Servers
AMD finishes its 'Piledriver' Opteron server chip rollout
AMD has had a wrenching couple of years, and its executives are wrestling with so many transitions in the processor market and inside AMD that they are just punting out the new Opteron 4300 and 3300 CPUs for entry servers without making a fuss with the press or analyst communities. No briefings, no fuss, no muss, here's the …
This topic is closed for new posts.
Excavator?
I was thinking more like 'DDT', 'Tombstone' and perhaps 'Chokeslam'.
No?
It has to fight Intel – and win.
Not sure about that. I reckon they think all they have to do is hold the line until their 64-bit ARM chips ship in volume.
I have a sneaking suspicion that they're looking back at what being first with 64-bit in x86 did for them and betting the farm on repeating it in ARM. If so, you have to suspect that they're on to something. The only thing holding ARM back in servers is the memory addressing limitations. 64-bit ARM could well own the high-density / low power server market.
Big plus: This is something that Intel do not already have half-done on the shelf.......
Re: It has to fight Intel – and win.
For high density those interconnects from Seamicro will help too. If the Fusion processors were already the 'moon shot' I can't think of a suitably more dramatic name for this effort... but I think you're right.
That would be Roadroller. The last steamrollers were withdrawn from service in the late 1960s due to the arrival of UNIX.
Actually its an interesting idea
If you look at the FX chips they don't make it to the top end matching intel, but they come pretty close by using more cores.
That's not helpful for gamers who often need high-performance on a single thread.
However, outside that arena, more cores are probably more useful, since more cores often limit a run-away single-thread's hogging of cpu time, leaving the other cores for other things. Applying that to servers running lots of threads (VDI, JVMs, web serving anyone?) and again, splitting things over more cores may actually help.
For multi-threaded apps, AMD's performance matches intel's not-far-off top end products and does it at a far lower price point. Now, if AMD could just get the thermals under control...
I'm not sure that ARM will save them directly, but it would be nice to see some of the ARM power-saving and process-migration engineering feeding into Opterons. I'd love to have a 16 core desktop which quietly migrates processes onto a low power core so I can leave the thing switched on but running just above hibernation state. It might also be useful for GPU systems: if the monitor has gone to sleep, migrate all critical processes to a lower power core and switch everything else (including the fans) off - we can just pretend that we rendered whatever we have been given.
Saying they have to fight Intel and win is a fallacy. They don't, as ARM is slowly taking on that role. AMD fills a niche market in providing low power, dense pack CPU's at a price point well below Intel. I can buy twice as many CPU's with AMD for my data center and run them cheaper. Their residential chips run laptops nicely and satisfy the OC crowd as well.
AMD is a huge thorn in Intels side and keeps their silly prices from getting much sillier. As long as AMD is around Intel will be forced to keep their prices competitive. No monopoly is good for the consumer.
We have a winner!
As Jobs said all those years ago "Microsoft doesn't have to lose for Apple to win" and the same applies here, if AMD keeps the "bang for the buck" while still making decent profits? Well its a lot easier to sell a low priced machine with more cores than a higher priced one with less, i don't care what market you are talking about.
I do have a question though, will those AM3+ Opterons work in a standard board, or are they server only? Because that 8 core one might be a nice upgrade to my Thuban, when you are doing video editing and transcoding no such thing as too many cores ya know.
But as far as ARM? I really wouldn't bet the farm on ARM, I think its time is just about over. they simply haven't been able to bump up the IPC high enough to keep up with the ever higher multimedia and gaming demands and that when combined with the fact you can only bolt so many cores onto ARM before blowing the power savings has them pretty much in a pickle. mark my words you'll see dual core ARM tablets for $50, because without speed or cores they will have a race to the bottom just like x86, you'll have a few companies like Apple make money in the ARM field but most will lose their shirts. if AMD bets the farm on ARM they may as well close their doors now, they have nothing to offer over Apple and Nvidia in that space. The fact that they are the only other source for X86 chips gives them an advantage, and if they can continue dropping the heat while bumping the IPC without pricing the chips out of their low to midrange market? Then they can carve out a nice niche without having to worry about Intel one way or another.
can't wait
got a 4 core Trinity and it deals much better with 4 Firefox sessions each with 15-20 tabs open than a 4 core i5-2320, and unpacks a large archives faster, on x86_64 GNU/Linux at least
pondering now if I should delay building a new workstation and wait for the new Opterons on AM3+ or get a Vishera
Looks good to me
I think a lot of enterprise will jump on these high performance, low power consumption chips.
building a new amd desktop is a complete waste of time, stick to the 970 chipset and put a fx 8 core in, they forgot to add pcie 3.0 to all newer chipsets and its not in 1070 specs, you cant do all the computer science with cuda, or play games with hd textures and so on
http://www.hardocp.com/article/2012/07/18/pci_express_20_vs_30_gpu_gaming_performance_review/
Indications are that there isn't much difference on an intel platform between pciev2 and v3, in terms of gameplay. If you're that interested, you'll have picked intel over amd anyway.
there is in gt -> gtx cards if you want sli, or use cruddy amd cards and crossfire, and have upto 1866mhz ram the FX core can handle, while all previous ram only can handle 1333mhz
there might not be much different comparing intel chips, pcie 3 would level the benchmark abit better with Fx vs i7
PCIe 3.0 not needed yet
There is no performance advantage to be had with PCIe 3.0 at this time. Current hardware can't saturate the existing PCIe 2.0/2.1 spec.
There is no gain with PCIe 3.0
There is nothing to "level" as PCIe doesn't deliver any performance gain with current hardware.
Re: There is no gain with PCIe 3.0
there is everything to level, 90% of desktop benchmark score are made of of read/write from RAM and HDD
ddr3 ram is :
1333mhz = 8.5 - 10 GB/s
1866mhz = roughly 29.8GB/s
for geforce cards pci-e 3.0 allow keplar GPU, pci-e 2.0 is fermi, pci-e 3.0 doubles the bandwidth, so sli or crossfire setups will be able todo alot more by middle of next year, its not just gamers that use top end graphic card setups
FX onboard memory controller is upto 1866mhz ram, phomem II is 1333mhz
etc etc etc, i7 bridge speed has always been in use as full, AMD chipsets never have, and never will be
This topic is closed for new posts.
