AMD officially announces the Radeon R7 250X, available today

[Scorpus]

Two days ago we got a first glimpse at AMD's entry-level Radeon R7 250X graphics card, designed to slot between the R7 250 and R7 260, and today the company has officially launched it. The card is essentially a re-brand of the Radeon HD 7770 at a new price point of $99, so its specifications shouldn't be too exciting.

The 'Cape Verde' GCN 1.0 GPU packs 640 stream processors, 40 texture units, 16 ROPs and a core clock of 1000 MHz. Memory-wise we're looking at either 1 or 2 GB of GDDR5 running at an effective clock speed of 4.5 GHz on a 128-bit bus. Unlike it's smaller brother, the Radeon R7 250, the 250X consumes up to 95W of power and so requires power input through a PCIe port.

We've heard reports that the R7 250X performs identically to the two-year-old HD 7770, so for a look at the GPU's performance, check out the review we posted at the card's launch. As this card is based on the GCN architecture, Mantle is supported, however TrueAudio is not.

The R7 250X occupies an interesting price point, coming in at the same price as the Nvidia GeForce GTX 650 but performing significantly better. For $10 less you can pick up the R7 250, which is quite a bit slower than the 250X, or for $20-40 more you can grab the R7 260, R7 260X or GeForce GTX 650 Ti.

AMD says the Radeon R7 250X should be available from retailers today, and board partners will already have custom-designed coolers ready to go.

Permalink to story.

https://www.techspot.com/news/55611-amd-officially-announces-the-radeon-r7-250x-available-today.html

 

[misor]

At last (in a local pcbuyer's guide magazine, Philippines; 120php, January - March 2014 issue; available at national book store) the following msi gaming boards based on the latest amd series are already available:
29950php = amd r9 290x 4gd5
22950php = amd r9 290 4gd5 **
16950php = amd r9 280x gaming 3g **

11450php = amd r9 270x hawk **
10950php = amd r9 270x gaming 2g **
9950php = amd r9 270 gaming 2g **

7450php = amd r7 260x 2gd5 oc
5950php = amd r7 250 1gd5 oc
4450php = amd r7 240 2gd3 lp

the listed price is in Philippines pesos at approximately 45 pesos to one u.s. dollar (give or take a few cents)

any listed Philippine price typically includes mandated 12% VAT (value-added tax)

funny thing for us who live in far-flung provinces away from the national capital region (greater manila area): only amd r7 240 is available and not at the suggested retail price.

my favorite computer parts store, pc express, does not yet carry the new amd items.

(NVidia gtx 760 and up are on order basis, 3-4 weeks waiting time but are available right away at selected branches in manila)

** amd series I like and within my budget
( I erroneously posted this reply here: https://www.techspot.com/news/55604-amd-set-to-launch-budget-minded-radeon-r7-250x-next-month.html )

 

[Dukenukemx]

If you can get your hands on a cheap 6850 or 6870, then that would be better then this. What I don't understand is why 256-bit memory isn't standard on everything yet?

 

[theBest11778]

If you can get your hands on a cheap 6850 or 6870, then that would be better then this. What I don't understand is why 256-bit memory isn't standard on everything yet?

Maybe not on sub $100 cards, but 128bit needs to be the minimum. 64bit Memory cards are still around, and for the life of me I don't know why. You'd be better off with an APU because that'll run at least 128bit.

 

[Guest]

I don't really get it. I thought that the built-in gpu of intel and amd processors were fast enough rendering entry level discrete video cards obsolete. am I wrong in this assumption?

