Tuesday, July 28th 2015

AMD Readies Radeon R7 370X to Counter GeForce GTX 950

AMD is reportedly giving final touches to the Radeon R7 370X, to preempt launch of NVIDIA GeForce GTX 950. The card will be based on the "Trinidad XT" silicon, and will max out components physically present on the chip. This means that the card will feature 1,280 Graphics CoreNext stream processors, 80 TMUs, 32 ROPs, and a 256-bit wide GDDR5 memory interface, holding 2 GB or 4 GB of memory. Leaked screenshots that disclose these specs suggest that AMD will carry over clock speeds from the R9 270X, working out to 1180 MHz core, and 5.60 GHz (GDDR5-effective) memory, working out to a memory bandwidth of 179.2 GB/s.
Sources: VideoCardz, Expreview
Add your own comment

34 Comments on AMD Readies Radeon R7 370X to Counter GeForce GTX 950

#26
GorbazTheDragon
HumanSmoke28nm is a pretty mature process - much more so in the 3.5 years since its debut. Nvidia's own silicon has shown similar boosts ( A1 vs B1 silicon GTX 780's for example)
NV clocks have scaled far better than AMD...
Posted on Reply
#27
arbiter
GorbazTheDragonIf you look at all of AMDs cards they are rubbish overclockers when compared to maxwell... NV's cards are all pushing steadily towards the 1.5GHz mark out of the box...
I'm quite curious as to what exactly is allowing those chips to be much faster
Reason is, Nvidia from what seen clock their cards pretty low of what they CAN do. Reason AMD cards don't overclock is so cards are competitive they clock them as high as they can. If Nvidia really wanted they could factory clock chip 1300mhz which all of them can do, most can do least 1400.
My evga gtx980 SC ACX 2.0 does 1442mhz right outta the box with no overclock.
HumanSmokeWouldn't make too much difference from a marketing point of view. Tahiti - like Pitcairn/Curacao/Trinidad, isn't overly DX12 compliant....and according to many AMD devotee's, DX12 is going to be the companies graphics performance saviour. Pretty difficult to make a case for DX12 being a prime reason to choose the Fury/Fury X, but dismiss DX12 for a card where the features would provide a solid performance leap.
Part of DX12 those cards really need they should support as ALL DX11 cards are said to be able to do speed which is likely to be what takes hold first then any graphic sides. Reason they go with it being savior is their cpu side more then gpu side.
Posted on Reply
#28
GorbazTheDragon
arbiterReason is, Nvidia from what seen clock their cards pretty low of what they CAN do. Reason AMD cards don't overclock is so cards are competitive they clock them as high as they can. If Nvidia really wanted they could factory clock chip 1300mhz which all of them can do, most can do least 1400.
My evga gtx980 SC ACX 2.0 does 1442mhz right outta the box with no overclock.
My point is that NVidia silicon gets to higher clocks than AMD stuff. Could be that Maxwell is a better architecture when it comes to stability at high clocks, however considering how early Kepler was quite a lot slower than the last ones, it could be that they just have better silicon.
Posted on Reply
#30
GorbazTheDragon
Higher junction temperatures reduce stability as leakage currents are higher. IIRC maxwell is somewhat more dense than Kepler (transistor density) which would mean that actual TJs are even higher than the measured temps at sensors (temp delta between Junction and sensor is higher). So a possible theory would be that the transistors are reaching higher temps at lower volts due to higher power density, this is also consistent with the evidence that extreme cooling is relatively unaffected, because TJs in those cases are sub-0. (IIRC typical TJ can be as high as 150c for normal silicon transistors, measured temps are lower because sensors are somewhat further from transistors and do not dissipate power themselves)

Another possibility is that due to the larger size of the chips they are more vulnerable to ripple, reducing stability at high power loads, however this is partially inconsistent with the evidence since extreme cooling is AFAIK not really affected from this scaling standpoint.

Either way, my original point still stands, late NV 28nm clocks higher at normal voltages than early NV 28nm and most AMD 28nm.

/rant
Overclocking margins are irrelevant IMO, what matters is the actual clock speed... Remember when the Titan was released and people were all gawking about how it went way faster than the stock 700-800ish MHz clock, then came the 780 and everyone realized GK110s could easily hit 1000-1100MHz out of the box. Or what about the 4790k, a lot of people thrashed it because it didn't have overclocking headroom. Oh it only gets 10% faster than stock clocks... 4.8GHz is 4.8GHz, and that was faster than 4770ks seemed to average for similar volts, but "oh no the 4770k has so much more headroom"... Oh yeah the stock clock is like 3.9GHz turbo... well 20% gets it to ~4.7GHz, more "headroom..."

/rant
Posted on Reply
#31
john_
From the thread about the new 0.8.5 GPU-Z
Added support for AMD R9 370, R9 370X
Posted on Reply
#32
P-40E
Wow! Just WOW! The third release for the tired old 7870! LET IT DIE ALREADY AMD! at 1180mhz factory clock some cards are going to have horrible artifacts because some Pitcairn/Curacao and now Trinidad chips could not go past 1100mhz without crashing or artifacts! Yeah it is a great old card! And it is still decent enough to run most games at decent settings, But the dam thing has been released twice already! I am glad I bought a GTX 970 this year! I do agree the Pitcairn Pro chip does very well when clocked past 1100mhz, My old R9 270 did well at 1150mhz, But there will be no overclocking room at all! There is no value in a GPU that you can't overclock!
Posted on Reply
#33
P-40E
arbiterReason is, Nvidia from what seen clock their cards pretty low of what they CAN do. Reason AMD cards don't overclock is so cards are competitive they clock them as high as they can. If Nvidia really wanted they could factory clock chip 1300mhz which all of them can do, most can do least 1400.
My evga gtx980 SC ACX 2.0 does 1442mhz right outta the box with no overclock.
The reason AMD cards don't overclock well is because they keep re-badging them as new cards with higher clocks! The 7870 was a good overclocker!
Posted on Reply
#34
P-40E
the54thvoidIt's giggles and all to think the term 'readying' is applied to a recycled 7870. Don't get me wrong, great little card but yeah, old tech being smothered in lipstick and given a latex red dress with a plunging neckline. C'mon AMD, you should have delivered more options on Fiji.
Still, Nano is still to show so perhaps innovation (of sorts) there will he a bit more appealing.
I was thinking the same thing! How long does it take to re-flash the bios on old 7870's and R9 270/270X cards? LOL!
Posted on Reply
Add your own comment
Jul 30th, 2024 14:26 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts