet GPU should they need better graphics.

Something like Johnny's first Steam machine that they can slot a graphics card into next Christmas.

 

The vast majority of PCs is used for office and casual graphics and multimedia-editing. That's the market.

 

For rendering ... just buy a caustic accelerator card. It's cheap and efficient . For the rest, I won't buy anything that do not have ECC memory for professionnal use, but that's just me.

 

For professional use you might consider a server with a few more cores as a render-node

 

Going to try and keep up with how these perform, currently use two APUs, one in a media pc I built for the family over Christmas and another (without the iGPU) in my main rig, think they're great for the money.

No idea if they compete with an i5, but then again they don't cost as much so they shouldn't need to.

 

Yup forgot to add those in as well its still what i think AMD chips are ideally suited for.
There is no denying Intel has AMD beaten when it come to shear grunt, but IMHO AMD has them beat on the value front others may (and probably will) disagree with me though

 

Use APUs in a few HTPCs that I have, Would I ever use one in a performance built pc no I would not. More intrested to see what there next FX cpu will do, A more competitive AMD is good for everyone wether id buy one or not is different matter.

Thats AMDs problem getting enthusiasts intrested in there cpus again. For most enthusaists though a A10-7850k is not going to be enough. A top end fx that can compete across a whole range of benchmarks instead of just Multi threaded stuff now that would get people intrested.

Im a stratagy and Race gamer how well it performs in fps games will never effect my purchase decisions. If they show its a wonderful CPU for PA or Starcraft 2. Or some of the race sims I play hell I might even buy one. As it is my aging i7950 is 20-30% quicker at its overclocked speed of 4ghz than any cpu AMD makes in those games.

Value coments are not relivent to me so I dont count them as pros or cons. Im in a position where a budget is not really what I care for in PC performance. Alot of people are in similar positions or AMD would be doing alot better than it currently is.

 

Ecc and server is a Pleonasm ... Server with loads of cores and load of PCIe 16x to drop loads of caustic card in it

 

Prof usage is Tesla cards if your work load supports them. Racks and Racks of Tesla cards thats what most of the top performing computers are using. Backed up by racks and racks of Intel cpus in most cases.

In a business sence ive seen more Tesla / Duel cpu intel builds in recent years than ever before but thats because the software has finally caught up to it. Ecc Memory costs more than the Cpus in most builds.

ECC memory + pro motherboards can be more cost than any GPU or Cpu you can care to list anyway.

 

Lets not for get power consumption. I don't know what idle is, but that Intel i5-4670k is 84w versus 95w for the AMD Kaveri. Lets not include the fact that the Haswell processors have been shown to consume signficantly less power under idle and light load than Ivy Bridge did, which were 77w TDP processors.

So odds are good that the Intel processors and whole systems, with identical hardware configurations other than processor and motherboard, are probably going to consume between 15-40% less power (and likely run quieter as a result). That isn't a big consideration for lots of people, but it is for plenty as well. Example a business where they look at their power bill and think about the fact that an Intel built system might end up costing them $15-30 less per year (both between the machine power consumption and possibly increased cooling costs for the office), per machine on their electric bill, which even if the AMD system is cheaper, means it would need to be at least $15-30 per year of life expectancy of the system cheaper to be worthwhile.

At least if the IT planner is looking at total cost of ownership. That means if the business is thinking of a 3 year desktop life cycle, the AMD system would need to be $45-90 cheaper to be competitive on total cost of ownership.

The AMD systems deffinitely have their place for really low cost systems and for HTPCs in some cases. Outside of that they aren't worthwhile IMHO. Intel's IGP with Ivy was good enough for very casual gamers and most typical users (actually most typical users who don't do anything heavier than something like FB games, FB, email, maybe word, IE and similar would/were happy with Sandy level graphics).

Haswell either what is included in regular desktop graphics now, or better yet the Iris pro graphics (in the even lower power consumption 65w TDP desktop chips, which are barely more expensive than high end core i5) are just fine for most people. Hell, they'd probably be almost good enough for me for most of what I do. The Iris pro is certainly faster than the 5570 GPU I am running right now (that said, I do want to look at getting a 650ti or 7790 or something sometime soon).

So I see AMD chips as something you are going for the cheapest build possible or for select cases where you either need the most compact, highest graphical performance HTPC you need, or a gamer on a very, very strict budget. That doesn't cover a whole lot of buyers/users these days unfortunately.

From a personal perspective, few things I use take advantage of OCL. Some, yes. Photoshop CS6 does, but frankly most things that can take advantage of the GPU see relatively little difference moving from one level of graphics to another. The HD2500 in my i5-3570 performs GPU accelerated filters nearly as fast as my 5570 or the HD4000 in my laptop. The difference from one to the other is at most 1 second even for significant liquify filter changes, for example. One might take 4 seconds on the 5570 and HD4000 and maybe at most 5 seconds on the HD2500.

I am sure there are plenty of OCL workloads where faster graphics will see significant speed up, but CS6 is about the only thing that supports OCL that I use. Yes, browsers have GPU acceleration now, there isn't much of a difference between GPUs though (that I have seen any evidence of). Handbrake mainline build doesn't use OCL and for the beta builds where you can use OCL for decompression the speed up on an overall encode isn't all that large and is signficantly less of an impact on overall encode speed in comparison to having a better CPU in the machine versus a better GPU.

That is pretty much it for what takes advantage of OCL in my experience in my environment.

 

What I really don't understand is that if AMD can get so much performance from the GPU side of their APUs, why are their dedicated GPUs so power hungry? Especially when looking at their lower to mid range cards which seem to be on par with their APU performance.

 

intel power consumption is `SDP` and not real world useage - overheating i7 haswell`s in sealed boxes on stock cooling when they came out.