 

[Guest]

Pc express has listed the following models based on the latest amd series:
PowerColor AXR7 240 OC 2gb ddr3 = 3730php
PowerColor AXR7 250 OC 1gb ddr5 = 4480php
PowerColor AXR7 260X 2gb ddr5 = 7020php
MSI AMD R9 280X OC 3gb ddr5 384bit = 18360php
MSI AMD R9 270X 2gb ddr5 256bit = 12420php

and its nvidia competitor:
Asus GTX750 1gb ddr5 128bit = 9400php
Asus GTX760 2gb ddr5 256bit = 15820php
Asus GTX770 2gb ddr5 256bit = 24840php
EVGA GTX770 2gb ddr5 SC hdmi = 23170php

 

[cliffordcooley]

Not all systems have the latest and greatest CPU's. Unfortunately there is still a market for these cards.

 

[Jad Chaar]

A 7770 for $100. Not bad.

 

[Dukenukemx]

I don't really get it. I thought that the built-in gpu of intel and amd processors were fast enough rendering entry level discrete video cards obsolete. am I wrong in this assumption?

Not everyone is buying CPUs with built in APUs. If you plan to get a better video card then why get a CPU with built in graphics? As well as the APU's perform, they do share memory and bandwidth with the CPU.

Currently, a FX 6100 with one of these would be a better system then just getting an APU. Also you may want to continue to buy Intel or Nvidia, which doesn't supply an APU option. I don't consider Intel an APU option. APU's require some serious fast ram, so I see that becoming a problem in the future. As it is the move to DDR4 makes less sense to me then to go GDDR or just putting a separate memory slot for video. I can't see people buying 4 memory sticks just to feed a cheap APU. They should make memory with 4 channels built into each memory stick. Not require 4 memory sticks to do the job. This system was introduced back in the Nforce days, and was a band-aid fix for low bandwidth.

 

[Dukenukemx]

Maybe not on sub $100 cards, but 128bit needs to be the minimum. 64bit Memory cards are still around, and for the life of me I don't know why. You'd be better off with an APU because that'll run at least 128bit.

If you plan to use a sub $100, then an APU would be a better solution. Anything over $100 should be 256-bit. It's not a huge price increase for them to do so. Remember when they substituted Radeon 9500's with 9700's, and they disabled the 256-bit? That was over 10 years ago, how long can they keep miking this?

 

[dividebyzero]

A 7770 for $100. Not bad.

There are already "actual" HD 7770's selling for less. Asus and MSI are both $90 after MIR at Newegg, and you can get a pretty decent OC'ed and cooled Vapor-X for $100 ( strangely enough, the base card is the same price, albeit without a MIR involved)

 

[TS-56336]

At last (in a local pcbuyer's guide magazine, Philippines; 120php, January - March 2014 issue; available at national book store) the following msi gaming boards based on the latest amd series are already available:
29950php = amd r9 290x 4gd5
22950php = amd r9 290 4gd5 **
16950php = amd r9 280x gaming 3g **

11450php = amd r9 270x hawk **
10950php = amd r9 270x gaming 2g **
9950php = amd r9 270 gaming 2g **

7450php = amd r7 260x 2gd5 oc
5950php = amd r7 250 1gd5 oc
4450php = amd r7 240 2gd3 lp

I assume the R7 260x is slightly affordable in your part?

 

[misor]

I assume the R7 260x is slightly affordable in your part?

yes, I have 25000php earmarked for NVidia gtx 760/770 or amd r9/r7 (which I have saved by ceasing from alcoholic beverage drinking). unfortunately since I live in the province, I have no choice but to wait or go to Cebu/manila.

I have not gone to Cebu city ion January/February 2014 (where a local pc express branch is located) or to the national capital region (where many pc express branches and its competitor stores are located).

I was in Cebu around October 2013 for my stonogram (computerized tomography scan of the kidneys, ureters, and urinary bladder) for an abdominal problem. the issue is almost resolved after medication and lifestyle change. also been there in November and December 2013 (I bought evga gtx 650 for one of my rigs)

 

[Dukenukemx]

There are already "actual" HD 7770's selling for less. Asus and MSI are both $90 after MIR at Newegg, and you can get a pretty decent OC'ed and cooled Vapor-X for $100 ( strangely enough, the base card is the same price, albeit without a MIR involved)

Re-branding graphic cards is usually to prevent a price drop, not to create one.

 

[cliffordcooley]

Re-branding graphic cards is usually to prevent a price drop, not to create one.

Using the same card eliminates the additional cost in modifying a new card. How is using a re-branded card going to prevent a price drop. Other than making it appear as if it is a new card with a new name?

 

[GhostRyder]

This happens pretty often if the same generation cards are used (Same architecture). GTX 570 was essentially a rebranded GTX 480 when you get down to it same with the R9 280X being a HD 7970 or GTX 770 being a GTX 680. When the rough architecture stays the same instead of changing up and making a new card it can be just easier to rebrand since the previous cards already fit into those categories just fine as it is.

Normally they drop it at least around the price of the previous gens are at or drop them down (In any case for instance the GTX 680 was still 500 at the time when the GTX 770 came out it was 400 and the GTX 680 eventually followed suit). The 250X to me though is not really needed because I had personally hoped that the 250 X was going to be a new card like what was on the R7 GPU inside the 7850k Chip (Which is close in specs and performance to an HD 7750) for a good card to do Hybrid with. However I do not feel this card was in all honestly needed because the price ranges of these cards already fill enough gaps as it is.

 

[Dukenukemx]

Using the same card eliminates the additional cost in modifying a new card. How is using a re-branded card going to prevent a price drop. Other than making it appear as if it is a new card with a new name?

These graphic cards are based on DX11, which has been out since 2009. Most people by now have some sort of DX11 graphics card, which to them is good enough for today’s games. The 7000 series cards were released in 2012, so tech wise they're pretty old.

AMD like Nvidia segregates their cards by stream processors and memory speed. Low ends have 128-bit memory with roughly ~300 stream processors. They don't want the new low end to be 256-bit with ~1000 stream processors. Smart consumers would continue to hold onto their DX11 cards, or buy HD 5000/6000 cards since it makes no difference.

So to confuse consumers, they create a new name for an existing series. This buys time for AMD to wait for a new api standard to work with. Cause neither Nvidia or AMD will make a new graphics card without knowing what the new DX or OpenGL standards are going to be, and DX12 doesn't look like to be happening any time soon. Otherwise these graphic cards would plummet in prices, and would have continued if it wasn't for bitcoin.

 

[dividebyzero]

Low ends have 128-bit memory with roughly ~300 stream processors. They don't want the new low end to be 256-bit with ~1000 stream processors.

AMD and Nvidia would love to have a low end using 256-bit - it's economics that stop it from happening. The uncore ( memory controllers, caches, thread dispatch, I/O) makes up around half -or slightly more, of the overall size of a GPU. Doubling the number of memory controllers and their I/O traces would send yields through the floor on parts that are sold for very little.

Smart consumers would continue to hold onto their DX11 cards, or buy HD 5000/6000 cards since it makes no difference.

That holds true for the mainstream and low end of the performance segments, but becomes largely irrelevant at the top 2 tiers (and dual GPU if in the product stack) since benchmarking, the simple need for the next best thing, and artificially inflated game I.q. options tend to keep the niche markets ticking over at a constant level. The Titan Black Edition will likely be a $300 more expensive 780 Ti with an extra 3GB vRAM and fully enabled 1:3 FP64...it won't stop people buying the card if it outperforms the 780 Ti in some esoteric benchmark at a ridiculous level of image quality at 7680x1440

So to confuse consumers, they create a new name for an existing series. This buys time for AMD to wait for a new api standard to work with

Nah, it's because product launches boost sales from PR and launch reviews. An overabundance of SKUs at $10 (or so) increments both confuses people - bringing out "new" models catches those who aren't particularly tech savvy, and provides a product at every price point a consumer might have set for themselves. AMD now have a lineup that ranges from the R7 240, R7 250, R7 250X, R7 260, R7 260X, R7 265 (imminent), R9 270, R9 270X, R9 280 (imminent), R9 280X, R9 290, and R9 290X. I can guarantee that even moderately tech savvy people would have a hard time naming all the GPUs in that range, and which are new, which are warmed over redesigns, and which are clock-for-clock rebrands.

As for the API argument, it is largely moot. DirectX11 will still be the dominant API when the next gen 20nm GPUs arrive, and likely the second generation of that process node also. DirectX11 hasn't really been implemented in any great depth to date with gaming. Once developers start to actually utilize the spec likely see additional features added that will push the envelope. An example:

[Source]

 

[Dukenukemx]

AMD and Nvidia would love to have a low end using 256-bit - it's economics that stop it from happening. The uncore ( memory controllers, caches, thread dispatch, I/O) makes up around half -or slightly more, of the overall size of a GPU. Doubling the number of memory controllers and their I/O traces would send yields through the floor on parts that are sold for very little.

Didn't stop ATI from selling Radeon 9700's as 9500 with the 256-bit disabled. They did this until they created the 9600. I believe Nvidia did the same thing with the 6800's.

As for the API argument, it is largely moot. DirectX11 will still be the dominant API when the next gen 20nm GPUs arrive, and likely the second generation of that process node also. DirectX11 hasn't really been implemented in any great depth to date with gaming. Once developers start to actually utilize the spec likely see additional features added that will push the envelope.

This can be said for DX10. How many games actually used DX10, or DX10.1 even? But this is because of Xbox 360 and PS3 stuck with DX9.0c graphics. So if you saw any DX10 or DX11 features, it would be additions to the game and not primarily used.

But historically, neither ATI/AMD or Nvidia just creates a whole new architecture unless it's based around a new API. They will continue to increase performance through manufacturing, clock speed, and memory speed. This is why I believe that the 5000/6000 series are the same architecture as the GCN graphic cards that AMD makes today.

As much as people say that vliw4 is different then GCN, I look at the open source drivers in Linux and see that the code between the two architectures is shared. It's not different enough for them to have to write entirely new code. Maybe when there's working OpenCL drivers, then we'll see the differences.

 

[dividebyzero]

Didn't stop ATI from selling Radeon 9700's as 9500 with the 256-bit disabled. They did this until they created the 9600. I believe Nvidia did the same thing with the 6800's.

The fact you had to go that far back to find an example should have been a clue.
Silicon wafer costs are well over triple what they used to be. The 9700 for instance used TSMC's CL015G (150nm) process. Cost per wafer was routinely in the $1000-1500 range...and that is not taking into account the overcapacity that TSMC had at the time. Wafer costs now?
$4000-5000.
Even the budget Radeon 9500 at $179 sold for more than most of the lower end cards of today. Die size for the R300 at 212mm^2 wasn't appreciably larger, and yields aren't that dissimilar either ( TSMC's 150nm process had yields exceeding 80%)

This can be said for DX10...[snip]

All rather tangential to what I posted. Which was a direct response to your assertion:

So to confuse consumers, they create a new name for an existing series. This buys time for AMD to wait for a new api standard to work with.

This series of cards will be long dead and buried before a new API arrives. Pirate Islands will likely have come and gone as well before MS even thinks about adding to the DirectX11.2 spec

But historically, neither ATI/AMD or Nvidia just creates a whole new architecture unless it's based around a new API. They will continue to increase performance through manufacturing, clock speed, and memory speed. This is why I believe that the 5000/6000 series are the same architecture as the GCN graphic cards that AMD makes today.

All rather contradictory don't you think? The 5000/6000 series are VLIW5 and VLIW4...the same base architecture has been in use since R600 (HD 2000 series) - an architecture that was designed to utilize DX10. By your reckoning, AMD should have changed architectures when DX11 arrived - they didn't. The first series to utilize DX11 (Evergreen/ HD 5000) still used the same µarch as it predecessors.